HK1138694B - Energy efficient ethernet method and energy efficient ethernet physical layer device - Google Patents

Description

Energy efficient ethernet method and energy efficient ethernet physical layer device

Technical Field

The present invention relates to an Ethernet system, and more particularly, to a system and method for allowing a legacy (legacy) Media Access Controller (MAC) to perform an Energy Efficient Ethernet (EEE) operation.

Background

The cost of electricity continues to rise, and this trend has developed rapidly in recent years. In such cases, many industries have become very sensitive to the impact of these rising costs. In this regard, one area in which attention is being given to increasing interest is the IT architecture area. Many companies look closely at the power usage of their IT systems to determine whether power costs can be reduced. For this reason, an industry has emerged that focuses on energy efficient networks, all with the aim of solving the problems caused by the ever increasing costs of using IT equipment (i.e. PCs, displays, printers, servers, network appliances, etc.).

One considerable problem in designing energy efficient solutions is the traffic model on the network links. For example, in the pauses between occasional bursts of data, many network links are typically idle, while in other network links there is regular or periodic low bandwidth traffic accompanied by bursts of high bandwidth traffic. Another consideration when designing energy efficient solutions is the degree of sensitivity of the flow to buffering and delay. For example, some traffic types (e.g., HPC clusters or high-end 24-hour data centers) are very delay sensitive, thereby causing difficulties in buffering. For these and other reasons, applying energy-efficient concepts to different traffic models requires different solutions. These different solutions seek to adapt the link, link rate and layers above the link to the preferred solution (based on different power costs and impact on traffic) which is also application dependent on itself.

It should be noted that EEE solutions typically require coordination between different layers. For example, the EEE mechanism may be implemented within a physical layer device (PHY) to hop the PHY between various power states. To support these different PHY power states, the MAC and its upper layers (including silicon, software, and firmware) also need to control its operation to implement EEE control strategies. Ideally, a device containing a MAC may be updated to accommodate such an EEE mechanism. Otherwise, any innovations made to the PHY to implement EEEs in systems containing legacy MAC silicon would be useless. Therefore, a mechanism is needed to allow legacy MACs to be able to accommodate a PHY that supports EEE.

Disclosure of Invention

According to an aspect of the present invention, there is provided an energy efficient ethernet method applied in a physical layer device, comprising:

detecting that a physical layer device needs to jump between different power consumption modes;

generating a pause frame in the physical layer device in response to a transition between different power consumption modes; and

transmitting the generated pause frame from the physical layer device to a medium access control device, wherein the generated pause frame instructs the medium access control device to reduce a number of traffic destined for the physical layer device to adapt the physical layer device to the hop.

Preferably, the hopping transitions to a low power idle mode.

Preferably, the hopping changes to a hopping to subset physical layer device mode.

Preferably, the physical layer device is a backplane device.

Preferably, the physical layer device is a twisted pair device.

Preferably, the detecting comprises listening to a traffic queue.

Preferably, the detecting comprises listening for a subsystem status.

Preferably, the hopping is part of a bi-directional symmetric hopping of the link.

Preferably, the hopping is part of unidirectional asymmetric hopping of the link.

According to one aspect of the present invention, there is provided an energy efficient ethernet method for linking an enhanced physical layer device to a legacy media access control device, comprising:

detecting that an enhanced physical layer device needs to jump to a low power consumption mode; and

in response to detecting the need, reducing traffic from a legacy media access control device to an enhanced physical layer device, wherein the reducing comprises signaling using a backpressure control mechanism supported by the legacy media access control device.

Preferably, the transmitting signal includes transmitting a pause frame.

Preferably, the detecting comprises listening to a traffic queue.

Preferably, the detecting comprises listening for a subsystem status.

Preferably, the transmission signal is not based on direct flow measurement.

Preferably, the transmission signal is not based on a command from a peer device (peer device).

Preferably, the backpressure control mechanism comprises on-chip memory buffering.

In accordance with one aspect of the present invention, there is provided an energy efficient ethernet physical layer device connected to a medium access control device, comprising:

the control module is used for generating a control signal to respond to the requirement that the physical layer equipment jumps to a low power consumption mode; and

and the pause generation module is connected with the control module and used for responding to the generated control signal, generating a pause frame and sending the pause frame to the medium access control equipment so as to reduce the inbound traffic sent from the legacy medium access control equipment to the physical layer equipment.

Preferably, the pause generation module is implemented by hardware.

Preferably, the pause generation module is implemented by software.

Preferably, the control module generates the control signal in response to a low power idle mode transition.

Preferably, the control module generates the control signal in response to a subset physical layer device mode transition.

Preferably, the device further comprises a buffer memory for storing traffic received from the medium access control device.

Preferably, the device further comprises a buffer memory for storing traffic received from the remote end.

Drawings

The foregoing brief description of the disclosure, in order to describe the invention and other advantages and features, will be described in detail with reference to specific embodiments thereof as illustrated in the accompanying drawings. It is appreciated that these drawings depict only typical embodiments of the invention and are therefore not to be considered limiting of its scope. The invention will be further described with reference to the accompanying drawings and examples, in which:

FIG. 1 is a schematic diagram of a controller according to a preferred embodiment of the present invention;

FIG. 2 is a diagram illustrating a switch according to a preferred embodiment of the present invention;

FIG. 3 is a diagram of a pause frame generated based on an energy efficient Ethernet PHY and assisted by a control policy using pauses in accordance with a preferred embodiment of the present invention;

FIG. 4 is a flow chart of the process of the present invention according to a preferred embodiment of the present invention;

fig. 5 is a diagram of an embodiment of an energy efficient ethernet PHY including receive buffering, according to a preferred embodiment of the present invention.

Detailed Description

A system and/or method for enabling Energy Efficient Ethernet (EEE) operation for legacy (legacy) media access control, substantially as shown in and/or described in connection with at least one of the figures, as set forth more completely in the claims.

Hereinafter, various embodiments of the present invention will be described in detail. Although specific implementations are discussed below, it should be understood that these descriptions are merely illustrative of the principles of the invention. Those skilled in the art will recognize that other components and configurations may be used without departing from the spirit and scope of the invention.

Ethernet technology has become increasingly popular and has been applied in a variety of environments (e.g., twisted pair, backplane, etc.). IEEE 802.3az Energy Efficient Ethernet (EEE) continues to evaluate various methods for reducing the amount of power consumed during periods of low link usage. In this process, a protocol may be defined that may respond to changes in network requirements to transition to and from a low power mode.

Generally, a subrate that reduces the link rate to the primary rate may reduce power consumption and thus may save power. In one embodiment, the subrate may be the zero rate, which saves the most energy.

One example of subrate (smoothing) is through the use of subset PHY techniques. In this subset PHY technique, a low link usage interval may be satisfied by hopping the PHY to a low link rate (which may be supported by a subset of parent PHYs). In one embodiment, the subset PHY technique may be implemented by turning off portions of the parent PHY to support operation at a low or subset rate. For example, a subset 1G PHY may be generated from a parent 10GBASE-T PHY by closing three out of four channels. In another embodiment, the subset PHY technique may be implemented by reducing the clock rate of the parent PHY. For example, the parent PHY has an enhanced core whose rate may be decreased 1/10 by a factor during low link usage and increased 1/10 when bursty data is received. In the example of 1/10, 1 enhanced core of 10G may jump down to the 1G link rate when idle and revert to the 10G link rate when data needs to be sent.

Another example of a subrate is through the use of Low Power Idle (LPI) techniques. LPI generally relies on changing an active channel to a silent channel when no data needs to be sent. Thus, when the link is shut down, the energy is also reduced. The refresh signal may be sent periodically to enable waking up from a sleep mode. In another embodiment, a synchronization signal may be used on the interfaces, namely the media independent interface (MDI) and the PHY/Media Access Control (MAC) interface, to wake up quickly from sleep mode and maintain frequency lock. For example, on the MDI interface for 10GBASE-T signals, a simple PAM2 pseudo-random bit sequence on line pair A (pair A) may be used in LPI mode. Such operation does not significantly increase power consumption. Typically, both the subset technology and the LPI technology involve shutting down or adjusting a portion of the PHY during low link usage.

Whatever low power mode the EEE PHY supports, when the EEE PHY is connected to the legacy MAC, the EEE mechanism within the PHY is disabled. It is therefore desirable to reuse existing MAC devices without damaging EEE PHY functionality during use of a new EEE PHY.

This feature of the present invention is very valuable because a large market share of the controller or switch chip integrates a MAC or legacy chip containing a MAC (but allows for external PHY connections). In such an environment, the external EEE PHY may be connected to an existing legacy MAC. By introducing a mechanism to enable EEE PHY functions to work with legacy MAC devices, existing legacy devices can take full advantage of the EEE without having to check the entire device.

To save power, the capacity of the link will be reduced. When the PHY is in a low power state, other layers above the PHY can burst out at full rate (initially negotiated at link startup). If the MAC and its higher layers are capable of supporting EEE, the subsystem above the PHY will have enough buffering to allow the link to revert to the original rate. In legacy systems that do not support EEE, subsystems on the PHY cannot use it in real time even if there is memory.

In accordance with the present invention, the functionality of a device (e.g., network switch, controller, etc.) that includes a MAC may be enhanced to accommodate EEE mechanisms. In one embodiment, a backpressure mechanism will be introduced that allows legacy MACs to be off (hold off) when the PHY is in a low battery state or is out of a low battery state. In addition, EEE control strategies may use a buffer system in the backpressure mechanism to trigger jumps to and from various PHY power states. In fact, according to one feature of the present invention, the chip memory that is routinely used for pause frames can be reused, so that the EEE does not need additional memory to buffer (during transitions between power states) and provide assistance to EEE control strategies.

Fig. 1 is a diagram of an embodiment of an apparatus including a MAC implemented as a controller according to a preferred embodiment of the present invention. In various embodiments, the controller may be part of a client (e.g., a notebook, desktop, or workstation), a server (e.g., an audio-video (AV) server, a high-performance computing (HPC) server), or a consumer device (e.g., HDTV, blu-ray, etc.). As shown, host system 130 is connected to integrated ethernet controller 110. Ethernet controller 110 further comprises a PHY111 connected to MAC 113. In the depicted embodiment, MAC 113 is coupled to PCI express device 116 through a memory controller, wherein MAC 113 is also coupled to buffer 114 and processor 115.

Fig. 2 is a diagram of an embodiment of a device including a MAC implemented as a network switch, in accordance with a preferred embodiment of the present invention. In various embodiments, switching system 200 may represent a router or any other device having multi-port switching functionality. In various examples, the switch may be a user level, SMB, enterprise level, metropolitan level, or access switch. In another embodiment, switching system 200 may represent a Voice over IP (VoIP) chip having a network interface (Port 0) and a PC interface (Port 1). In another embodiment, switching system 200 may represent Customer Premise Equipment (CPE) within a service provider access network that includes an optical Central Office (CO) interface (port 0) and multiple interfaces (ports 1-N) to connect homes and/or gateways (e.g., the CPE may simply be considered part of a media converter and/or home gateway). Further, switching system 200 may also represent an access point, such as a WLAN base station.

As shown, the switching system 200 includes a switch 210 that supports an internal port and a plurality of external ports 0-N through MAC and PHY interfaces. It should be appreciated that the internal interfaces supported may vary depending on the implementation. For example, a VoIP phone may include an internal interface, while a switch does not. As shown in fig. 2, switch 210 is also supported by buffer 220 and controller 230.

As shown, the PHY in fig. 1 and 2 is an enhanced EEE PHY device. These enhanced EEE PHY devices may be used into existing integrated ethernet controllers 110 or switching systems 200. From a system perspective, it is desirable to improve the performance of enhanced EEE PHY devices over the use of entirely new chips and associated software. For this reason, it is necessary to use the existing MAC device and the enhanced EEE PHY device together. As mentioned above, this situation represents the vast majority of the share today that encompasses the MAC market.

In accordance with the present invention, existing MAC devices can be used with enhanced EEE PHY devices by adding functions to the EEE enhanced PHY to generate pause frames that can be passed up the protocol stack to the control MAC of the PHY. To describe the features of the present invention, the technical solution of the present invention will now be described with reference to fig. 3.

As shown in fig. 3, server 310 communicates with switch 320 through an enhanced EEE PHY. Here, the enhanced EEE PHY in server 310 includes a pause circuit 312 for initiating and sending pause frames to the MAC in server 310.

Here, it should be noted that conventional systems typically transmit pause frames that are generated by a remote device (remote/link partner), such as switch 320. This is the case, for example, when the receive buffer in switch 320 overflows, a request is sent from switch 320 to server 310 over the link to request server 310 to stop the subsequent transfer. In the present invention, the pause frame is generated by the enhanced EEE PHY in the same device on the same side of the link where the transmission operation request is paused. Obviously, the pause frame may be triggered to be generated by the EEE control strategy. Another advantage of using pause frames is that they enter the system buffer to perform pause operations, which are typically much larger than other frames that may be input into the PHY.

As shown in fig. 3, pause frames are sent by pause circuitry 312 to the MAC in server 310. The pause frame is applied in the backpressure mechanism of the server 310. The backpressure mechanism is implemented by flow control (flowcontrol)314 within the MAC. After receiving the pause frame, flow control 314 stops subsequent transmissions until pause timer 316 times out. Suspending subsequent transmissions results in traffic being deposited in the buffer of server 310.

In one embodiment, the pause timer 316 may be set to the value indicated by the pause frame, thereby pausing transmissions for a specified period of time. For an EEE PHY entering or leaving a low power mode, the particular time period may be set to be sufficient to support completing a transition between two different PHY power states. In one example, the specific time period allows for re-establishing/maintaining synchronization with the EEE PHY when it returns to an active state from a low power mode. In one embodiment, upon receipt of a pause 0 frame, the paused transmission will resume.

As previously described, the generation of the pause frame by the pause circuit 312 is triggered by the EEE control strategy. As shown in fig. 3, the EEE control strategy may be at least partially implemented in the EEE control module 318 in the EEE PHY. In one embodiment, the entire EEE control strategy is contained in the EEE PHY. In another embodiment, the EEE control policy may be triggered entirely by a higher layer that has access to the traffic model but cannot control the buffer in real time.

During operation, the EEE control module 318 alerts the pause circuit 312 that a pause frame needs to be generated. For example, the EEE control module 318 may prompt the pause circuit 312 to generate a pause frame when the EEE control module 318 determines that the EEEPHY enters the low power mode. It should be appreciated that the decision to transition to or from the low power mode may be based on various EEE considerations. Typically, the EEE control mechanism has access to a number of devices and software in the protocol stack and on the link. Regardless of the type of EEE control strategy implemented, the EEE control module 318 may generate a trigger signal to trigger the pause circuit 312 to generate a pause frame for the MAC.

For example, assume that the 10G ethernet controller does not contain any hardware that supports the EEE enhanced PHY low power mode. The transition to the low power mode may be based on a request from a link partner or the device's own EEE control policy (e.g., triggered when PCIE enters an L1 state, buffer level touches the watermark, a change in traffic queue rate reaches a threshold, etc.). When the EEE enhanced PHY begins to transition to the low power mode, the EEE control module 318 instructs the pause circuit 312 to generate a pause frame and send it to the 10G Ethernet controller. The suspended backpressure mechanism prevents the local MAC from sending data to the EEE enhanced PHY. When the EEE enhanced PHY exits the low power mode (requested by a local control policy or link partner), the pause timer is set to 0 when the PHY returns and enters a steady state and is ready to send data.

To further describe the features of the present invention, reference is now made to FIG. 4. As shown, the process begins at step 402, where the EEE control strategy indicates a need to transition to a low power mode. It should be appreciated that the EEE control strategy may be based on an analysis of various link-related parameters at either end of the link. Regardless of the EEE control strategy used, the PHY will indicate that a transition to a low power mode is required.

Upon receiving the indication, the EEE control in the EEE PHY then signals a pause circuit in the EEE PHY to indicate a power mode transition at step 404. In response to the received signal, a pause circuit within the EEE PHY then generates a pause frame (step 406). The EEE PHY then transmits the generated pause frame to the legacy MAC at step 408. Upon receipt of the pause frame, the legacy MAC suspends transmitting traffic to the EEE PHY that issued the pause frame at step 410. This transmit pause state will continue until either the pause timer times out or a 0 pause frame is received. At this point, the EEE-related functions then begin and the EEE PHY transitions to a low power mode (e.g., LP1 or subset PHY mode) at step 412. In this low power mode, the EEE control strategy will monitor the conditions at step 414. In this monitoring process, the EEE control policy monitors whether the pause timer has timed out to determine whether it has expired or whether another pause frame should be sent. In addition, the EEE control strategy may also determine whether a 0 pause frame should be sent to jump out of this mode.

As described above, the EEE control strategy may be used to trigger the generation of pause frames, thus regulating existing backpressure mechanisms in a unique way. In another embodiment, a software mechanism may be used to simulate receipt of a pause frame generated by the EEE PHY. In this manner, a similar result is achieved in software without triggering a hardware suspend mechanism. In another embodiment, a PHY without a pause circuit may be used, with a software mechanism built into the PHY to simulate the reception of pause frames generated by an EEE PHY.

In one embodiment, the pause mechanism generated by the EEE PHY may be used in conjunction with other mechanisms (for handling traffic from the MAC) that cannot accommodate the PHY at this time. For example, the subject matter of the present invention can be used with other buffering mechanisms (either the MAC or the buffering mechanisms available among PHYs for receiving traffic destined for the PHY). These other buffers may be used to receive traffic from the conventionally generated pause frames, wherein use of the pause frames for EEE purposes is not excluded. Other buffers may also be used to reduce latency.

In one embodiment, as shown in fig. 5, the EEE PHY also includes a buffer on the Receive (RX) side. RX buffer 512 in EEE PHY510 is used to buffer traffic received from EEE PHY 520. One benefit of this RX buffering is that EEE PHY520 may send traffic to EEE PHY510 when the EEE PHY decides to transition to a low power mode. This hopping process is also accompanied by the generation of pause frames 532 at the local side, as described above. In this arrangement, the RX buffer 512 may be used to receive inbound data on the receive side to ensure that the pause frame 532 does not miss any packet data from the EEE PHY520 on the far-end side. In one embodiment, RX buffer 512 is a relatively shallow buffer that receives receive-side traffic while pause frame 532 (e.g., a 64 byte packet) is addressed to the MAC. Note that one way that RX buffering can be avoided is to wait Y seconds to monitor RX before sending the pause frame.

Another benefit of using RX buffer 532 in EEE PHY510 is that EEE PHY510 may be enabled to examine frames from EEE PHY520 to determine whether a pause frame is contained therein. This checking procedure is advantageous when the response of the remote side to the previous burst number is delayed. This traffic burst will cause EEE PHY520 to generate a conventional pause frame 534. Typically, the value of the pause frame 534 generated by the far side will be different from the pause frame 532 generated at the local side. For example, the value of the pause frame 534 may be less than the pause frame 532. In another embodiment, the remote side may have issued a pause frame that is greater in value than the pause frame issued by the local side. In these schemes, EEE PHY510 may interpret and rewrite the pause value for the pause frame sent to the MAC to accommodate (account for) other pause frames. Through this process, the EEE PHY510 may track various pause requests.

It should be noted that to accommodate jumbo frames (e.g., 9k) that exceed the size of RX buffer 512, EEE PHY510 may be used to generate conventional pause frames for transmission to the far-end side before transmitting the locally generated pause frames to the MAC.

It should be noted that the teachings of the present invention can be used in a variety of PHY/MAC interfaces. For example, the PHY/MAC signals of the present invention may be implemented in register-based communications with an out-of-band signaling mechanism over a connection unit interface (AUI), Media Independent Interface (MII), Serial MII (SMII), Reduced MII (RMII), Gigabit MII (GMII), Reduced GMII (RGMII), Serial GMII (SGMII), 10 gigabit MII (XGMII), 10-GAUI (XAUI), or similar interface. Further, the subject matter of the present invention can be used with a variety of PHY types (e.g., backplane, twisted pair, optical, etc.) and standard or non-standard link rates (e.g., 2.5G, 5G, 10G, etc.) as well as future link rates (e.g., 40G, 100G, etc.).

It should be noted that the subject matter of the present invention is applicable to symmetrical or asymmetrical applications of EEEs. In a symmetrical application of EEE, both directions of the link may hop in a consistent manner between multiple power consumption modes. In the asymmetric application of EEE, the two directions of the link can independently transition between the various power consumption modes.

The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents and improvements made within the spirit and principle of the present invention are intended to be included within the scope of the present invention.

Claims (9)

1. An energy efficient ethernet method for use in a physical layer device, comprising:

detecting that a physical layer device needs to jump between different power consumption modes;

generating a pause frame in the physical layer device in response to a transition between different power consumption modes; and

sending the generated pause frame from the physical layer device to a medium access control device, wherein the generated pause frame instructs the medium access control device to reduce the amount of traffic directed to the physical layer device to adapt the physical layer device to the hop; the pause frame is generated by an enhanced energy-efficient Ethernet physical layer device in the same device on the same side of the link, the pause frame can be triggered and generated by an energy-efficient Ethernet control strategy, and the energy-efficient Ethernet control strategy can be based on analysis of various link related parameters at any end of the link; the enhanced energy efficient ethernet physical layer device is connected to an existing legacy media access control device.

2. The method of claim 1, wherein the hopping transitions to a low power idle mode.

3. The method of claim 1, wherein the hopping transitions to a hopping to subset physical layer device mode.

4. The method of claim 1, wherein the physical layer device is a backplane device.

5. The method of claim 1, wherein the physical layer device is a twisted pair device.

6. The method of claim 1, wherein the detecting comprises listening to a traffic queue.

7. An energy efficient ethernet method for linking an enhanced physical layer device to a legacy media access control device, comprising:

detecting that an enhanced physical layer device needs to jump to a low power consumption mode; and

in response to detecting the need, reducing traffic from a legacy media access control device to an enhanced physical layer device, wherein the reducing comprises signaling using a backpressure control mechanism supported by the legacy media access control device; the transmitting signal includes transmitting a pause frame; the pause frame may be triggered by an energy efficient ethernet control policy, which may be based on an analysis of various link related parameters at either end of the link.

8. The method of claim 7, wherein the detecting comprises listening to a traffic queue.

9. An energy efficient ethernet physical layer device connected to a media access control device, comprising:

the control module is used for generating a control signal to respond to the requirement that the physical layer equipment jumps to a low power consumption mode; and

a pause generation module, connected to the control module, configured to generate a pause frame in response to the generated control signal, and send the pause frame to the mac device, so as to reduce inbound traffic sent from a legacy mac device to a physical layer device; the pause frame may be triggered by an energy efficient ethernet control policy, which may be based on an analysis of various link related parameters at either end of the link.