CN108604940B - optoelectronic switch - Google Patents

Abstract

The present invention provides a switch module for use in an optoelectronic switch, the switch module having: a client portion for connecting to an input device or an output device; a first component (fabric) and a second component each for processing signals and communicating with other switch modules, the first component having a transmit side and a receive side, the transmit side having: a transmit side input for receiving a first electronic signal carrying information, the information comprising information about a destination switch module of the first electronic signal, the first electronic signal being received from an output or input device of the second component part via the client part; a transmission-side conversion means for converting the first electronic signal into a first plurality of optical signals containing the same information; a transmit side multiplexer for converting the first plurality of optical signals into a multiplexed fabric output signal for transmission to an active switch, and the receive side having: a receive-side demultiplexer to receive a multiplexed fabric input signal from an active switch and to separate the multiplexed fabric input signal into a second plurality of optical signals; a receiving-side converting means for converting the second plurality of optical signals into second electrical signals; and a receive side output for sending the second electronic signal via the client portion to a transmit side input or output of the second component portion.

Description

optoelectronic switch

CROSS-REFERENCE TO RELATED APPLICATIONS

This application claims priority to and the benefit of US Provisional Application No. 62/251,572, filed November 5, 2015, entitled "Optical Switch Architectures," the entire contents of which are incorporated herein by reference.

technical field

One or more aspects of embodiments in accordance with the present invention relate to switch modules usable in optoelectronic switches, and also to optoelectronic switches incorporating the switch modules.

Background of the Invention

The current and continuing increase in data traffic and data center requirements for switching speeds and reduced energy consumption have resulted in a large number of recent innovations. In particular, it has been recognized that optical switching provides many desirable properties, but that optical devices need to be controlled by and used in conjunction with electronic devices including conventional electronic data servers.

The optics themselves do not necessarily reduce the size or complexity of the switch. In order to improve flexibility in assembling and applying optical switching units, it is desirable to improve the scalability of optical switches. An improved method involves the topology of components within a switch network. It is desirable to produce highly scalable optical switching units. Therefore, there remains a need for a packet switch that optimally benefits from the speed of optics and the flexibility of CMOS electronics assembled in an architecture suitable for enormous scalability.

In order to most clearly describe a network topology, such as a computer network or an optically switched network, as in embodiments of the present invention, the following terms and notations may be used:

A graph G is a set of vertices V and a set of edges E, which connect pairs of vertices. This graph can be expressed as G=(V,E). Thus, networks can be modeled as graphs in which nodes (ie, individual switching elements) are represented by vertices, and the links between pairs of nodes are graph edges.

The physical topology of the network is the location in a true 3D space of nodes and links. The logical topology of a network is mathematically represented as a graph of the network G=(V, E).

The base R of a single switch element is the number of ports on that switch element. Switch ports can be client ports (connected to external clients such as hosts or servers) or fabric ports (connected to other switching elements), or unconnected. Number of client ports per switching element = C, and number of fabric interfaces per switching element = F.

A path is a sequence of links connecting a source node to a destination node, and the length of the path is the number of links in the sequence. The minimum path between two nodes is the path with the shortest length, and the diameter of the network is the longest minimum path between any two nodes. Switching elements in a switch may be arranged in N dimensions (also referred to herein as tiers).

A known named network topology is the folded Claus network. This is a common topology currently used in data center networks and multi-chip switches. The topology is also referred to as a k-ary n-tree. The network can be described with just R and N:

C _total = total number of client ports, i.e.

Diameter, D=2(N-1)

Table 1 below shows the value of N, the number of client ports for various values of the parameter, which, as described above, indicates the number of external clients connectable using such a network with a given parameter.

Table 1: Values of N for folded Claus networks with varying values of R and N.

Figures 15A to C show examples of folded Claus network arrangements with different values of N and R, and their corresponding values of _Ctotal , P and D. Obviously, when used in an actual switching network, the number of switching elements used is much greater than that shown in these figures. In these examples, client ports are represented in each case by unconnected links on the bottom row of the switch element. In this document, the term "leaf" may be used for switching elements connected to both clients and other switching elements, and "backbone" may be used for only those switches connected to other leaves. In this application, the terms "backbone" and "active switch" are also used interchangeably.

Summary of Invention

Most generally, embodiments of the present invention provide an optoelectronic switch having a plurality of switching elements organized using an improved (physical) network topology that yields great scalability improvements. In order to realize an optoelectronic switch using the improved physical topology, the switching elements that make up the optoelectronic switch must have specific characteristics, and specific connection capabilities. In particular, switching elements acting as leaves, referred to herein as "switch modules", may require certain internal components for optimal execution. Accordingly, a first aspect of embodiments of the present invention provides a switch module for use in an optoelectronic switch, the switch module having:

a client part for connecting to an input device or an output device;

a first fabric portion and a second fabric portion each for processing signals and communicating with other switch modules, the first fabric portion having a transmit side and a receive side,

The transmission side has:

a transmission-side input for receiving a first electronic signal carrying information including information about the destination switch module of the first electronic signal via which The client part of the description is received from:

the output of the second fabric portion, or

input device;

a transmission-side conversion member for converting the first electronic signal into a first plurality of optical signals containing the same information;

a transmit-side multiplexer for converting the first plurality of optical signals into multiplexed fabric output signals for transmission to an active switch, and

The receiving side has:

a receiving-side demultiplexer for receiving the multiplexed fabric input signal from the active switch and separating the multiplexed fabric input signal into a second plurality of optical signals;

a reception-side conversion member for converting the second plurality of optical signals into second electronic signals, and

a receiving-side output for sending the second electronic signal via the client part to:

the transmit-side input of the second fabric portion, or

output device.

To avoid confusion, it should be noted that in the following, the term "fabric part" is used to describe the fabric port itself, ie the interface between the switch module and the network fabric (link between switch modules), and all the associated parts. Similarly, the "client part" is used to describe the client port itself, ie the interface to external clients, and all associated components within the switch module. Input devices and/or output devices may refer to external clients such as servers or hosts.

Preferably, the second fabric portion is configured to perform the same optical and electronic processing as the first fabric portion, such that, for example, when the second electronic signal is output from the first fabric portion The same processing can occur on the electronic signal when the terminal is sent to the input terminal of the second fabric portion, so that the electronic signal is transmitted elsewhere. In this way, the switch module can act as an intermediate switch module (or backbone), wherein the data received is not forwarded directly to the output device, but to another switch module for subsequent further transmission to another switch module, or transmission to the output device. Accordingly, the second constituent part may comprise:

a transmission-side input for receiving a second electronic signal carrying information, the information including information about the destination switch module of the electronic signal, the second electronic signal being sent via the client The end part receives from:

the output of the first component part, or

input device;

a transmission-side conversion member for converting the second electronic signal into a third plurality of optical signals containing the same information;

a transmission-side multiplexer, the transmission-side multiplexer is configured to convert the third plurality of optical signals into multiplexed fabric output signals for transmission to the active switch;

a receiving-side demultiplexer for receiving the multiplexed fabric input signal from the active switch and separating the multiplexed fabric input signal into a fourth plurality of optical signals;

a reception-side conversion member for converting the fourth plurality of optical signals into third electronic signals, and

a receiving-side output for sending the third electronic signal via the client part to:

the transmit-side input of the first fabric portion, or

output device.

The switch module may also include more than one client section, and preferably includes two client sections. Having an increased number of client sections on each switch module increases the number of external devices that can be connected to each switch module when the switch module is used in an optoelectronic switch.

The switch module may comprise more than the first and second fabric parts, depending on the dimensionality of the optoelectronic switch in which the switch module will be used. Thus, the output of the second fabric portion may alternatively be configured to send a signal to a third fabric portion instead of the first fabric portion.

A switch module according to the first aspect of the present invention provides the functionality required to construct a scalable multi-dimensional (ie N>1, to use terminology introduced in the Background section of this application) optoelectronic switches capable of converting Optical signals received at the client portion of one switch module are transmitted to the client portion of the other switch module. The conversion associated with the transmit/receive side conversion means allows most of the data transfer to take place in the optical rather than the electronic domain. Thus, it is possible to transmit data over long distances with lower power at high data rates and with lower power losses than in the electronic domain. Additionally, using the optical domain enables the use of wavelength division multiplexing. Another important advantage of using the optical domain during active switching is bit rate independence, where switch plane data operates at packet rate rather than bit rate.

As mentioned at the beginning of this chapter, the optoelectronic switch of the present invention (of the second aspect) uses a new topology that provides relative to the aforementioned folded Claus technique and other topologies known to be used in optical switching networks several improvements. In general, a second aspect of the present invention provides an optoelectronic switch comprising an array of switch modules and active switches arranged in an improved physical topology. The switch modules are those according to the first aspect of the invention, which switch modules comprise both client ports ("client part") and fabric ports ("fab part") and are therefore connected to optical Fabric and any external clients.

Accordingly, accordingly, a switch module according to the first aspect of the present invention may be interconnected to an optoelectronic switch according to the second aspect of the present invention, wherein the N-dimensional optoelectronic switch for transmitting optical signals from the input device to the output device comprises multiple interconnected switch modules according to the first aspect of the invention, wherein:

The switch modules are arranged in an N-dimensional array, the i-th dimension has size Ri ( _i =1,2...N), each switch module has a position giving its position relative to each of the N dimensions The associated coordinate set of ;

Each switch module is a member of N sub-arrays S _i , each sub-array S _i includes R _i switch modules that differ only with respect to the coordinates of the position in the i-th dimension, and each of the N sub-arrays associated with different dimensions;

each of the switch modules is configured to generate a multiplexed fabric output signal,

Each sub-array S _i also includes an active switch with R _i inputs and R _i outputs,

Each input of each active switch is configured to receive a multiplexed fabric output signal from each of the R _i switch modules in the sub-array,

The active switch is configured to, based on the destination information contained in the first electronic signal received at the transmit-side input of the switch module, multiplex a fabric output signal therefrom. Any of the R _i inputs leads to any of its R _i outputs, the active switch receiving the multiplexing fabric output signal from the switch module.

Signals sent from each of the _Ri outputs may form the multiplexed fabric input signal that may be received by another of the _Ri switch modules within the sub-array.

Here, the "size" Ri of the _i -th dimension is most easily understood by considering, for example, 120 switch modules organized into a 4x5x6 array. Therefore, R ₁ =4, R ₂ =5, and R ₃ =6. In other words, the size of the ith dimension can also be viewed as the length of the array in the direction associated with that dimension. It must be emphasized that this does not mean that the modules are physically arranged in eg a 3D array, it only represents the connections between switch modules, as will be described in more detail below. This will be apparent from the fact that switch modules can eg be arranged in a 5D array which is clearly not achievable in real space. In the arrangement as described above, it will be appreciated that the size R _i of the _i -th dimension is the same as the number of inputs/outputs of the active switches associated with the sub-array Si in which the coordinates change. In this way, active switches can be guaranteed to be connected to all switch modules in the sub-array.

The overall interconnection mesh formed between the fabric portions of all switch modules and including active switches may be referred to as an "optical fabric" or "switch fabric" and includes optical links connecting the various components. Preferably, the optical link is an optical fiber. Also, the optical links are preferably bidirectional, which can be achieved by bundling two or more optical links within a single cable. Alternatively, the optical link may be in the form of an optical polymer waveguide embedded in eg a PCB or a silicon waveguide formed on or in the substrate. Similarly, an "active switch" refers to a type of switch that is capable of actively controlling the path a signal travels through. Thus, active switches can provide full mesh connectivity without necessarily providing a full mesh of interconnecting fibers, or similar fabrics. Furthermore, if an active switch simultaneously receives _Ri different multiplexed signals at its _Ri inputs, each with a different intended destination, then the active switch is able to transmit all of the signals simultaneously. Active switches preferably operate in a non-blocking manner, and more preferably operate in a strictly non-blocking manner rather than a rearrangeable non-blocking manner. The active switches described herein perform the functions of the "backbone" described above because the active switches are only connected to switch modules, not external client devices.

In an illustrative example, where each active switch in the entire array has the same cardinality R (ie, each active switch in the entire array has R inputs/outputs, regardless of the dimension/sub-port associated with the active switch arrays). These are the base R and the number N of dimensions of each switch module. Thus, using the notation as used in the Background of the Invention section, the following relationship applies:

Leaf, i.e. the total number of switch modules, P ₁ =R ^N

Backbone, i.e. the total number of active switches,

D=2N

Table 2 below shows _the value of Ctotal for exemplary values of R and N, and shows that the number of clients that an optoelectronic switch according to the present invention can support is greatly improved over a folded Claus network. It should be noted that this improvement is noticeable even if the cardinality of switches in different arrays is not identical. However, for manufacturing reasons, it is preferred that all active switches used in the optoelectronic switches of the present invention are identical or substantially identical.

Table 2: Values of C _total for the optoelectronic switches of the present invention for varying values of R and N

In a 2D embodiment of the invention, an optoelectronic switch comprises an array of interconnected switch modules according to the first aspect of the invention arranged in an X x Y array, the array having X columns of Y switch modules, and Y rows of X switch module. Thus, each of the Y rows has an associated active switch (with X inputs/outputs), and each of the X columns has an associated active switch (with Y inputs/outputs). Each active switch provides connectivity between all active switches in its associated row or column. In this case, it can be seen that each switch module is connected to two different active switches, namely the active switch associated with its column and the active switch associated with its row.

Therefore, switch modules used in such optoelectronic switches each require two fabric parts.

Similarly, in a 3D embodiment of the invention, an optoelectronic switch comprises an interconnected array of switch modules according to the first aspect of the invention arranged in an X x Y x Z array, the array having:

X columns (each column consists of a Y x Z array),

Y rows (each row consists of an X x Z array) and

Z layers (each layer consists of an X x Y array).

Thus, each sub-array of switch modules has an associated active switch, each of the switch modules being located, for example, in the same column/row, but at a different level. Thus, there are X x Y active switches associated with a sub-array containing Z switch modules that differ only in layers, each active switch having Z inputs/outputs. Also, there are X x Z active switches associated with a sub-array containing Y switch modules that differ only in row, each active switch having Y inputs/outputs. Also, there are Y x Z active switches associated with a sub-array containing X switch modules that differ only in columns, each active switch having X inputs/outputs. Similar to the 2D case above, each switch module is connected to three different active switches, the active switches associated with each of the sub-arrays described earlier in the paragraph. Therefore, in the 3D implementation of the optoelectronic switch according to the present invention, three fabric parts are required for each switch module.

It can be seen from the above that it is preferred that each switch module has at least N fabric parts, each of the N fabric parts being associated with a different sub-array of which the switch module is the sub-array a member of. Thus, via the dedicated fabric portion for the sub-array, the switch module is able to communicate data to any other switch module in the sub-array via the active switch associated with the sub-array. After the optical hopping has occurred, the signal reaches a different switch module that is a member of a set of sub-arrays different from the first sub-array, and the same processing can then be performed to send the data to another switch module with the sub-array in common. In this way, all data transfer from one switch module to another switch module can take place in a series of optical and electronic hops.

In this arrangement, it is possible to send data from any switch module in the array to any other in the array through up to N optical hops (where optical hops are hops involving signals passing through an optical fabric via an active switch) switch module. This is possible because individual switch modules can act as intermediate switch modules, ie because the output of a first fabric part can send electronic signals (eg data packets) to the input of another fabric part on the same switch module terminal, and correspondingly, the input terminal of the first fabric portion on the switch module can receive data from the output terminal of the second switch module. Electronic signals may be transferred between the two fabric parts using an integrated switch, such as an electronic crossbar switch, or to provide connections between two fabric parts, two client parts, or one fabric part and a Electronic shared memory switch for connections between client parts. Thus, during a data transfer operation, data can perform optical hopping to another fabric portion located in the same sub-array via an active switch associated with the self-sub-array. The data can then perform an electronic hop through the switch module itself to a fabric portion associated with a different sub-array, then a second optical hop can occur - the process is repeated up to N times until the data packet reaches its final destination, That is, the switch module with the client part through which data (eg, in the form of packets) is transmitted to the output device.

With more than N fabric sections on each switch, flexibility is provided for extending optoelectronic switches into higher dimensions. As an example, consider the case of a 2D optoelectronic switch with ^M2 modules organized into a square array. This situation can be extended by connecting spare fabric portions on switch modules with the same row and same column in each (newly defined) tier via new active switches to define new sub-arrays and associated active switches into a 3D optoelectronic switch with M ³ switch modules organized into a cube array (ie, N layers of N ² switches). Excess fabric ports can also be utilized by providing more than one fabric section to connect one switch module to another switch module within the same subarray.

The transmit-side conversion member of each switch module preferably further includes a transmit-side packet processor configured to receive the first electronic signal in the form of a packet containing the destination information. The packet header of the packet, preferably in the form of a raw packet. Except for the data itself. The information included in the packet may include information about the destination of the packet, eg the part of the client to which the packet should finally be sent. The packet header may also include various pieces of information including source and destination addresses, packet length, protocol version, sequence number, payload type, hop count, quality of service indicators, and other information.

The transport-side packet processor may be configured to perform packet fragmentation in which data packets having the same destination switch module are arranged into frames having a predetermined size, and wherein the data packets may be divided into a plurality of corresponding Multiple packet fragments in a frame, and where optionally a frame may contain data from one or more data packets. Each packet fragment preferably has its own packet fragment header that includes at least information identifying the packet to which the packet fragment originally belonged, so that the packet can be reconstructed after subsequent processing and transmission. As an example, consider the case where the packet handler is configured such that the frame payload size is 1000B, and three packets of 400B, 800B, and 800B are input into the switch module. If each of these packets would be sent one packet at a time in separate frames, this could represent an efficiency of (400+800+800)/3000=67%. However, by using packet slicing, the first frame may include the 400B packet, and 600B of the first 800B packet, and then the second frame may include the second 800B packet and the remaining 200B of the first 800B packet. This yields 100% efficiency. Frames constructed by this process represent data packets by their own ability, and thus, when packets undergo more than one optical hop in order to reach the destination switch module, further fragmentation can take place at intermediate switch modules.

To maximize efficiency, subsequent processing of the frame (eg, forwarding the frame for conversion into the first plurality of optical signals) may not be performed until the fill ratio of the frame reaches a predetermined threshold, preferably greater than 80%, more preferably Greater than 90%, and most preferably 100%. After a predetermined amount of time has elapsed, packets may be sent alternately for subsequent processing. In this way, if packets for a given switch module stop reaching the packet processor, frames that are still below the threshold fill ratio can still be sent for subsequent processing, rather than stalling the packet processor. The predetermined amount of time may be between 50ns and 1000ns, but is preferably between 50ns and 200ns. Most preferably, the time interval is approximately approximately 100 ns. Accordingly, the transmit-side packet processor may include or be associated with a transmit-side memory for temporarily storing incomplete frames during frame construction. The elapsed time can vary depending on the communication needs; generally, the higher the rate of communication traffic, the shorter the elapsed time, and the lower rate of communication traffic can result in increased time intervals.

When the packet processor is configured to perform packet fragmentation, the receiving-side conversion means preferably further includes a receiving-side packet processor configured to disperse the original packet among more than one The original packet is reconstructed from the packet fragment while in frame. This operation can be performed with reference to the packet fragment header described above. As a packet undergoes several separate fragmentations by successive intermediate switch modules on its journey from source to destination, the final reassembly of the packet by the receiving-side packet processor may be delayed until the original packet's All components have reached the destination switch module. Accordingly, the receiving-side packet processor may include or be associated with a receiving-side memory for temporarily storing the constituent parts.

The transmission-side conversion member may comprise a modulator configured to receive light from the light source, and more preferably, a plurality of modulators, preferably light modulators. The light modulator may be a phase or intensity modulator, such as an electroabsorption modulator (EAM), a Franz-Keldysh modulator, a quantum-confined Stark effect based modulator, a Mach-Zeng A Mach-Zehnder modulator, and the plurality of modulators preferably includes 8 modulators. Each modulator may be associated with only a single light source or may be illuminated by fewer light sources, where the light sources are shared among the modulators. Each modulator may be configured to receive an electronic signal from the input or the transmit-side packet processor, and unmodulated light from a light source. By combining the electronic signal and unmodulated light, the modulator generates a modulated light signal that has the same wavelength as the unmodulated light from the light source and that carries the information carried by the original electronic signal. This modulated optical signal may then be transmitted to the transmit side multiplexer. The light source is preferably in the form of a laser in order to generate a substantially monochromatic light beam limited to a narrow wavelength band. To minimize losses, the modulator is preferably configured to receive light with wavelengths in the C-band or L-band of the electromagnetic spectrum, ie 1530 nm to 1625 nm. More preferably, the light has a wavelength within the C-band or "erbium window", with a wavelength of 1530 nm to 1565 nm.

The laser can be a fixed wavelength laser or a tunable laser. In an array of modulators, the light sources associated with each modulator should have different wavelengths and have non-overlapping bandwidths in order to minimize crosstalk in the multiplexer. When the light source is a laser, the modulator may be in the form of an electro-absorption modulator (EAM), which uses a varying voltage to modulate the intensity of the laser to carry the information contained in the electronic signal. Using EAM means only changing the intensity of the laser, not the frequency, and thus prevents any change in the wavelength of the modulated optical signal.

When multiple modulators are present, the transmit-side packet processor may also be configured to perform packet slicing, wherein a frame (as constructed by the packet slicing process described above) or data packet is sliced into a first plurality of electronic signals. Each of the first plurality of electronic signals is then sent to a different one of the plurality of modulators, whereby the electronic signals are converted into a first plurality of optical signals. The receiving-side conversion member may include a photodetector such as a photodiode for converting the second plurality of optical signals into the second plurality of electronic signals. More preferably, the receiving-side conversion member may include a plurality of photodetectors. The receive-side packet processor may be configured to reassemble the second plurality of electronic signals representing packet slices into the second electronic signal. By dividing a packet or frame into multiple slices before sending to another switch module, data can be sent using many different wavelengths multiplexed into a single optical link by a multiplexer. In this way, several pieces of information can be sent in parallel and result in increased bandwidth and more efficient data transfer.

In the case where the transport-side packet processor is configured to perform both packet slicing and packet slicing, the packet slicing step (ie, forming the frame of data) is performed first, followed by slicing the frame. Accordingly, at the destination (or intermediate) switch module of the received signal, the packet processor reassembles the second plurality of electronic signals (ie, packet slices) into a single second electronic signal, which is then reconstructed from the frame the original package.

After fragmentation, frames are constructed that each contain data intended for only a single destination switch module. After that, the data is converted into the first plurality of optical signals with different wavelengths, and the data is wavelength-multiplexed by the transmission-side multiplexer to form the multiplexed fabric output signal. Preferably, the switch module is configured to operate in burst mode, wherein the switch module is configured to transmit the multiplexed fabric output signal in a series of consecutive bursts, each burst comprising packets from a single data frame and/or packet fragmentation such that each burst includes only packets and/or packet fragments with the same destination module. Each successive burst may include data frames with different destination switch modules. The paired sequential bursts may be separated by predetermined time intervals, which may be between 50ns and 1000ns, but preferably between 50ns and 200ns. Most preferably, the time interval is approximately approximately 100 ns. Preferably, all fabric parts of an active switch connected to a single sub-array are configured to operate synchronously, ie each fabric part sends bursts to the inputs of the active switch simultaneously. In this way, the active switch can route each signal to the next switch module in one switching action.

The transmit-side packet processor may also be configured to perform error correction on incoming data packets. This operation may be performed by methods such as error correction and retransmission or forward error correction (FEC). Additionally, the switch module may further include a management portion configured to perform fabric management procedures including initialization, program routing/forwarding tables, fault reporting, diagnostics, statistics reporting, and metering .

In order to control the exchange of data by the active switches, each sub-array of switch modules may include an arbiter configured to control inclusion in the sub-array based on destination information stored in the data packets to be exchanged operation of the active switch. This allows to provide a route that ensures that all data reaches its destination in a non-blocking manner and minimizes the occurrence of bottlenecks. The arbiter may be connected to a switch driver that controls the operation of the switch. The arbiter is connectable to a transmit-side packet processor in each switch module of the sub-array that includes it. Alternatively, each fabric portion of each switch module may further include a controller via which the arbiter may be connected to the transmit-side packet processor. When a data packet is received at the transmit-side packet processor, the transmit-side packet processor is configured to send a request to the arbiter, the request preferably identifying the destination switch module of the data packet. The transmit-side packet processor may look up in a lookup table or otherwise which output of the active switch to which the transmit-side packet processor is connected corresponds to the destination switch that is the subject of the request module. More specifically, the output connected to the destination switch module or the intermediate switch module where the next optical hop should take place then requests the output itself from the arbiter.

Thus, one or both of the transmit-side packet processor and the arbiter may include a look-up table containing a look-up table that associates switch modules in a sub-array with R _i outputs of an active switch information. When a request is made, the arbiter then establishes a scheme that ensures, to the greatest extent possible, that each packet can perform its next optical hop. More specifically, the arbiter may be configured to execute a bipartite graph matching algorithm to compute pairings between R _i inputs and R _i outputs of an active switch such that each input is associated with at most one output. paired and vice versa. Naturally, in some cases, such as when several fabric parts send large amounts of data all intended for the same output of the active switch, the request cannot be satisfied. Accordingly, the arbiter may be configured to store information related to unsatisfied requests in the request queue. Then, until these requests are satisfied, the associated data is buffered on the switch module, eg in the transport side packet handler or in a separate transport side memory. In this way, requests that cannot be satisfied are delayed, rather than dropped, eg when a local bottleneck occurs at an active switch. In other words, the arbiter maintains the state of a buffer memory or virtual output queue (VOQ) on the switch module, which can be in the form of a counter (counting, for example, the number of packets or bytes per VOQ), or a storage The first-in, first-out (FIFO) form of the packet descriptor. However, the actual packets themselves remain stored on the switch module, not at the arbiter.

When it is necessary for a packet to perform more than one hop in order to reach its destination switch module, the route can be completely inferred from a comparison between the coordinates of the source switch module and the destination switch module. For example, in a process known as dimensional order routing, a first hop may match first coordinates of source and destination switch modules, a second hop may match second coordinates of source and destination switch modules, and so on, until All coordinates match, ie until the packet has been delivered to the destination switch module. For example, in a four-dimensional network, if a source switch module would have coordinates (a,b,c,d) and a destination switch module would have coordinates (w,x,y,z), then the dimensional ordering route could be: (a,b,c,d)→(w,b,c,d)→(w,x,c,d)→(w,x,y,d)→(w,x,y,z). At any point along the route, the packet processor can compare the coordinates of the source switch module and the coordinates of the destination switch module and determine which coordinates do not yet match. The packet processor will then decide to route along the non-matching direction with, for example, the lowest index or the highest index.

The active switch of the present invention may be in the form of an optical active switch. Such optically active switches may be based on an arrangement of Mach-Zehnder interferometers (MZIs) and, more specifically, may be in the form of MZI cascaded switches. The MZI cascade switch includes a plurality of MZIs, each having: two arms split at the input coupler, and the two arms feed the split paths into an output coupler used to recombine the paths; and two outputs part. The plurality of MZIs are preferably arranged to provide access from each input to each output of the MZI cascade switch. To the greatest extent possible, the arms are of the same length. Alternatively, the arm may be unbalanced where it is preferable to have a default output. Each MZI may include electro-optic regions at one or both arms, where the refractive index depends on the voltage applied to the regions via one or more electrodes. The phase difference of the light traveling through the electro-optic region can thus be controlled by applying a bias voltage via the electrodes. By adjusting the phase difference and thus the resulting interference at the outcoupling, light can be exchanged from one output of the MZI to the other. Preferably, the MZI cascade switch has _Ri inputs and _Ri outputs, and these may for example consist of a plurality of 1 x 2 and 2 x 1 MZIs arranged to provide connections from each input to each output path. When R _i is 5 or more, MZI cascade switches or any other active switches such as this are beneficial in a full mesh for connecting R _i interconnecting switch modules because a full mesh requires 1/2·R _i (R _i -1) fibers are used to connect all the fabric parts, while an active switch requires only 2R _i fibers. It is possible to form an MZI cascade switch with R _i = ²ⁿ inputs and outputs by building R _i "1x R _i demultiplexing trees" and R _i "R _i x 1 multiplexing trees", where Each tree includes n levels with 1x2 (demultiplexing) or 2x1 (multiplexing) switches, with ^2k switches at the kth level. By building R _i +1 trees on each side and ignoring internal connections, additional ports can be supported on each cascaded switch so that the input is not connected to the output that is connected to the switch itself. MZI cascade switches such as this are largely wavelength-agnostic and are therefore able to switch the entire multiplexed fabric output signal from the input without any demultiplexing/multiplexing at the input and output to the output.

Alternatively, the active switch may be in the form of an electronic active switch, such as an electronic crossbar switch. More preferably, the electronic active switch may be an electronic shared memory switch. An electronic shared memory switch is an electronic crossbar switch that also includes memory. The presence of memory within the switch is advantageous because it means that the switch can not only perform switching, but also buffering, ie storing queues of packets when a bottleneck occurs at the electronic shared memory switch, as described above. This means that the electronics on the packet processor can be simplified.

In order to use electronic rather than optical active switches in the architecture of the present invention, the multiplexed fabric output signal must be converted into a signal that can be switched electrically. Accordingly, an electronically active switch may comprise an opto-electrical converter at each input for converting the multiplexed fabric output signal from an optical signal to an electronically active switching signal; and an electrical-to-optical converter at each output , the electro-optical converter is used to convert the electronically active switching signal into an optical signal in the form of a multiplexed fabric input signal, wherein the electronically active switching signal is configured to convert the electronically active switching signal from one of its R _i inputs. Either swaps to any of its _Ri outputs. Furthermore, in order to handle the multiplexed nature of the signal, the optical-to-electrical converter may include a demultiplexer for demultiplexing the multiplexed fabric output signal into a first plurality of intermediate optical signals, the demultiplexer Each of the intermediate optical signals is preferably converted by a corresponding plurality of photodetectors into intermediate electronically actively switched signals for switching to a desired output, and the electro-optical converter may be configured to convert the plurality of switched The intermediate electronically actively switched signals are converted into a second plurality of intermediate optical signals, and also includes a multiplexer for multiplexing the second plurality of intermediate optical signals to form a multiplexed fabric input signal. In a preferred embodiment, the electronically active switch may be configured to temporarily store a queue of packets or frames of data when requests related to said packets or frames cannot be satisfied.

Any or all of the multiplexer, transmit-side multiplexer, demultiplexer, and receive-side demultiplexer are preferably in the form of an arrayed waveguide grating (AWG), which is a passive device. AWGs allow a single fiber to carry multiple optical signals of different wavelengths along the river. Because the wavelengths of the multiple modulated optical signals produced by the modulator are all different, the multiplexed fabric output signal produced by the AWG suffers from little to no crosstalk, as the light of different wavelengths interfere only linearly. Alternatively, instead of an AWG, the multiplexed signal can be broadcast to a number of wavelength selective filters, each tuned to receive one of the desired split signals at a wavelength.

An important consideration in a switching system, such as the switch of the present invention, is its bandwidth. In the following discussion, "bandwidth" is used to refer to the maximum rate of data transfer that a particular portion can achieve, and is typically measured in gigabits per second (abbreviated herein as "Gbps"). Specifically, it is important to ensure that bandwidth is reserved on both a local and global scale. In order to ensure that more data than can simultaneously be transmitted from a given switch module cannot enter the switch module at a given time (ie, creating a bottleneck localized to the switch module), the client portion of the switch module has a The total bandwidth preferably does not exceed the total bandwidth of the fabric parts on the same switch module. More preferably, the total bandwidth of the fabric parts on the switch module exceeds the total bandwidth of the client parts on the same switch module, and most preferably, the bandwidth of each fabric part on the switch module exceeds or equals the bandwidth on the switch module. The total bandwidth of all client parts of . In this way, local bottlenecks caused by unexpectedly large amounts of incoming data from multiple client devices all being directed to the same fabric portion on the same switch module can be avoided. In particular, this allows all signals to be multiplexed together for subsequent transmission in a non-blocking manner.

In a preferred embodiment, the active switch is located on or in the optical backplane and is preferably connected to the optical backplane. Preferably, the backplane contains optical links for connecting switch modules to active switches, thus providing a connection between each switch module and each active switch, each of the switch modules being shared with the active switch subarray. More specifically, each of the optical links may provide a connection for conveying a multiplexed fabric output signal between the transmit-side multiplexer on the switch module and the input of the active switch. Active Optical Backplane Modules (AOBMs) can be used when the backplane is used in conjunction with the optical active switches described above or the like. The switch modules may be detachable or detachable from the backplane so that the switch modules can be rearranged according to external requirements. Accordingly, the switch module may also comprise connection members for connection to the optical backplane. The connection member may comprise an array of single mode fibers joined using MPO connectors or similar means.

According to a third aspect of the present invention, there is provided an optical backplane for use in an N-dimensional optoelectronic switch, the optical backplane being arranged to provide connections between switch modules of an N-dimensional array arranged in an N-dimensional array, the i-th dimension has a base R _i (i=1,2...N), where each switch module has an associated set of coordinates giving its position relative to each of the N dimensions, and each switch module is N sub- a member of an array S _i , each sub-array S _i comprising R _i switch modules that differ only in coordinates with respect to their position in said i-th dimension, and each of said N sub-arrays is associated with a different dimension, The optical backplane includes:

An array of active switches, each having R _i inputs and I _i outputs, the active switches are associated with each sub-array Si of switch modules _, the switch arrays being configured such that when all When the switch module array described above is connected to the optical backplane:

Each input of each active switch is connected via an optical link to each of the R _i switch modules in the associated sub-array Si _,

The active switch is configured to direct signals received from the optical link from any of its _Ri inputs to any of its _Ri outputs.

According to a fourth aspect of the invention there is provided a method for switching data packets from a first switch module to a second switch module using an N-dimensional optoelectronic switch according to the second aspect of the invention. It should be noted that any of the above optical features may be used in conjunction with the method of the fourth aspect of the invention and the hardware of the first, second and third aspects. The method includes the following steps:

(a) receiving at the input of the fabric portion of the first switch module a data packet carrying information identifying the intended destination switch module;

(b) converting the packet into a first plurality of optical signals containing the same information;

(c) multiplexing the first plurality of optical signals into a multiplexed fabric output signal;

(d) transmitting the multiplexing fabric output signal to the input of an active switch, the active switch and both the first and second switch modules being located in the same sub-array _Si ;

(e) switching the multiplexing fabric output signal to an output of the active switch corresponding to the second switch module to generate a multiplexing fabric input signal;

(f) demultiplexing the multiplexed fabric input signal into a second plurality of optical signals at the first fabric port of the second switch module;

(g) converting the second plurality of optical signals back into the original datagrams; and

(h) forwarding the data packet to the input of the client portion of the second switch module or the second fabric portion of the second switch module.

Such methods are suitable for switching signals from a source switch module to a destination switch module, eg, when the source and destination switch modules are in the same sub-array. Alternatively, the method may be adapted to switch from a source switch module to an intermediate switch module. The method may be adapted to transmit the signal to a third switch module, eg a destination switch module, by adding the following steps:

(i) receiving at the input of the second fabric port of the second switch module a data packet carrying information identifying the destination switch module;

(j) converting the packet into a third plurality of optical signals containing the same information;

(k) multiplexing the third plurality of optical signals into a multiplexed fabric output signal;

(1) transmitting the multiplexing fabric output signal to the input of an active switch located in the same sub-array _Si as the second and third switches;

(m) switching the multiplexing fabric output signal to an output of the active switch corresponding to the third switch module to generate a multiplexing fabric input signal;

(n) demultiplexing the multiplexed fabric input signal into a second plurality of optical signals at the first fabric port of the third switch module;

(o) Converting the second plurality of optical signals back into the original data packets.

Since the optoelectronic switch is adapted to transmit the optical signal from the input device to the output device, the optical signal can be received from the input device before step (a), and after step (h) or step (o), depending on whether an intermediate light is required or not Hop, which forwards the packet to the output device.

In some embodiments, rather than each sub-array including a single active switch, the sub-array may include multiple or a set of active switches, thus, another aspect of the present invention may provide for routing optical signals from an input an N-dimensional optoelectronic switch that transmits the device to an output device, the optoelectronic switch comprising a plurality of switch modules as claimed in claim 1, the switch modules being interconnected, wherein:

The switch modules are arranged in an N-dimensional array, the i-th dimension has size Ri ( _i =1,2...N), each switch module has a position giving its position relative to each of the N dimensions The associated coordinate set of ;

Each switch module is a member of N sub-arrays S _i , each sub-array S _i includes R _i switch modules that differ only with respect to the coordinates of the position in the i-th dimension, and each of the N sub-arrays associated with different dimensions;

each of the switch modules is configured to generate a multiplexed fabric output signal;

Each sub-array _Si also includes one or more active switches arranged to provide connections between all switch modules in the sub-array;

The input of each active switch is configured to receive a multiplexed fabric output signal from one or more of the R _i switch modules in the sub-array; and

Each of the one or more active switches is configured to, based on the destination information contained in the first electronic signal received at the transmit-side input of the switch module, A multiplexing fabric output signal is directed from any switch module in the sub-array to any other switch module in the sub-array from which the active switch receives the multiplexing fabric output signal.

In particular, in some embodiments of the above aspects of the invention, a sub-array of R _i switch modules includes only a single active switch having R _i inputs and R _i outputs, and:

Each input of the active switch is configured to receive a multiplexed fabric output signal from each of the R _i switch modules in the sub-array,

each of the switch modules is configured to receive a multiplexed fabric output signal from one of the R _i outputs of the active switch, and

The active switch is configured to, based on the destination information contained in the first electronic signal received at the transmit-side input of the switch module, multiplex a fabric output signal therefrom. Any of the R _i inputs leads to any of its R _i outputs, the active switch receiving the multiplexing fabric output signal from the switch module.

In other embodiments, at least one of the sub-arrays S _i may include P _sub active switches preferably arranged to form a network connecting the modules in the sub-array with the modules in the sub-array of each other switch module. P _sub may be the same for all sub-arrays and may be equal to the number of switch modules in a given sub-array, or alternatively, P _sub may be greater or less than the number of switch modules in a given sub-array. In some embodiments, P _sub may be the same for all subarrays associated with a given dimension, but P _sub may vary from dimension to dimension. In some embodiments, for sub-arrays associated with some dimensions, there may be only a single active switch interconnecting switch modules within those sub-arrays, and a plurality or set of P _sub active switches exist in other dimensions in the associated subarray. This enables optoelectronic switches to adapt to client needs. The use of multiple active switches within a subarray eliminates the need for a large cardinality active switch, replacing the large cardinality active switch with several active switches with a small cardinality. In a preferred embodiment, P _sub may be equal to the number of client ports on each switch module.

In some embodiments of the invention, in a given sub-array, there may be additional "layers" of switches between the switch modules and one or more active switches. The switches of the additional layers are referred to herein as "intermediate switches". In other words: each switch module in a given subarray is an active switch connected to that subarray via an intermediate switch. The intermediate switch may be an optical active switch, or an electronic active switch, such as an electronic packet switch. The nature of the switching elements (ie, switch modules, intermediate switches, and active switches) will be discussed in more detail in this application. For clarity, in the following paragraphs, the terms "leaf and spine" are used, whereby the following terminology is employed: switch modules are referred to as "leaf switches" and active switches are referred to as "spine switches".

In one or more preferred embodiments, the intermediate switches are bidirectional, ie the intermediate switches carry signals in both directions. An example of a bidirectional switch is an electronic packet switch which will be described in detail later. In an embodiment that includes a bidirectional intermediate switch, a signal may be sent from the source leaf switch to the backbone switch through the same "layer" of intermediate switches as the signal hops, and from the backbone switch to the destination leaf switch on the signal journey. In such embodiments, the structure of the sub-arrays can be represented by a folded Claus network that allows optoelectronic switches to scale without the need to use bulky and potentially costly large-cardinality active switches (i.e. large cardinality). Backbone switches), ie the number of external clients that can connect to each other is greatly increased.

In implementations using unidirectional intermediate switches, it may be desirable in such cases to include a first-tier intermediate switch for hopping from a source leaf switch to a spine switch, and a first-tier intermediate switch for a hop from the spine switch to the destination leaf switch Layer 2 intermediate switch. Such interconnections are best represented using unfolded Claus networks (or "partially folded" Claus networks), but are not within the scope of the present claims.

The use of a folded/unfolded/partially folded Claus network or any Claus-like network ensures an advantageous non-blocking operation of the optoelectronic switch. In fact, the leaf switches, intermediate switches, and backbone switches in a given subarray may be interconnected with any other network configuration with suitable path redundancy to ensure or improve the scope of non-blocking operation.

Using the terms leaf and backbone described above, an embodiment can be defined as an N-dimensional optoelectronic switch for carrying optical signals from an input device to an output device, the optoelectronic switch comprising:

A plurality of leaf switches, each having a cardinality R and arranged in an N-dimensional array, where each dimension i has a corresponding size R _i (i=1, 2, . . . , N), each leaf switch has given its relative an N-tuple (x ₁ , . . . , x _N ) of the associated coordinates of the location of each of the N dimensions;

wherein each leaf switch is a member of N sub-arrays, each of the N sub-arrays is associated with a different one of the N dimensions, and includes:

a plurality of R _i leaf switches whose coordinates differ only with respect to the i-th dimension, each leaf switch having C client parts for connecting to input devices or output devices, and F for connecting to backbone switches the constituent part of;

a plurality of Si _backbone switches each having R fabric parts for connecting to the fabric parts of the leaf switches, and

Wherein, in a given sub-array, each leaf switch in the sub-array is connected to each backbone switch via an intermediate switch.

To understand the beneficial effects of an embodiment that includes intermediate switches connected between leaf switches and spine switches, consider a sub-array where spine switches are connected via only a single spine switch or a set of parallel spine switches ("parallel" means that the spine switches each connected only to leaf switches, not to other backbone switches). Each of these spine switches is relatively connected to each leaf switch in the subarray under consideration, and each leaf switch must be connected to each spine switch, otherwise it may not be possible to provide each leaf switch with all the connections between other leaf switches.

Therefore, as mentioned above, for large sub-arrays, a large number of large-cardinality backbone switches are required. Each leaf switch still needs to have a connection to each spine switch, but in the embodiments discussed here, the leaf switches are connected to the spine switches via intermediate switches. Each intermediate switch may have multiple inputs and multiple outputs, the number of inputs being the same as the number of outputs. The "cardinality" of an intermediate switch refers to the number of inputs or the number of outputs, not the total number of inputs and outputs. Thus, the output of a given intermediate switch can be connected to multiple or a group of backbone switches in a sub-array. A given leaf switch can be connected to a cluster of intermediate switches, each of which is connected to a set of backbone switches. More specifically, each of the intermediate switches of the cluster to which a given leaf switch is connected may be connected to a different disjoint set of backbone switches ("disjoint" means that no two intermediate switches in a given cluster are connected to The same backbone switch, that is, the set of backbone switches does not overlap).

In this way, the leaf switch under consideration can be connected to all backbone switches in the sub-array, but via intermediate switches of a smaller cardinality. Because the intermediate switches also have multiple inputs, these inputs can be shared among multiple leaf switches. In other words, for a given cluster of intermediate switches whose outputs provide connections to all backbone switches in a subarray, each input of each intermediate switch in the cluster may be connected to a corresponding (ie, different) leaf switch. Thus, a leaf switch can be divided into multiple clusters, and in particular: within a given subarray, a leaf switch can be divided into multiple clusters, each of the clusters containing multiple leaf switches. Each cluster can have its own associated intermediate switch cluster that provides connectivity between each leaf switch cluster and each spine switch in the array via the cluster of intermediate switches, in other words, each leaf switch cluster can communicate with A cluster of one or more intermediate switches is associated to form a line card assembly, each leaf switch in the cluster can be connected to each intermediate switch in the line card assembly, and the intermediate switches can be arranged such that signals are passed from the leaf switches to the backbone. Pass through intermediate switches during switching.

Optionally, in a given sub-array, each backbone switch may be connected to an intermediate switch in a line card assembly located in the sub-array, and no more than one fabric entry intermediate switch in the line card assembly is connected to the given sub-array. Fixed backbone switch. Within a line card assembly, there may be M distinct sets of intermediate switches, each set being configured to transmit signals within respective sub-arrays containing the line card assembly, each of those sub-arrays being related to the M dimensions. are associated with the corresponding dimensions.

To maximize topological regularity, preferably all or substantially all intermediate switches used in embodiments of the present invention have the same cardinality, or more specifically, the same number of inputs and outputs. In particular, when all intermediate switches in a given sub-array have a given cardinality, it is possible to support the same number of backbone switches without compromising the bisection bandwidth, and the number of leaf switches increases in multiples of the cardinality. For example, if a base 3 intermediate switch is used, the sub-array can be tripled in size. In other words, the opportunity to scale up the array is greatly increased without increasing the size of the backbone switches used.

The above connections are related to the transmission of signals from the outputs of the leaf switches to the inputs of the backbone switches. In order to complete the routing of a signal from one leaf switch in the subarray to another leaf switch in the subarray, the signal must be routed from the output of the backbone switch to the fabric portion on the destination (or intermediate) leaf switch. This transmission also takes place via an intermediate switch, which is preferably different from the one through which the signal enters the backbone switch, although it is possible that the intermediate switches are members of the same cluster.

When leaf switches, intermediate switches, and backbone switches are as described above, within a given subarray, the leaf switches are effectively connected via a five-level Claus network, ie, leaf switches→intermediate switches→spine switches→intermediate switches→leaf switches. As discussed, using a Claus network such as this (or a partially folded Claus network) means that it is possible to accommodate any one-to-one pairing between R _i leaf switches in a sub-array in a non-blocking fashion combination.

The arrangements described in the previous paragraphs do not have to be used for exchanges in all N dimensions. In some embodiments of the invention, each leaf switch in a given subarray may be connected to each backbone switch via intermediate switches only with respect to the subarrays associated with M dimensions, where M<N. The other optional features presented above are applicable to any or all dimensions where exchanges between leaf switches are via intermediate switches.

Any or all of the leaf switches, backbone switches and intermediate switches can be placed on the optical backplane. Thus, leaf switches, intermediate switches and spine switches can be located on the card. The card may be a printed circuit board on which, for example, electronic components, optical components and control components (ie arbiters) are formed. The card may also house optical and electronic components therebetween. In particular, embodiments of the present invention may include two types of cards: line cards and fabric cards. More specifically, the components may be located on line cards or fabric cards. Line cards are "client facing" cards, and fabric cards are "fabric facing" cards.

Leaf switches and intermediate switches, ie line card assemblies as described above, may be located on respective line cards. More specifically, a single line card includes at least one leaf switch and at least one intermediate switch. In some embodiments, a single line card may include multiple leaf switches and/or multiple intermediate switches. In embodiments where there is a leaf switch cluster and an associated intermediate switch cluster, the leaf switch cluster and its associated intermediate switch cluster are preferably installed on the same line card. In an N-dimensional embodiment of the present invention, each leaf switch is a member of N sub-arrays, as discussed previously in this application.

In embodiments where there are leaf switch clusters and intermediate switch clusters, each leaf switch cluster and its associated intermediate switch cluster may be located on its own line card. In a given sub-array, the backbone switches may be located on fabric cards. The fabric card may also include an arbiter for controlling the path of signals through backbone switches located on the fabric card. The fabric card may also include a plurality of arbiters, each of which is configured to control the path of a signal through a corresponding backbone switch on the fabric card. It is not required that all backbone switches be on the same fabric card, but more than one backbone switch can be on a given fabric card. There are two control elements in the current embodiment of the invention: routing/load balancing, and arbitration.

The packet processor can make routing decisions based on the packet's destination address and current location. On the path from the leaf to the backbone, routing decisions select the backbone to route to (usually trying to balance the load on the available backbones), which in turn determines specific output ports on both the local leaf switches and intermediate switches. The output identifier of the intermediate switch is passed to the arbiter so that the arbiter can determine which input of the intermediate switch needs to be connected to which of its outputs. On the path from the backbone to the leaf, routing decisions select the appropriate leaf based on the destination of the packet, which in turn determines the backbone and local output ports on the appropriate intermediate switches between the backbone and the destination leaf switch.

Arbitration is performed by the arbiter and is the process by which the arbiter determines the path a signal should take through an intermediate switch (i.e. from which input to which output) in order to ensure that all signals hitting the intermediate switch are directed towards the correct next switch Elements (the switching elements can be spine switches or leaf switches, depending on which "level" the signal is at) are directed. Thus, in some embodiments, there may be an arbiter associated with each intermediate switch in a given sub-array, in other words, a line card may include means for controlling the passage of signals through a line card component included in a line card assembly located on the line card. An arbiter for the paths of the intermediate switches, or there may be a plurality of arbiters, each configured to control the paths of the signals through the respective intermediate switches. Alternatively, the arbitration process is relatively simple (eg, compared to a radix 24 switching element) since the intermediate switches may have a small radix (eg, 2, 3, 4, 5, 6, 7, or 8), and thus multiple arbiters Can be combined into a single arbitration component, which can be an ASIC. In some embodiments, there may be a single arbiter, or a single arbitration component as described above, on each line card. In the embodiment described in this paragraph, the control performed by the arbiter is constrained within the boundaries of the line card under consideration. This minimizes latency and synchronization issues associated with control surfaces: distance/time of flight on a card can be controlled to tighter tolerances within the physical dimensions of a single card, which spans can be considerably apart from each other Distance positioning card. Furthermore, by having several arbiters, each associated with a small number (eg, one) of intermediate switches, a large number of small problems can be solved quickly and in parallel, as compared to having to focus on more complex problems relatively slowly.

Generally speaking, the intermediate switches are controlled by an arbiter configured to control the actions of at least one of the intermediate switches and backbone switches within a given sub-array based on destination information stored in the packets to be exchanged . This thus allows to provide a route that ensures that all data reaches the appropriate leaf switch in a non-blocking manner and minimizes the occurrence of bottlenecks. The packet processors in the leaf switches may each be connected to an arbiter. When a data packet is received at the transmit-side packet processor, the packet processor may send a request to an arbiter to which the packet processor is connected, the request preferably identifying the destination leaf switch, or alternatively Identify the next leaf switch to which the packet should be sent (which may be the destination leaf switch). The arbiter can then generate a scheme that ensures, to the greatest extent possible, that each packet can perform its next hop. The structure of the leaf switch will be described in more detail below.

The arbiter can use dedicated control channels to connect to other components, such as packet processors and intermediate switches. The arbiter may also be connected to a driver chip configured to control the actions of the intermediate switch.

For ease of manufacture of optoelectronic switches according to embodiments of the present invention, it is preferred that each of the leaf switches contain the same components as each of the backbone switches. Furthermore, in other embodiments, each of the intermediate switches may contain the same components as each of the leaf switches and/or the backbone switches. Effectively, all switching elements (ie, switch modules, active switches, leaf switches, intermediate switches, and backbone switches, depending on the terminology used) may be the same or substantially the same. As such, optoelectronic switches can be constructed by assembling these elements, and then different functionalities of the switching elements (eg, the different functions of leaf switches, intermediate switches and backbone switches as described above) can be controlled using eg software. Intermediate switches and spine switches may differ from leaf switches in that they do not have a client part because intermediate switches and spine switches are only connected to other switching elements within the optoelectronic switch and not to clients end (ie, external) device. Furthermore, it must be emphasized that although the exchange elements may have identical components, or may be structurally identical or substantially identical, the functionality of the different types of exchange elements may vary. It should be noted that the reason for the use of the terms leaf/spine/intermediate is to illustrate that the three types of switching elements can be the same/substantially the same, rather than using terms such as "switch modules" and "active switches" which may appear to be distinct components .

To use the term "leaf and backbone" used earlier in the description, three different types of switches can be defined as follows:

■ Leaf switches are switching elements that have:

a client part for connecting to an input device or an output device;

a first fabric portion and a second fabric portion each for processing signals and communicating with other switching elements, the first fabric portion having a transmit side and a receive side,

The transmission side has:

a transmission-side input for receiving a first electronic signal carrying information including information about the destination switch module of the first electronic signal via which The client part of the description is received from:

the output of the second fabric portion, or

input device;

a transmission-side conversion member for converting the first electronic signal into a first plurality of optical signals containing the same information;

a transmit-side multiplexer for converting the first plurality of optical signals into multiplexed fabric output signals for transmission to an intermediate switch or backbone switch, and

The receiving side has:

a receiving-side multiplexer, the receiving-side multiplexer is configured to receive a multiplexed fabric input signal from an intermediate switch or a backbone switch, and separate the multiplexed fabric input signal into a second plurality of optical signals;

a reception-side conversion member for converting the second plurality of optical signals into second electronic signals, and

a receiving-side output for sending the second electronic signal via the client part to:

the transmit-side input of the second fabric portion, or

output device.

■ A backbone switch is a switching element that has:

a first fabric portion and a second fabric portion each for processing signals and communicating with other switching elements, the first fabric portion having a transmit side and a receive side,

The transmission side has:

A transmission-side input for receiving a first electronic signal carrying information, the information including information about the destination switch module of the first electronic signal, the first electronic signal being receiving at the output of the second configuration part;

a transmission-side conversion member for converting the first electronic signal into a first plurality of optical signals containing the same information;

a transmit-side multiplexer for converting the first plurality of optical signals into multiplexed fabric output signals for transmission to an intermediate switch or leaf switch, and

The receiving side has:

a receive-side multiplexer configured to receive a multiplexed fabric input signal from an intermediate switch or leaf switch, and to separate the multiplexed fabric input signal into a second plurality of optical signals;

a reception-side conversion member for converting the second plurality of optical signals into second electronic signals, and

A receiving-side output terminal, the receiving-side output terminal is used for sending the second electronic signal to the transmitting-side input terminal of the second configuration part.

■ An intermediate switch is a switching element that has:

a first fabric portion and a second fabric portion each for processing signals and communicating with other switching elements, the first fabric portion having a transmit side and a receive side,

The transmission side has:

A transmission-side input for receiving a first electronic signal carrying information, the information including information about the destination switch module of the first electronic signal, the first electronic signal being receiving at the output of the second configuration part;

a transmission-side conversion member for converting the first electronic signal into a first plurality of optical signals containing the same information;

a transmit-side multiplexer for converting the first plurality of optical signals into multiplexed fabric output signals for transmission to a backbone switch or leaf switch, and

The receiving side has:

a receiving-side multiplexer, the receiving-side multiplexer is configured to receive the multiplexed fabric input signal from the backbone switch or the leaf switch, and separate the multiplexed fabric input signal into a second plurality of optical signals;

a reception-side conversion member for converting the second plurality of optical signals into second electronic signals, and

A receiving-side output terminal, the receiving-side output terminal is used for sending the second electronic signal to the transmitting-side input terminal of the second configuration part.

As can be seen from this section, the order of switching elements that a signal encounters on its journey from the source to the destination leaf switch can be:

■ In an embodiment that includes an intermediate switch: source leaf switch→intermediate switch→backbone switch→intermediate switch→destination leaf switch.

■ In embodiments that do not include intermediate switches, ie where there is a single spine switch, or a single set of parallel spine switches: source leaf switch→spine switch→destination leaf switch.

From the above, it is evident that only leaf switches include the client part, because then the leaf switches are the only "client-facing" switching elements. Intermediate switches and backbone switches are defined as interconnect fabrics that connect all leaf switches together, and thus only include fabric parts. It will be noted in the above that all switching elements in an optoelectronic switch can be the same or substantially the same. Although the backbone switches and intermediate switches do not have client parts, this is possible because the term "client part" is a functional term used to describe the fact that external devices can connect to these parts. Each of the switching elements may have the same physical structure, but may differ for various uses. Similarly, when switching elements act as leaf/spine/intermediate switches depending on the particular implementation of the invention, a different number of clients/fabric parts may be used. It is also worth noting that the optical features set forth above with reference to the switch module of the first aspect of the invention are equally applicable to leaf switches, backbone switches and intermediate switches as defined herein.

It is envisaged that similar advantages may be achieved by providing an optoelectronic switch for transmitting optical signals from an input device to an output device, the optoelectronic switch comprising a plurality of interconnected switch modules according to the first aspect of the invention, wherein:

The switch modules are arranged in an N-dimensional array, the i-th dimension has a cardinality R _i (i=1,2,...N), and each switch module has its relative relative to each of the N dimensions given The associated set of coordinates for the location of ;

Each switch module is a member of at most N sub-arrays, the N sub-arrays being associated with the coordinate set of the associated switch module;

the switch modules are connected by an array of active switches each having an input and an output, wherein each active switch is associated with a given sub-array or given sub-arrays of the array of switch modules,

And, in use:

Each input of each active switch is configured to receive a multiplexed fabric output signal from the switch module to which it is connected,

Each active switch is configured to switch a signal from any of its inputs based on the destination information contained in the first electronic signal received at the transmit-side input of the switch module directed to any of its outputs, the active switch receives a multiplexed fabric output signal from the switch module, and

The signals sent from the output of the active switch form a multiplexed fabric input signal, which can be received by another switch module to which the active switch is connected.

Schematic Description

These and other features and advantages of the present invention will be understood and appreciated with reference to the specification, claims and drawings, wherein:

Figure 1 is a schematic diagram illustrating the manner in which two switch modules may be connected in an embodiment of the present invention.

Figure 2 is a schematic diagram of the switch module, identifying the different functional parts.

Figure 3 is a schematic diagram showing the components of the constituent parts inside a switch module according to an embodiment of the first aspect of the present invention.

4 is a schematic diagram showing components of two different architectural portions of a switch module according to an alternative configuration.

5 is a schematic diagram of a one-dimensional switch that may be constructed using a switch module according to an embodiment of the present invention.

Figure 6 is a schematic diagram of a two-dimensional switch according to an embodiment of the second aspect of the present invention and which may be constructed using a switch module according to an embodiment of the first aspect of the present invention.

7 is a schematic diagram of an alternative layout of a two-dimensional switch according to another embodiment of the second aspect of the present invention and which may be constructed using a switch module according to an embodiment of the first aspect of the present invention.

8 is a schematic diagram of a three-dimensional switch according to another embodiment of the second aspect of the present invention, and which may be constructed using a switch module according to an embodiment of the first aspect of the present invention.

Figures 9A, B and C are schematic diagrams showing further examples of switch architectures according to the second aspect of the invention, wherein all active switches have the same number of inputs/outputs.

Figure 10 is a schematic diagram showing the manner in which an arbiter may be connected to a switch module arranged in accordance with the second aspect of the present invention.

Figure 11 is a schematic diagram illustrating the connection between an arbiter and a spatial optical switch according to an embodiment of the present invention.

Figure 12 shows an exemplary setup of a Mach-Zehnder switch that can be used as a spatial optical switch as part of a switch array according to an embodiment of the second aspect of the present invention;

13A, B, and C each show an example of an embodiment of the present invention in which an electronic active switch or multiple electronic active switches are used instead of optical active switches.

14A, B and C show representations of 2D optoelectronic switches. Figures 14A and B are folded; only 14C is unfolded.

15A, B and C show schematic diagrams of known folded Claus networks.

Detailed ways

The detailed description set forth below in connection with the accompanying drawings is intended as a description of exemplary embodiments of switch modules and optoelectronic switches provided in accordance with the present invention and is not intended to represent the only forms in which the present invention may be constructed or utilized. The description sets forth the features of the invention in conjunction with the illustrated embodiments. It should be understood, however, that the same or equivalent functions and structures may be implemented by different embodiments, which are also intended to be encompassed within the spirit and scope of the present invention. As indicated elsewhere herein, similar reference numerals are intended to indicate similar elements or features.

Figure 1 is a schematic illustration depicting typical connections between two switch modules of a switch architecture according to an embodiment of the present invention. In this figure, only two photodetectors P and two modulators M are shown on each switch module to illustrate the connections between the switch modules.

The switch module 1 has a fabric side F1 for connection with other switch modules present in the optoelectronic switch (in this schematic diagram, only switch module 2), and a client side C1 for connection to external devices. On the fabric side F1 of the switch module 1, there are two electro-absorption modulators M1, M2, the outputs of which are incident on the multiplexer MUX1, which in this case is an AWG. MUX1 combines the signals exiting M1 and M2 and transmits the signals (dashed arrows) to the R _i x R _i optical active switch (herein, "optical active switch" unless the context clearly dictates otherwise) 4 , which An optical active switch has _Ri inputs and _Ri outputs. The characteristics of this switch are described in more detail below.

The optical active switch 4 passes the signal from the input where the multiplexed signal arrives from the MUX1 to the output, depending on the desired destination switch module of the signal, in this case switch module 2 . The control scheme used to determine the destination switch module will be described in more detail later. The signal is transmitted from the output of the optical active switch 4 to the destination switch module 2 (dashed arrow). The signal is incident on the demultiplexer DEMUX2 of the switch module 2 . Here, the multiplexed signal is demultiplexed into its constituent individual signals, each of which is incident on a single photodetector P3, P4. The signal can be further transmitted from the photodetectors P3, P4 to the client part on the client side C2 of the switch module 2, or (in the case of optoelectronic switches multidimensional) the signal can be transmitted back to the fabric side F2 for further transmission . The solid arrows in FIG. 1 show an alternative transmission of signals from switch module 2 to switch module 1 . Arrows (both solid and dashed) in the drawings represent WDM optical connections.

Figure 2 shows a schematic diagram of a typical switch module used in embodiments of the present invention. Each switch module includes an integrated switch segment with a client side and a fabric side, as shown in the previous figures. The number of client ports and fabric ports depends on the needs of the product, and bandwidth constraints. In the switch module shown in Figure 2, there is also a management section that is configured to perform fabric management procedures such as initialization, program routing/forwarding tables, fault reporting, diagnostics, statistics reporting, metering, and the like .

Figure 3 shows a more detailed view of the fabric side Fl of a typical switch module 1 used in the architecture of an embodiment of the present invention. First, the structure of the switch module 1 will be described, followed by a description of the paths of signals in the switch module 1 . The fabric side F1 is divided into two parts, the transmit side Tx and the receive side Rx. The transmit side Tx includes an array of packet processors PP-Tx, EAMMOD1, MOD2, ..., MODQ, each of the EAMs receiving an input from one of the arrays of light sources LS1, LS2, ..., LSQ. Each of the EAMs of the array is connected to a signal multiplexer WDM-MUX, which in turn outputs its WDM signal to an optical active switch, which can be considered to implement the present invention. The "fabrication" of the interconnection between the switch modules 1 of the optoelectronic switch of the embodiment. The receiving side Rx has a similar structure. More specifically, the receiving side Rx includes a packet processor PP-Rx that receives input from an array of photodetectors PD1, PD2, . . . , PDQ each receiving input from a single demultiplexer input of the device WDM-DEMUX. The demultiplexer receives input from an optical active switch (not shown in Figure 3). The controller CTRL is also included in the switch module 1 and is not bound to either the transmission side Tx or the reception side Rx. The controller CTRL is bidirectionally connected to the two packet processors PP-Tx, PP-Rx and the arbiter shown by the arrows marked AR.

At a higher level, it should be noted that all data transfers that take place on the left side of the diagram take place in the electrical domain, and that all data transfers that take place on the right side of the diagram take place in the optical domain, ie all data transfers occur in the Between the multiplexer WDM-MUX and the demultiplexer WDM-DEMUX.

Now, the journey of the packet through the various components of the switch module 1 will be described. Contains information that will be transferred from the source switch module to the destination switch module. Specifically, the inclusion includes information about the intended destination switch module. In the following description of the journey the packet goes through, it is assumed that all data associated with that packet has the same intended destination switch module.

The following process is performed in the electrical domain. The packets may be incident on the transmission side Tx of the switch module 1 eg from the client part, which is connected to the client side of the switch module 1 . Alternatively, packets may be received from the receiving side Rx of switch module 1 (ie the same switch module) via an integrated switch such as shown in Figure 2, so that the packets can be forwarded to another switch module (not shown) for use by Teleport to a different dimension. This packet transfer between dimensions will be explained in more depth later. The packets incident on the transmission side Tx enter the packet processor PP-Tx, where they are sliced into a first plurality of Q-electronic signals in the form of packet slices, each having the same destination switch module. Each of the electrical signals is then transmitted to one of the Q EAM MOD1, MOD2, . . . , MODQ. At this time, each of the electrical signals contains information corresponding to the data in the packet slice and information about the destination switch module of the packet.

Now consider the packet slice incident on MOD1. MOD1 has two inputs: (a) a slice of electrical packets, and (b) light of a given wavelength λ ₁ from light source LS1. The optical channels are chosen to minimize crosstalk and relatively easy to fabricate the waveguides in good yield. Optical channel spacings between 0.4 nm and 2 nm are preferred. The laser light may have a line width as narrow as that particularly for this application and preferably no less than 1 KHz. In other configurations, the frequency resolution and spacing will depend on the sophistication of the device and thus on passive components. If there are, for example, 8 wavelengths, the device can be rather "rudimentary", but if more wavelengths are to be used, then higher specifications will be required.

MOD1 then modulates the light from light source LS1 to carry the information contained in the packet slices, thereby producing an optical signal having a given wavelength λ ₁ . From this point on, data transfer is in the optical domain. Each modulator operates similarly to generate the first plurality of Q optical signals. The Q optical packet slices from each of EAM MOD1, MOD2, ..., MODQ are incident on a multiplexer MUX where wavelength division multiplexing occurs to combine the Q optical signals (from each EAM) into a single output fiber. Each of the Q optical signals has a different wavelength, so crosstalk between the signals is minimized. The multiplexed signal forming the output signal of the multiplexed fabric is then transmitted to an optical active switch (described in more detail later). The optical signal generated in the switch module 1 is then transmitted by the optical active switch to its destination switch module or an intermediate switch module routed to the destination switch module. The control flow and associated hardware architecture to ensure that each signal eventually reaches the correct destination will be described in more detail later.

For the purposes of this description, we will continue to refer to Figure 3, but in general use, the source and destination switch modules will not be the same switch module. However, the source and destination modules can be the same module, eg for testing purposes. However, the source and destination switch modules should be substantially identical to each other, so that the description based on FIG. 3 still applies equally. The optical multiplexing fabric input signal from the optical active switch is incident on the demultiplexer DEMUX located on the receiving side Rx of the switch module 1 . The multiplexed fabric input signal is demultiplexed by the demultiplexer DEMUX into a second plurality of Q optical signals, which are equivalent to those combined at the multiplexer MUX on the source switch module 1 those light signals. The Q demultiplexed signals are then incident on each of the array of photodetectors PD1, PD2, ..., PDQ. In the photodetector, the demultiplexed signal is converted back into a second plurality of Q electrical signals, again containing the information contained in the original packet slice. The electrical signal is then transmitted to the packet processor PP-Rx, where it is reassembled into packets incident on the source switch module 1 using the information contained in the header of the packet slice The original package on the processor PP-Tx.

In some embodiments, each fabric section on a given switch module 1 has its own associated multiplexer and demultiplexer.

However, in an alternative configuration, as shown in Figure 4, it can be seen that this is not the case. In this case, the EAM MOD1, MOD2, ..., MODQ (and their associated light sources), the photodetectors PD1, PD2, ..., PDQ, the multiplexer WDM-MUX and the demultiplexer WDM-DEMUX are in Shared among N fabric ports. The diagram is divided into two sections to show which processes occur in the optical domain and which processes occur in the electrical domain. In this embodiment, there are additional arrays of multiplexers and demultiplexers, shown to the left of the dashed line. Compared to the multiplexer MUX located at the output of the EAM MOD1, MOD2, ..., MODQ for wavelength division multiplexing, the multiplexer to the left of the dashed line is configured to combine the signal in the electrical domain rather than the optical domain. reuse together. The same applies to the demultiplexer DEMUX. In another embodiment, the multiplexers and demultiplexers may be in the form of CMOS combinational logic circuits integrated into the switch module. The journey of a packet from a source switch module to a destination switch module will now be described with reference to FIG. 4 . In the case where the process or components are the same as those in FIG. 3, the description will not be repeated here. The packets entering the first fabric part pass through the packet processor PP-Tx as before, wherein in this case the packets are divided into three packet fragments, each in the form of an electrical signal. Similarly, at the same time, packets arriving at fabric part F2 enter the packet processor PP-Tx on the second fabric part and are also split into three packet fragments, again all in the form of electrical signals. The three optical packet fragments generated by the packet processor PP-Tx of each of the first and second fabric parts are then sent out to three different multiplexers MUX. In other words, each of the multiplexers MUX receives two electrical signals, each corresponding to a packet fragment from a different packet, one electrical signal incident on the PP-Tx on the first fabric portion, and one electrical signal Incident on the PP-Tx on the second fabric. These two signals are then multiplexed into a single multiplexed electronic signal, which is then transmitted to one of the EAM MOD1, MOD2, ..., MODQ. As in Figure 3, the EAM modulates the signals from the light sources LS1, LS2, ..., LSQ such that an optical signal is generated carrying information previously carried by the electrical signal, each EAM MOD1, MOD2, ..., MODQ produces a signal with a different wavelength Signal. Thus, as in Figure 3, the optical signals output from the EAM MOD1, MOD2, ..., MODQ are wavelength division multiplexed into a single fiber by the multiplexer WDM-MUX. For each time slot used to transmit the signal, an arbitration step is necessary in order to determine which fabric part is eligible to use the optical transmission path. Only one input to each multiplexer MUX can be reached at any given time to avoid data loss. Equivalently, on the reverse path, the demultiplexer DEMUX must be similarly controlled to send incoming packets to the correct receive fabric section, etc.

In addition to the NxN optical active switches, 1xK additional optical multiplexers/demultiplexers are required because in this configuration each switch module has only one optical transmitter and receiver, which must be optically Coupled to K different fibers (in both directions) for different dimensions, these multiplexers also need to be properly controlled to properly direct the signal. For the demultiplexer DEMUX, this means choosing the dimension along which the transmission is to be taken. For the multiplexer WDM-MUX, this implies that all switch modules connected to this module need to be coordinated in such a way that only one of the incoming fibers carries a valid signal in any given time slot. To achieve this, the configuration shown in Figure 4 requires arbiters to be connected along all dimensions.

In addition, as above, the signal is received by a switch module different from the switch module that transmits the signal, but for convenience and brevity, the receiving-side Rx process will be described here with reference to the same drawing. The demultiplexer WDM-DEMUX demultiplexes the optical signals received from the 1xK demultiplexer into the same Q signals entering the multiplexer WDM-MUX. One of the demultiplexed optical signals is then incident on each of the photodetectors PD1, PD2, . . . , PDQ, which convert the optical signal back into a corresponding electrical signal. Each of the photodetectors PD1, PD2, ..., PDQ outputs the electrical signal to one of the three electrical domain demultiplexers DEMUX, thereby demultiplexing into its two constituent electrical signals, namely an electrical signal. The signal originally came from (in Figure 4) the first fabric and an electrical signal originally came from the second fabric. Each of the three electrical domain demultiplexers DEMUX outputs two signals, packet fragmentation from the packet processor PP-Tx on each of the fabric sections. The three received packet fragments for each fabric part are then combined on the packet processor PP-Rx on each fabric part to reproduce the original packet originally incident on the source switch module. Thereafter, the packets may be transmitted to another fabric part for transmission into another dimension, or to a client part for transmission to an external device connected to the optoelectronic switch.

The configuration shown in Figure 4 requires time division multiplexing as well as wavelength division multiplexing in order to grant access to each dimension to the fabric part. This can be done using strict time-division multiplexing rules, ie going through successive fabric parts in sequence. Alternatively, time division multiplexing can be performed in a more flexible manner - as long as only a single fabric part is eligible to send a signal in a given time slot. In an alternative and more advanced configuration, it is possible to split the available wavelengths and then have multiple fabric parts transmit simultaneously, but at different wavelengths. Then, by using a cyclic AWG for the multiplexer WDM-MUX and the demultiplexer WDM-DEMUX, more than one fabric portion can be sent along different dimensions simultaneously by using disjoint subsets of the available wavelengths.

Figure 5 shows an example of a 1D optoelectronic switch. This example demonstrates the basic connectivity of the optoelectronic switch architecture of embodiments of the present invention, as well as the notation that can be used to conveniently describe more complex multi-dimensional optoelectronic switches.

Each of the small squares in the middle row of the diagram represents a single switch module as shown, for example, in FIGS. 3 and 4 . The ovals below these squares represent client ports that can be connected to external devices. As for the description of the connectivity of the switch modules, the fabric part and the client part are mostly independent of each other, so the fabric part and the client part will not be discussed in the following description. The switch module represents the smallest building block of an optoelectronic switch of some embodiments, and is referred to herein as a layer 0 switch. In the following description, a layer-i switch, where i>0, is an active switch that provides connections between switch modules (ie, layer-0 switches) along the i-th dimension, i.e. at the same coordinates (i-th except for the coordinates in the direction) between the switch modules. Each of the layer 0 switches (marked with S1 ) is connected to an optical active switch (marked with S2 ) represented by a long rectangle. This switch is referred to in this figure as a layer 1 switch and has 8 bidirectional inputs/outputs.

换句话说，t_i等于向量R中的不涉及第i维度的项的乘积。The following notation will be used to describe the array of layer 0 switches in various configurations/architectures of embodiments of the present invention. The switch fabric as a whole can be described using the notation (N, R), where N is the number of layers of optical switches in the switch fabric, which is equal to the number of dimensions, and R is of the form {R ₁ ,R ₂ ...R _N } A vector of , giving the cardinality of each layer, which is the same as the "size" of the dimension as defined in the "Summary of the Invention" chapter, where the cardinality gives each active in layer i (i.e. layer i, where i >0) The number of layer 0 switches to which the switch is connected. More specifically, a tier is an array of active switches or switch modules. In the following description, a tier 0 switch represents a switch module (eg, as shown in FIG. 3 ), and a tier i switch (where i>0) represents an active switch. A layer includes all switches associated with switching optical signals within a given dimension, and thus, there are N layers in an N-dimensional switch. In this notation, the optoelectronic switch shown in Figure 5 can be described as a (N=1, R={8}) switch because the top-level switch is a level-1 switch, and the top-level switch is connected to all eight level-0 switches switch. The number of switches t _i in the i-th layer is equal to the product of the cardinality of each dimension up to the i-th dimension, i.e.

In other words, ti is equal to the product of the entries in the vector R that do not involve the _ith dimension.

Each individual optical switch can be labeled as follows: S(i;C), where i represents the layer at which the switch is located, eg, layer 0, layer 1, etc., and C is a vector with (N-1) elements corresponding to the layer i switch The position within its layer, in the coordinate system of the layer's cardinality, except for the layer to which the switch corresponds. For example, in a layer 3 network, the switches in layer 2 have C=(c ₁ , c ₃ ), where c ₁ and c ₃ represent the labels used for the switches on the layers.

Figure 6 shows a schematic example of a 2D optoelectronic switch classified as (N=2, R={8,4}) according to the above notation. In this particular embodiment, there are 32 (ie 8x4) layer 0 switches connected together. Each of the 32 layer 0 switches has two fabric ports, one fabric port for connecting to a switch in layer 1 and one fabric port for connecting to a switch in layer 2. Because the 32 layer 0 switches are organized into 4 groups of 8 switches, there are 4 layer 1 switches and 8 layer 2 switches. This diagram clearly demonstrates an important property of optoelectronic switches according to embodiments of the present invention, namely that the maximum number of optical hops required from one layer 0 switch to any other layer 0 switch is the number of layers in the switch architecture (ie, N) . As an example, consider the transfer of data from an S1-labeled switch to an S2-labeled switch, where hops are shown with thicker lines. First, data is transferred from switch S1 to switch S3 via switch S4. Next, in a second hop, data is transferred from switch S3 to S2 via switch S5. Thus, it can be seen that in switches according to embodiments of the present invention, data can be transferred in a series of optical hops, each time via a layer i switch in a different layer.

More specifically, at each stage, packets are passed from one layer 0 switch to another layer 0 switch, as described above with reference to FIG. 3, and then the packets may need to pass through the layer 0 switch itself before the next optical hop can take place. Electronic hopping for transmission; however, electronic hopping does not significantly slow down operation because electronic data transfer has lower latency, integrated switches have lower associated cardinality, and time-of-flight does not need to be considered. Furthermore, no external arbitration or control is required since the transfer is simply from one fabric part to another within the same layer 0 switch.

Figure 7 shows an alternative schematic diagram of a 2D optoelectronic switch now with (N=2, R={8,8}). This switch has the same properties as the switch shown in FIG. 6 . This switch also more clearly shows the interrelationship between the tier i switch and the tier 0 switch. In particular, it can be seen that tier 0 switches are arranged in an 8x8 array, with tier 1 switches associated with each row and tier 2 switches associated with each column. More specifically, since each tier 0 switch has a fabric portion associated with each of the tiers, it can be seen that a tier i (i≠0) switch provides a given tier 0 switch with every other tier 0 Routes between switches (which have the same coordinates in all layers except layer i). (where the coordinates within tier i are values in the range 0 to t _i -1, where t _i is the number of active switches in tier i). This can also be seen from Figure 8, which shows a 3D optoelectronic switch with (N=3, R={8,4,2}). Here, it is possible to get from any layer 0 switch to any other layer 0 switch with up to 3 optical hops, each optical hop via a different layer i (i≠0) switch. It can be seen that the 3D optoelectronic switch of Figure 8 is formed by placing two 2D switches as shown in Figure 6 side-by-side and introducing an array of 32 layers 3 to provide the required interconnecting rows. Thirty-two tier 3 switches are required because tier 3 switches are required for sets of tier 0 switches with the same coordinates in tier 1 and tier 2. Since there are effectively 2 groups each consisting of 8x 4 layer 0 switches, it can be seen that 32 layer 3 switches are required, each connected to a layer 0 switch in the first group and a second group A layer 0 switch in the . Thus, the Layer 3 switches shown in Figure 8 each have two connections. More simply, the number of switches in each layer i (i≠0) is equal to the product of the cardinality of each of the other layers i (i≠0).

9A to C illustrate further arrangements of optoelectronic switches according to embodiments of the present invention. In these examples, all active switches have the same cardinality, referred to herein as R. In the notation above, this is expressed as R={R,R,R}={4,4,4}.

In order for a layer i (i≠0) switch to operate correctly and send the optical signal to the correct destination layer 0 switch, the switch must be controlled by an arbiter. Figure 11 shows a schematic diagram of how the arbiter is connected to a layer i (i≠0) optical active switch. For example, the input to the arbiter is connected to the controller CTRL as shown in FIGS. 3 and 4 . These controllers CTRL receive input from the packet processors, eg PP-Tx and PP-Rx related to the intended destination to which the packet is incident. This information is then forwarded to the arbiter, which calculates the best operating scheme for the layer i (i≠0) optical active switch so that all signals reach the correct destination layer 0 switch, i.e. provide the data transfer route , so that each transmit side of the fabric part is paired with the correct receive side of the fabric part, providing non-blocking operation. This calculated operating scheme is then transmitted to the switch driver which drives and controls the operation of the layer i (i≠0) optical active switch in order to achieve an efficient exchange of the optical signals received at its input.

Figure 10 shows the connections between the layer 0 switch and the arbiter in an exemplary 3x3 optoelectronic switch. Much like tier i (i≠0) optical active switches, there is an arbiter associated with each sub-array of tier 0 switches with the same coordinates in all but one tier. In the 2D case, ie with only two layers, this means that each row is associated with a row arbiter RA and each column is associated with a column arbiter CA. Thus, data transfers between layer 0 switches in the same row can be controlled by the associated row arbiter RA, and subsequent optical hops between rows (via layer i (i≠0) optically active switches) can then be controlled by the associated column Arbiter CA to control. As explained elsewhere in this application, switches may be connected using optical or electronic switches. Where electronic switches such as shared memory are used, a separate arbiter as shown in Figure 10 may not be required.

FIG. 12 shows an example of the arrangement of MZIs inside an optical MZI cascade switch, which can be used as an optical active switch in an embodiment of the present invention. Solid rectangles indicate individual MZIs. Using the notation of the "Summary of the Invention" system, it can be seen that in this particular configuration, the MZI cascade switch has R _i =4=2 ² (ie n=2) inputs and outputs. The input side may consist of four 1x4 "trees" (one tree is highlighted with a dashed square), each tree comprising two levels of 1x2MZI. The output side has a mirrored arrangement. The internal two layers of 1x 2MZI are connected so that routing from all inputs to all outputs can be provided simultaneously in a non-blocking manner. In other words, there are 4 possible inputs-outputs between the four inputs and outputs! = Each of the 24 combinations will be accommodated by this MZI cascade switch. A switch driver such as that shown in Figure 11 is configured to control which of the 24 combinations should be selected by controlling the voltages applied to the electro-optical regions of each 1 x 2MZI.

FIG. 13A shows the arrangement of components that would be used if an electronic active switch was used instead of the optical active switch as shown in FIG. 12 . For simplicity, only one switch module is shown. The bidirectional link shown carries the multiplexed fabric output signal towards the (electronic) shared memory switch SMS. At the SMS, the signal is incident on a demultiplexer DEMUX, which is configured to split the multiplexed signal into a plurality of optical signals. The DEMUX has substantially the same structure as the MUX (shown in the enlarged view), but reversed. The equivalent of the module labeled "Rx" or "Tx" on the DEMUX acts as an optical-to-electronic (O/E) converter for converting an optical signal into a plurality of electronic signals, which are then exchanged by SMS to the correct output. The modules "Rx" or "Tx" then act as electronic-to-optical (E/O) converters to convert the switch electronic signals to optical signals, which are then multiplexed to form a multiplexed fabric input signal. This signal is then carried by optical (WDM) fibers to the correct switch module.

Figure 13C shows a setup similar to that in Figures 13A and 13B, where instead of using a single electronically active switch to connect switch modules located in sub-arrays of switch modules, multiple or a set of electronically active switches are used type switch. It should be noted that it is also possible to use a set of optical active switches, eg of the type described in the previous paragraph. Topologically, the two approaches are identical, and for brevity only the embodiment using an electronic active switch will be described in detail. Using multiple switches to interconnect the switch modules in each sub-array results in a larger bisection bandwidth. This situation will be best understood from a comparison between Figures 13B and 13C, which were drawn using a similar layout. In these examples, there are R sets of R switch modules, which can be viewed, for example, as a square array with R columns and R rows, (switch modules are labeled 1 to R ² ). In this particular case, the subarrays are:

Dimension 1: each of the R sets, containing R switch modules, and

Dimension 2: Sets of switch modules that are identically located within each of the R sets.

In the arrangement shown in Figure 13B, a single electronically active switch is used to connect all switch modules in a given sub-array, as also shown, for example, in Figure 9B. However, in an alternative embodiment, as shown in Figure 13C, instead of using a single electronically active switch to interconnect the sub-arrays, an array of S electronically active switches is used instead. In the embodiment shown, the connections between switch modules in a given sub-array via electronically active switches are in the form of a Claus network, and more specifically a folded Claus network, since the links are bidirectional . However, other network topologies may be used to interconnect the sub-arrays. In this embodiment, there are S electronic active switches within each group. Preferably, S is chosen equal to the number of client ports on each of the switch modules.

Figure 14A shows a two-dimensional example of connections in an optoelectronic switch according to an embodiment of the invention as described in more detail above, with the backbone of the second (blue) dimension not shown. In Figure 14A, the leaf switch is depicted only once in the collapsed configuration. Similarly, Figure 14B shows a folded representation of a two-dimensional optoelectronic switch according to the present invention. Figure 14C shows an alternative expanded representation of a 2-dimensional optoelectronic switch comprising an array of sixty-four leaf switches at the edges of the schema, and two sets of sixteen backbone switches at the center (each associated with the exchange of each of the dimensions). Red lines represent connections in one dimension, while blue lines represent connections in another dimension.

While exemplary embodiments of switch modules and optoelectronic switches have been specifically described and illustrated herein, many modifications and variations will be apparent to those skilled in the art. Accordingly, it should be understood that switch modules or optoelectronic switches constructed in accordance with the principles of the present invention may be implemented otherwise than as specifically described herein. The invention is also defined in the claims and their equivalents.

Claims (27)

1. A switch module for use in an optoelectronic switch, the switch module having: a client part for connecting to an input device or an output device; a first fabric portion and a second fabric portion each for processing signals and communicating with other switch modules, the first fabric portion having a transmit side and a receive side, The transmission side has: A transmission-side input for receiving a first electronic signal carrying information, the information including information about the destination switch module of the first electronic signal, the first electronic signal being received by an output of the second fabric portion or received from an input device via the client portion; a transmission-side conversion member for converting the first electronic signal into a first plurality of optical signals containing the same information; a transmit-side multiplexer for converting the first plurality of optical signals into multiplexed fabric output signals for transmission to an active switch, and The receiving side has: a receiving-side demultiplexer for receiving the multiplexed fabric input signal from the active switch and separating the multiplexed fabric input signal into a second plurality of optical signals; a reception-side conversion member for converting the second plurality of optical signals into second electronic signals, and a receiving-side output, which is used to send the second electronic signal to the transmitting-side input of the second configuration part, or send the second electronic signal via the client part to the output device. 2. The switch module of claim 1, wherein the transmit-side conversion means comprises a transmit-side packet processor configured to receive the first electronic signal in the form of a raw packet, the The original packet has a packet header containing destination information. 3. The switch module of claim 2, wherein the transport-side packet processor is configured to perform packet fragmentation, wherein: Data packets with the same destination module are arranged into frames of a predetermined size; and The data packet is divided into a plurality of packet fragments arranged in corresponding plurality of frames; wherein the receive-side conversion means includes a receive-side packet processor configured to reconstruct the packet from the packet fragments when the original packet is scattered in more than one frame. 4. The switch module of claim 3, wherein the switch module is configured to send the multiplexed fabric output signal in a series of consecutive bursts, each burst comprising packets and/or packets from a single frame Fragmentation so that each burst includes only packets and/or packet fragments with the same destination module, and pairs of sequential bursts are separated by time intervals. 5. The switch module of claim 2, wherein: the transmission-side conversion member includes a plurality of modulators; the transmit-side packet processor is configured to perform packet slicing, wherein a frame or packet is sliced into a first plurality of electronic signals; and The transmit-side packet processor is configured to send each of the first plurality of electronic signals to a different one of the plurality of modulators, whereby the first plurality of electronic signals are converted into the the first plurality of optical signals. 6. The switch module of claim 5, wherein the receive-side conversion member includes a plurality of photodetectors configured to convert the second plurality of optical signals into a second plurality of optical signals. electronic signals, and the receive-side conversion means further includes a receive-side packet processor configured to reassemble the second plurality of electronic signals into the second electronic signal. 7. The switch module of claim 6, wherein the transmit-side packet processor and/or the receive-side packet processor are connected to a controller for connection to an arbiter. 8. The switch module of claim 7, wherein the transmit-side packet processor is configured to send a request to the arbiter, the request identifying the destination switch module of the packet. 9. The switch module of claim 8, wherein the transport-side packet processor is configured to use a look-up table or otherwise to look up which output of the active switch to which the transport-side packet processor is connected corresponds to the destination switch module that is the subject of the request. 10. The switch module of claim 1, wherein either or both of the transmit-side multiplexer and the receive-side demultiplexer are arrayed waveguide gratings (AWGs). 11. The switch module of claim 1, wherein the switch includes means for connecting to an optical backplane. 12. An N-dimensional optoelectronic switch for transmitting optical signals from an input device to an output device, the optoelectronic switch comprising a plurality of switch modules as claimed in claim 1, the switch modules being interconnected, wherein: The switch modules are arranged in an N-dimensional array, the i-th dimension has a size R _i , i=1, 2...N, each switch module has a value giving its position relative to each of the N dimensions. associated coordinate set; Each switch module is a member of N sub-arrays S _i , each sub-array S _i includes R _i switch modules, and the coordinates of the R _i switch modules are only different with respect to their positions in the i-th dimension, and all of the R i switch modules have different coordinates. each of the N subarrays is associated with a different dimension; each of the switch modules is configured to generate a multiplexed fabric output signal, Each sub-array S _i also includes an active switch with R _i inputs and R _i outputs; Each input of each active switch is configured to receive a multiplexed fabric output signal from each of the R _i switch modules in the sub-array; and The active switch is configured to multiplex a fabric output signal from its _Ri based on destination information contained in the first electronic signal received at the transmit-side input of the switch module. Any of the inputs leads to any of the _Ri outputs, the active switch receiving the multiplexing fabric output signal from the switch module. 13. The optoelectronic switch of claim 12, wherein each switch module has at least N fabric parts, each of the N fabric parts being associated with a different sub-array _Si , the switch modules being members of the _subarray Si. 14. The optoelectronic switch of claim 12, wherein the active switch is located on and connected to an optical backplane, the optical backplane further comprising a switch for providing a connection between each switch module and each active switch. A plurality of optical links are connected, and the switch module shares the sub-array Si with each active _switch . 15. The optoelectronic switch of claim 12, wherein the active switch is an optical active switch or an electronic active switch. 16. The optoelectronic switch of claim 15, wherein the active switch is a Mach-Zehnder interferometer (MZI) cascade switch comprising a plurality of MZIs, each MZI having: an input coupler two arms separated at and two arms feeding the separate paths into output couplers in which the paths are recombined; and two output sections, the plurality of MZIs being arranged to provide a cascade from the MZIs The path from each input end to each output end of the switch. 17. The optoelectronic switch of claim 15, wherein each electronic active switch further comprises: an optical-to-electrical converter at each input for converting the multiplexed fabric output signal from an optical signal to an electronically actively switched signal; and an electrical-to-optical converter at each output, the electrical-to-optical converter for converting the electronically active switching signal into an optical signal in the form of the multiplexed fabric input signal; wherein the electronically active switch is configured to switch the electronically active switch signal from any of its _Ri inputs to any of its _Ri outputs, and in: The photoelectric converter includes: a demultiplexer configured to demultiplex the multiplexed fabric output signal into a first plurality of intermediate optical signals, and a plurality of corresponding photodetectors, the the plurality of photodetectors for converting each of the intermediate optical signals into intermediate electronically actively switched signals for switching to a desired output, and The electro-optical converter is configured to convert the plurality of converted intermediate electronically actively switched signals into a second plurality of intermediate optical signals, and further includes means for multiplexing the second plurality of intermediate optical signals to form the A multiplexer that multiplexes the input signal of the fabric. 18. The optoelectronic switch of claim 12, wherein each sub-array Si of switch _modules further comprises an arbiter configured to control the said arbiter based on destination information stored in the data packets to be switched Operation of the active _switches included in the sub-array Si. 19. The optoelectronic switch of claim 18, wherein the arbiter is connected to at least one of a transmit-side packet processor and a receive-side packet processor on each switch module in the _subarray Si, and is configured to receive a request from each of the transmit-side packet processors to which the arbiter is connected. 20. An N-dimensional optoelectronic switch for transmitting optical signals from an input device to an output device, the optoelectronic switch comprising a plurality of switch modules as claimed in claim 1, the switch modules being interconnected, wherein: The switch modules are arranged in an N-dimensional array, the i-th dimension has a size R _i , i=1, 2...N, each switch module has a value giving its position relative to each of the N dimensions. associated coordinate set; Each switch module is a member of N sub-arrays S _i , each sub-array S _i includes R _i switch modules, and the coordinates of the R _i switch modules are only different with respect to their positions in the i-th dimension, and all of the R i switch modules have different coordinates. each of the N subarrays is associated with a different dimension; each of the switch modules is configured to generate a multiplexed fabric output signal; Each sub-array _Si also includes one or more active switches arranged to provide connections between all of the switch modules in the sub-array; The input of each active switch is configured to receive a multiplexed fabric output signal from one or more of the R _i switch modules in the sub-array; and Each of the one or more active switches is configured to multiplex the multiplexer based on destination information contained in the first electronic signal received at the transmit-side input of the switch module. A fabric output signal is directed from any switch module in the sub-array to any other switch module in the sub-array from which the active switch receives the multiplexed fabric output signal. 21. The optoelectronic switch of claim 20, wherein the sub-array of R _i switch modules includes only a single active switch having R _i inputs and R _i outputs, and: Each input of the active switch is configured to receive a multiplexed fabric output signal from each of the R _i switch modules in the sub-array, each of the switch modules is configured to receive a multiplexed fabric output signal from one of the R _i outputs of the active switch, and The active switch is configured to, based on the destination information contained in the first electronic signal received at the transmit-side input of the switch module, multiplex a fabric output signal therefrom. Any of the _Ri inputs leads to any of the _Ri outputs, the active switch receiving the multiplexing fabric output signal from the switch module. 22. The optoelectronic switch of claim 20, wherein at least one sub-array of _Ri switch modules comprises a plurality of P _sub active switches arranged to form a connection to the _sub -arrays The network of each switch module in the sub-array and each other switch module in the sub-array. 23. The optoelectronic switch of claim 22, wherein: The value of P _sub is the same for all sub-arrays comprising multiple active switches, and/or all said switch modules of said at least one sub-array of R _i switch modules have the same number of client ports, and P The value of _sub is equal to the number of client ports on each of the switch modules. 24. The optoelectronic switch of claim 22, wherein each switch module in a given sub-array is connected to an active switch via an intermediate switch. 25. The optoelectronic switch of claim 24, wherein the intermediate switch is one of an optical active switch, an electronic active switch, and an electronic packet switch. 26. The optoelectronic switch of claim 24, wherein the intermediate switch is bidirectional. 27. The optoelectronic switch of any one of claims 24 to 26, wherein the switch modules, intermediate switches and active switches are arranged in one of: a folded Claus network, an unfolded gram A Rolls network, a partially folded Claus network, or a Claus-like network.

Images