IT201600073909A1 - CLUSTERING SYSTEM. - Google Patents

Description

CLUSTERING SYSTEM

FIELD OF THE INVENTION

The present invention relates to the networking of computer systems and components. In particular, networking of components for use in high-density computer networks.

BACKGROUND OF THE INVENTION

The grouping system of the present invention applies to computing in general and should therefore be considered in this context. However, the invention will generally be discussed in relation to server systems, being one of the most beneficial applications of the invention.

Servers are processing devices specially built and used to perform particular tasks. They are normally mounted on standard 19-inch server racks that offer a high-density support structure to house multiple servers, conventional server systems are active twenty-four hours a day to provide uninterrupted availability of resources. They can be used for a variety of computing purposes, for example, providing additional computing power over a network, storage, communications, electronics, web content, printing and gaming.

The key issue related to current server systems is the cost of running the server farm on an uninterrupted basis. Due to the high computing power density in a standard server farm, high power requirements must be met in order to maintain the system. In addition, the high density and continuous use leads to high temperatures that can damage server components, resulting in server downtime. As a result, high-powered cooling systems are a necessity within server farms, further increasing power requirements. One of the biggest costs associated with maintaining a server farm is the power required to run and cool the servers. As such, server systems are commonly rated on the basis of performance per watt. Other rating systems are also used, such as performance per unit of cost, depending on the user's needs.

In order to increase the performance per watt worth of server systems, Server Fods provide disassembled server computers with modular designs willing to make good use of the space. Multiple servers can be housed within a single server fag, which greatly increases performance per unit of volume, and decreases cooling requirements, improving performance per watt.

By standardizing server sizes, the size of a server is indicated in rack units, U, at 19 inches (480 mm) wide and 1.75 inches (44 mm) high. Common server racks have a large 42U form factor, which limits the number of servers that can fit into a single rack. Some high-end biade systems today can achieve a density of 2352 compute cores per 42U rack.

The present invention aims to reduce the density constraints of current systems by providing an ultra-high density modular cluster server system.

SUMMARY OF THE INVENTION

In order to reduce the limitations associated with the prior art, a computing cluster according to a first aspect of the present invention comprises: a motherboard comprising a plurality of interface card connectors; a plurality of interface cards adapted to engage with the connectors of the interface card, each interface card comprising a plurality of connectors of the computing module; and a plurality of computing modules, the plurality of computing modules being arranged to engage with the connectors of the computing module.

As it will be appreciated, the present invention provides several advantages over the prior art. For example, due to the hierarchical structure of the system, the cluster system can use low-power components with a relatively small footprint compared to typical server systems. This leads to a higher density of modules than current systems, greatly increasing the amount of computing power or space storage available per unit of floor area. Additionally, the use of lower electrical power modules reduces the cooling requirements of the system.

The plurality of interface cards can be arranged to communicate with each other through a network passage. At least one of the modules can be engaged with the module connector, so that it is positioned substantially parallel to the interface board. The motherboard may further comprise a network controller. The network controller can be arranged to receive inputs from the plurality of interface cards. Interface cards may include power, Ethernet, serial, enable, power fault, and ground pins. Module connectors can include pins for power, Ethernet, serial data, enable, shutdown, universal serial bus, serial peripheral interface, and ground. A cluster system may comprise at least one compute cluster, in accordance with the above; and a network card, in data communication with the or each of the computing clusters.

An interface card according to a second aspect of the invention, adapted to be inserted and powered by a motherboard, comprises: a plurality of connectors arranged to receive a plurality of processing modules; and a network controller arranged to drive the network between the interface card and the plurality of connectors, in which the network controller is also arranged to act as a network through between one or more other interface cards.

Thanks to the ability of interface cards to pass through the network, any interface card can communicate directly with any other interface card in the server system at no external cost, allowing for greater scalability beyond a single unit.

The plurality of connectors can be further arranged to contain one or each of the units positioned substantially parallel to the interface card. The network controller can be further organized to send inputs to the motherboard. The connectors may include pins relating to power, Ethernet, for serial data, for enabling, for shutdown, for power failures, for universal serial bus, for serial peripheral interfaces and grounding.

BRIEF DESCRIPTION OF THE DRAWINGS

Other advantages and benefits of the present invention will become evident from the examination of the following description and the attached drawings, in which:

Figure 1 shows an isometric view of a cluster, according to the present invention;

Figure 2 shows a circuit diagram of a motherboard according to the present invention;

Figure 3 shows a circuit diagram of a daughter card according to the present invention;

Figure 4 shows a circuit diagram of a processing module according to the present invention;

Figure 5 shows a circuit diagram of a memory module according to the present invention;

Figure 6 shows a circuit diagram of a network card according to the present invention;

Figure 7 shows a diagram of the general network structure according to the present invention; And

Figure 8 shows a schematic view of a cluster structure according to the present invention.

DETAILED DESCRIPTION OF THE FORMS OF REALIZATION

EXAMPLE

A cluster system according to an exemplary embodiment of the present invention is shown in Figures 1 to 8.

Figure 1 shows a cluster 100 comprising a motherboard 200, intended to serve as the base structure, as well as the network base and power of each cluster 100. The motherboard includes four 80-pin slot connectors 205 and a modular connector 201 to allow external Ethernet connections to cluster 100. Cluster 100 also includes four interface cards (also known as daughter cards) 300 that are individually connected to the motherboard via 80-pin slot 205 connectors. Each 300 daughter board includes four M.2 slot 303 connectors and is connected to four modules 400, 500, via the M.2 slot 303 connectors.

The purpose of the motherboard 200 is to distribute power and control data and communication channels within the cluster 100. Referring now to FIG. 2, the motherboard 200 has several communication interfaces as well as other components.

Modular connector 201 has higher bandwidth that can handle data and control information for motherboard 200 on a single channel. In this embodiment, the modular connector 201 is an RJ45 connector, however, it can, of course, be of any type of modular connector.

In an alternative embodiment (not shown), there are two modular connectors 201, which provide two data channels to the motherboard 200, one channel being used as the command channel and the other is for communicating data external to the local network. .

In one embodiment, the motherboard includes 202 Ethernet hubs that control the data channel and allow access to 300 daughter cards. 202 Ethernet hubs operate at the 10/100/1000 base, allowing for a maximum data transfer rate of 1 GB per second for cluster 100. A microcontroller 206, or network controller 206, controls the Ethernet hubs 202, although the input for the microcontroller 206 comes from the daughter cards 300 due to limitations in the Ethernet network.

In an alternate embodiment, the microcontroller 206 on the motherboard 200 is Ethernet enabled. In this embodiment, the microcontroller 206 can directly receive inputs, allowing a cluster 100 to monitor and control even if the interface cards 300 are in an "off" state.

The motherboard 200 also includes power connectors 203, to provide 12V input power to the motherboard 200. A power regulator 204 takes 12V input from the power connectors 203 and provides a regulated 3.3V output to the motherboard. motherboard electronics 200.

The 80-pin slot connectors 205 provide the communication interface between the motherboard 200 and the daughter boards 300. Each of the 80-pin slot connectors 205 has the following connections: four-pin 12V power; eight Ethernet pins forming two channels, three I C pins (or other Serial Bus); one enable pin for each daughter board; a power failure pin; and seven ground pins. The remaining pins of the 205 connectors are unallocated, available for use when configuring or customizing the system.

Referring now to Figure 3, the daughter board 300 is shown, which can be connected to the 80-pin slot connector 205 on the motherboard 200 with an 80-pin board connector 301. The 80-pin board connector 301 has corresponding specifications to the 80-pin connector of slot 205 described above.

The 300 daughter board has additional communication interfaces in four M.2 slot 303 connectors, two on each face of the 300 daughter board, keyed B, and a 6-pin 307 programming header. 1 M.2 303 slot connectors they are arranged with the slots at right angles so that the modules 400, 500, when connected, are substantially parallel to the daughter card 300. The key B encoding the M.2 303 slot connectors specifies where the orientation notch is located. Although the M.2 303 slot connectors have standard pin formats, in the present embodiment they are used in a bespoke configuration, which is: two 12V power pins; four pins of Ethernet for one channel; three pins I C (or other Serial Bus); an enable pin; a broken power pin; a power failure pin; two universal serial (USB) bus pins; three serial peripheral interface (SPI) bus pins; and six ground pins. This pin configuration allows for the possibility of multiple alternative communication lines to be used in alternative embodiments in addition to those described herein. It should be noted, however, that alternative encoding modes can be used.

In one embodiment, the daughter board 300 also includes an Ethernet hub port 5 302 for channel distribution of received data through the 80-pin edge connector 301 through each of the four M.2 slot connectors 303. In an alternate embodiment, the Ethernet hub 202 of the motherboard 200 communicates directly with the modules 400, 500, eliminating the need for an Ethernet hub 302 on the daughterboard 300.

It should be noted that, in alternative embodiments, the 80 pin edge connectors 205, 301 can be replaced with other types of connectors. For example, in one embodiment, the connector 205, 301 may be a small inline dual memory module outline (SODIMM), as this allows for orientation and control ratio.

An Ethernet-to-SPI 305 controller uses the control channel from the motherboard 200 and converts the SPI communication standard to be used by a 306 microcontroller (or 306 network controller). The microcontroller 306 then routes the control data for each connector of the M.2 slot 303 through the communication lines I C. The microcontroller 306 is also able to communicate with all the other microcontrollers 306, 206 of the cluster 100 via communication lines I C. As such, the microcontroller 306 of the daughter board 300 can communicate external instructions to the microcontroller 206 of the motherboard 200, which, in one embodiment, has no direct external connection, and transfers data to or from other daughter boards. 300 of cluster 100 or larger system. In this way, the microcontroller 306 of the daughter board 300 can act as a network pass through (or network interface), shunting data through other daughter boards 300 or motherboards 200. With the microcontroller 306 acting as a pass network, each board Daughter 300 can network directly with any other 300 daughter card or 200 motherboard, allowing for a simple network structure.

The 300 daughter board also includes a 12 V to 3.3 V 304 power regulator for the daughter board electronics.

The 300 daughter board also includes 308 hot swap controllers. These are located on the main power lines in each 300 daughter board and act as a switch, allowing it to be turned on and off. The 308 controller also protects against overvoltage and overcurrent, as well as delayed start-up functionality. The 308 hot swap controller allows each 300 daughter card to be fully tuned, this allows for energy savings when a 300 daughter card is not needed, but also allows the 300 daughter card to be replaced without shutting down the entire system. The hot swap controller 308 receives control inputs from the motherboard 200 microcontroller 205.

Referring now to Figures 4a, 4b, 5a and 5b, the modules 400, 500 comprise on-board M.2 connectors 403, 503 arranged to engage the M.2 slot connectors 303 of the daughter board 300. A module 400, 500 can be used as a calculation module or as a memory module. Figure 4 shows a computation module 400, and Figure 5 shows a storage module 500.

A processing module 400 allows the cluster 100 to operate as an ultra-high-density processing platform. As mentioned above, this could be for hardware server, or it could be, for example, desktop rendering. The specific processor used varies depending on the system requirements and, for example, it can be a CPU for a laptop / tablet due to the high performance and cost ratio, a high spec GPU or an FPGA.

The computing module 400 has a single communication interface, the M.2 connector being 403, the pin for which corresponds to the output pin described for the aforementioned daughter board 300. The computing module 400 also includes all the necessary electronics required for the operation of the processor 406: a power regulator 404; a RAM 401 for the processor; a communication converter 402; and a memory 405 for storing firmware.

In one embodiment, the communications converter 402 may be a USB Ethernet hub 402. A USB Ethernet hub 402 can be used in conjunction with a quad core processor to facilitate communication. However, other processors do not require this form of communication conversion.

The storage module of 500 acts as an ultra-high-density memory array, where, for example, each module may comprise one or more solid state drives (SSDs) associated with corresponding integrated 505 NAND Flash chips. 500 has M.2 503 pinout corresponding to the pin out described for the daughter board 300 above, and also includes all the electronics necessary for a functioning storage module: a power regulator 504; a CPU 501 to interface with the rest of the cluster 100; and one or more integrated NAND Flash chips for storing 505 data. The SSD chips associated with the integrated NAND Flash 505 chips can be up to 2TB in size with current technologies. The CPU 501 only needs a basic storage controller or a low spec processor used for direct data to the memory of a 500 module.

Figure 6 shows a network card 600 according to the present invention. The network card 600 includes modular connectors 601 for receiving data and control signals and cluster connectors 602 for transmitting data and control signals to the clusters 100. The network card 600 also includes communication hubs 604 to control the flow of data. signals.

The network card 600 also includes inter-card connectors 605 arranged to facilitate communication with other network cards 600. In this way, a number of identical network cards 600 can be connected together for control of network networks. systems of any size. The network cards 600 are adapted to communicate directly with any cluster 100 connected to the card 600, through the cluster connectors 602, and also with any other cluster 100 connected to any other card 600, through the inter-card connectors 605.

Each 604 communication center includes four ports to facilitate communication within the 600 network card. One port supports Γ data feed to / from a cluster 100 via a 602 cluster connector. One port supports an external communication power supply across the modular connectors 601. Two ports support communication to / from other 600 network cards via 605 inter-card connectors.

In a preferred embodiment, the network card 600 is connected directly to two clusters 100, and thus requires two communication hubs 604 and two cluster connectors 602. However, in alternative embodiments, the network card 600 may be adapted to directly connect any number of clusters 100, with different numbers of communication hubs 604 and cluster connectors 602, as required.

In a preferred embodiment, data is transmitted over Ethernet using Ethernet hub 604 through RJ45 connectors 601, 602. In one embodiment, some of the connectors of cluster 602 are double RJ45 connectors, adapted to transmit data and communication signals. control on a single channel.

The 600 network card also includes a 606 power connector, a 607 power regulator and a 608 hot swap controller. The 608 hot swap controller allows the cluster 100 to be powered on / off and removed without affecting the rest of the system .

Figure 7 shows an example of a server system 700 comprises ten clusters 100 of the type described above. The 700 server system is designed to use cluster 100 mounted in a rack format. Several 100 clusters are equipped with additional infrastructure to operate in a server environment.

A power supply (PSU) 701 receives power from an external power input 702 and supplies power directly to each cluster 100 through the motherboard power connectors 203. The 701 power supply also provides power directly to a 600 network card (or a series of 600 interconnected network cards) and to a 705 system and fan management card.

The 600 network card (s) interfaces directly with each cluster 100 and the system and fan management card 705.

The 705 System Management Card and Fan Card sends and receives data through the 600 Network Card (s) to control the system temperature of the 700 Server. The 705 System Management Card and Fan Card also provides control of the 'system power on / off and regulation of the controlled with the fan. The server system also includes a frame for hosting the clusters.

The frame was designed to fit into a standard rack mount cabinet. Therefore, each server system 700 is 450 mm wide, which includes mounting rails for mounting the system 700 inside a cabinet. Each cluster is 80mm tall, and the server system can be standardized to a height of 2U, where U is a unit of height in a rack-mounted server (approximately 45mm).

As a consequence of the high density, modular nature and enhanced network capabilities of the present invention, conventionally high power, high cost components can be replaced with low power, low cost components networked directly with each other.

The depth of system 700 is dictated by the number of clusters 100. In a single server system 700, four clusters 100 can fit side by side along the width of the cabinet. As such, 700 server systems are designed in groups of four 100 clusters up to a maximum of 32 100 clusters, for a total of 512 modules 400, 500, in a single 2U chassis. Therefore, a 42U rack can contain up to 672 clusters 100, comprising 10752 modules 400, 500.

Figure 8 shows a schematic view of a cluster 100. The power connectors 203 of the motherboard 200 supply power to the components of the motherboard 200, for each daughterboard 300 and each module 400, 500. The 12V input is adjusted to a 3.3V output before sending it to the board components.

The data and control signals are received on the modular connector 201 of the motherboard 200. In one embodiment, these signals are transmitted over 10G Ethernet to the Ethernet hub 202. The Ethernet hub 202 sends command signals and signals of data to the microcontroller 206. In one embodiment, the data signals are transmitted over an Ethernet link to the SPI controller. In this preferred embodiment, the control signals can be sent directly to the microcontroller 206 as Ethernet is enabled.

The control signals are sent to the hot swap controller 308 of each daughter board 300, by the microcontroller 206, to control the on / off state. Ethernet communications are sent directly to the daughter cards 300 and modules 400, 500 from the Ethernet hub 202 on the motherboard 200. As with the motherboard 200, the data and control signals are sent to the microcontroller 306, which further routes the control data to modules 400, 500 via I C communication lines. Data signals are sent to modules 400, 500 via 1G Ethernet.

In one embodiment of the present invention, a daughter board 300 may comprise four quad core ARM® CPUs with 4GB of RAM, one CPU for each processing module 400. Each cluster 100 of four daughter boards 300 requires approximately 16W of power. Taking all the peripheral components into consideration, a system 700 comprising 32 clusters 100 will have a theoretical power endowment of about 600W. Therefore, a rack of 10752 compute modules 400 will have a theoretical total power allocation of 12.6kW providing 43TB of RAM.

Due to the modular nature of the modules 400, 500, and the related structure, each cluster 100 can contain any combination of processing modules 400 and storage modules 500 according to the needs of a user. For example, a single daughter board 300 could carry two processing modules 400 and two storage modules 500, four processing modules 400, or four storage modules 500.

It is to be understood that although various specific connectors with particular pin formations are discussed in the description of the embodiments of the present invention, other connectors and pin formations could alternatively be used and, as such, the present invention is not to be regarded as limited to these specific connectors and pin formations. For example, although the present invention is described, in one embodiment, using an Ethernet to receive data and control channels, it is to be understood that an alternative network connection may be used.

Further, although the present invention has been described in connection with a particular embodiment in which a motherboard can comprise four interface cards, and an interface card can comprise four modules, it is to be understood that this is an example. In another embodiment, a daughter card may comprise six storage modules. In a further embodiment, a motherboard can comprise five interface cards. The present invention is not to be considered as limited to the specific structure described above.

Claims (14)

CLAIMS 1. A computing cluster (100), comprising: a motherboard (200) comprising a plurality of interface card connectors (205); a plurality of interface cards (300) arranged to engage with the connectors of the interface card (205), each interface card (300) comprising a plurality of module connectors (303); And a plurality of modules (400,500), the plurality of modules (400, 500) being adapted to be engaged with the connectors of the modules (303). A computing cluster (100) according to claim 1, wherein the plurality of interface cards (300) are adapted to communicate with each other through a network passage. A computing cluster (100) according to claim 1 or 2, wherein at least one of the modules (400, 500) is engaged with a connector of the module (303) such that it is positioned substantially parallel to the interface board (300) . A computing cluster (100) according to any one of claims 1 to 3, wherein the motherboard (200) further comprises a network controller (206). A computing cluster (100) according to claim 4, wherein the network controller (206) is arranged to receive inputs from the plurality of interface cards (300). A computing cluster (100) according to any one of claims 1 to 5, wherein the connectors of the interface board (205) include pins relating to power, Ethernet, serial data, enable, power failure and grounding. A computing cluster (100) according to any one of claims 1 to 6, wherein the module connectors (303) include pins relating to power, Ethernet, serial data, enable, power failure, serial bus universal, serial peripheral interface and grounding. A computing cluster (100) according to any one of claims 1 to 7, wherein each of the plurality of interface cards (300) comprises a hot swap controller (308) adapted to control power supply in the plurality of interface cards ( 300). 9. Cluster server system (700), comprising: at least one computing cluster (100), according to any one of claims 1 to 8; and a network card (600), in data communication with the or each of the computing clusters (100). 10. An interface card (300), adapted to be inserted and powered by a motherboard (200), comprising: a plurality of connectors (303), adapted to receive a plurality of modules (400, 500); And a network controller (306), suitable for network control between the interface card (300) and the plurality of connectors (303), wherein the network controller (306) is also arranged to act as a network through between one or more interface cards (300). An interface card (300) according to claim 10, wherein the plurality of connectors (303) are further arranged to contain one or each of the modules (400, 500) positioned substantially parallel to the interface card (300). An interface card (300) according to claims 10 or 11, wherein the network controller (306) is further adapted to send input to the motherboard (200). An interface card (300) according to any one of claims 10 to 12, wherein the connectors (303) include pins relating to power, Ethernet, serial data, enable, shutdown, power fail, bus universal serial, peripheral serial interface and grounding. An interface card (300) according to any one of claims 10 to 13, wherein the interface card (300) further comprises a hot swap controller (308) adapted to control power supply in the interface card (300).