Hardware resource pool communication network architecture and communication method

Abstract

The invention provides a hardware resource pool communication network architecture. The network architecture comprises a controller, in-pool interconnection equipment, inter-pool interconnection equipment and at least two resource pools. And the controller is in signal connection with the in-pool interconnection equipment and the inter-pool interconnection equipment. Each resource pool comprises at least two hardware devices, and the in-pool interconnection device is in signal connection with the at least two hardware devices in the single resource pool. And the inter-pool interconnection device is in signal connection with all hardware devices in the at least two resource pools. The controller is configured to calculate a connection topology according to a preset communication request. And in response to the fact that the connection topology is in-pool connection, controlling the interconnection equipment in the pool to open a corresponding path, and establishing connection between the hardware equipment needing to be connected. And in response to the fact that the connection topology is inter-pool connection, controlling the inter-pool interconnection equipment to open a corresponding path, and establishing connection between the hardware equipment needing to be connected. In this way, efficient utilization of the hardware resource pool is achieved.

Description

Technical Field

[0001] The present application relates to the field of communication data exchange technology, and in particular to a hardware resource pool communication network architecture. Background Art

[0002] Traditional data centers, with servers as the smallest hardware unit, suffer from low resource utilization. This problem stems from the excessive granularity of hardware resources. When a server (assuming it contains one CPU, eight GPUs, and 8TB of memory) is assigned a task requiring one CPU, six GPUs, and five TB of memory, the remaining GPUs and memory are left idle, unable to execute other tasks. The proposed disaggregated data center architecture provides a path to improving data center hardware utilization. In this architecture, different types of hardware resources are disaggregated from servers, and hardware resources of the same type are pooled and managed uniformly. Based on task requirements, the required resources are dispatched from the corresponding hardware resource pool and networked to form a virtual server dedicated to that task.

[0003] While disaggregated data center architectures maximize resource utilization, they also present significant challenges for the networks that carry communication between resource pools. A key challenge is achieving cross-node networking with minimal degradation in virtual server performance. This requires networks to provide ultra-low latency, ultra-high bandwidth, and exceptional stability. In the post-Moore's Law era, the trend of bandwidth growth for electrical switches is unsustainable. While it is possible to build high-port electrical switches using multi-stage switching architectures, this comes at the expense of a large number of interconnected modules, resulting in higher costs, power consumption, and latency, while also sacrificing scalability. Summary of the Invention

[0004] In view of this, the purpose of this application is to propose a hardware resource pool communication network architecture and communication method.

[0005] Based on the above-mentioned purpose, the present application provides a hardware resource pool communication network architecture, which includes: a controller, an intra-pool interconnection device, an inter-pool interconnection device and at least two resource pools. The controller is signal-connected to the intra-pool interconnection device and the inter-pool interconnection device. Each resource pool includes at least two hardware devices, and the intra-pool interconnection device is signal-connected to at least two hardware devices in a single resource pool. The inter-pool interconnection device is signal-connected to all hardware devices in at least two resource pools. The controller is configured to: calculate the connection topology according to a preset communication request. In response to the connection topology being an intra-pool connection, the intra-pool interconnection device is controlled to open a corresponding path and establish a connection between the hardware devices that need to be connected. In response to the connection topology being an inter-pool connection, the inter-pool interconnection device is controlled to open a corresponding path and establish a connection between the hardware devices that need to be connected.

[0006] In some embodiments, each resource pool includes at most N hardware groups, and each hardware group includes at most N hardware devices.

[0007] In some embodiments, the intra-pool interconnect device includes at most N intra-group connectors and at most N inter-group connectors. Each intra-group connector is signal-connected to all hardware devices in a hardware group, and each hardware device is connected to only one intra-group connector. Each inter-group connector is signal-connected to only one hardware device in all hardware groups, and each hardware device is connected to only one inter-group connector.

[0008] In some embodiments, the intra-pool interconnection device is a fast optical switch, which includes N first connection ports. The fast optical switch is connected to the hardware device through the first connection ports.

[0009] In some embodiments, each resource pool includes N hardware groups, each of which includes N hardware devices. An intra-group connector corresponds one-to-one with each hardware group, and the N first connection ports of the intra-group connector are sequentially connected to the N hardware devices within the hardware group. The N first connection ports of the inter-group connector are connected to one hardware device in each hardware group, and each hardware device is connected to only one inter-group connector.

[0010] In some embodiments, the network architecture includes at most N 2 +1 resource pool, each resource pool contains at most N 2 A hardware device.

[0011] In some embodiments, the inter-pool interconnection device includes at most N 2 +1 leaf connector and at most N 2 +1 spine connector. Each leaf connector is connected to all hardware devices in a resource pool, and each hardware device is connected to only one leaf connector. Each spine connector is connected to all leaf connector signals.

[0012] In some embodiments, the inter-pool interconnection device is an optical circuit switch, each of which includes at least N 2 +1 second connection port, the optical circuit switch is connected to the hardware device through the second connection port.

[0013] In some embodiments, the network architecture includes N 2 +1 resource pool, each resource pool includes N 2 Hardware devices. Leaf connectors correspond to resource pools one by one, and there are at least N leaf connectors. 2 +1 second connection port and N in the resource pool 2 Each hardware device is connected in sequence. Each spine connector is connected to all leaf connectors.

[0014] A hardware resource pool communication method includes: calculating hardware device resource call information based on a preset communication request. The resource call information indicates the type and quantity of hardware devices to be called by the communication request. A connection topology is calculated based on the resource call information, and a connection path is established based on the connection topology.

[0015] From the above description, it can be seen that the hardware resource pool communication network architecture provided by this application is connected to all hardware devices in the resource pool by setting up intra-pool interconnection devices and inter-pool interconnection devices respectively, thereby reducing the number of forwarding required for communication between different hardware devices; and then identifying the communication request by setting up a controller, so that the intra-pool connection and the inter-pool connection links are separated, thereby reasonably distributing the communication load. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in this application or related technologies, the following briefly introduces the drawings required for use in the embodiments or related technical descriptions. Obviously, the drawings described below are merely embodiments of this application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0017] Figure 1 A schematic diagram of the interconnection topology structure within the network architecture pool provided in an embodiment of the present application;

[0018] Figure 2 This is a schematic diagram of the network architecture inter-pool interconnection topology provided in an embodiment of the present application.

[0019] Description of the accompanying drawings: 1-intra-pool interconnection device; 11-intra-group connector; 12-inter-group connector; 2-inter-pool interconnection device; 21-leaf connector; 22-spine connector; 3-resource pool; 31-hardware device; 32-hardware group. DETAILED DESCRIPTION

[0020] In order to make the objectives, technical solutions and advantages of this application more clear, this application is further described in detail below in combination with specific embodiments and with reference to the accompanying drawings.

[0021] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the embodiments of the present application should have the usual meanings understood by people with ordinary skills in the field to which this application belongs. The "first", "second" and similar words used in the embodiments of the present application do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word cover the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0022] The present application provides a hardware resource pool 3 communication network architecture, the network architecture including: a controller, an intra-pool interconnection device 1, an inter-pool interconnection device 2 and at least two resource pools 3.

[0023] In this embodiment, resource pool 3 is used to pool different types of hardware. To meet different computing requirements, resource pool 3 can generally include one or more of CPU resource pool 3, GPU resource pool 3 and memory resource pool 3. By pooling physical hardware resources, the hardware in resource pool 3 can be freely called according to computing requirements, thereby achieving efficient utilization of hardware resources.

[0024] The intra-pool interconnection device 1 is used to connect the signals of all devices in a single resource pool 3, thereby realizing the interconnection of the hardware devices 31 in the resource pool 3, and the inter-pool interconnection device 2 is used to connect the signals of devices in different resource pools 3, thereby realizing the interconnection between different resource pools 3.

[0025] The controller is used to control the devices connected within the pool and the devices connected between pools, and adjust the connection topology according to the communication request to establish a connection between two or more devices.

[0026] The controller is signal-connected to the intra-pool interconnection device 1 and the inter-pool interconnection device 2.

[0027] Each resource pool 3 includes at least two hardware devices 31 , and the intra-pool interconnection device 1 is signal-connected to at least two hardware devices 31 in a single resource pool 3 .

[0028] The inter-pool interconnection device 2 is signal-connected to all hardware devices 31 in at least two resource pools 3 .

[0029] The controller is configured to calculate a connection topology according to a preset communication request.

[0030] In response to the connection topology being an intra-pool connection, the intra-pool interconnection device 1 is controlled to open a corresponding path, and establish a connection between the hardware devices 31 that need to be connected.

[0031] In response to the connection topology being an inter-pool connection, the inter-pool interconnection device 2 is controlled to open a corresponding path, and establish a connection between the hardware devices 31 that need to be connected.

[0032] The type of the intra-pool interconnection device 1 is not limited and can be reasonably selected according to actual application needs. For example, it can be one of a router, an electrical switch, and an optical switch.

[0033] The type of the inter-pool interconnection device 2 is not limited and can be reasonably selected according to actual application needs. For example, it can be one of a router, an electrical switch, and an optical switch.

[0034] In some embodiments, each resource pool 3 includes at most N hardware groups 32 , and each hardware group 32 includes at most N hardware devices 31 .

[0035] In this embodiment, in order to facilitate the calling of the hardware devices 31 contained in the resource pool 3, by setting up to N hardware groups 32, and making each hardware group 32 contain up to N hardware devices 31, the numbers of the hardware devices 31 are matrixed for easy calling and management.

[0036] In some embodiments, as Figure 1 As shown, the intra-pool interconnection device 1 includes at most N intra-group connectors 11 and at most N inter-group connectors 12 .

[0037] Each intra-group connector 11 is signal-connected to all hardware devices 31 in one hardware group 32 , and each hardware device 31 is connected to only one intra-group connector 11 .

[0038] Each inter-group connector 12 is signal-connected to one hardware device 31 in all the hardware groups 32 , and each hardware device 31 is connected to only one inter-group connector 12 .

[0039] In this embodiment, by providing N intra-group connectors 11, all hardware devices 31 in at most N hardware groups 32 can be signal-connected to the intra-group connectors 11. By providing N inter-group connectors 12, all N hardware groups 32 in the resource pool 3 can be connected.

[0040] As an optional connection method, all hardware in the hardware resource pool 3 can be numbered, and the inter-group connectors 12 can be numbered separately, that is, the inter-group connector 121 is connected to all the first hardware in at most N hardware groups 32, the inter-group connector 122 is connected to all the second hardware in at most N hardware groups 32... the inter-group connector 12N is connected to all the Nth hardware in at most N hardware groups 32.

[0041] In some embodiments, the intra-pool interconnection device 1 is a fast optical switch, which includes N first connection ports. The fast optical switch is connected to the hardware device 31 through the first connection ports.

[0042] In this embodiment, in order to ensure the data exchange efficiency of the interconnected devices 1 in the pool and reduce the data exchange delay, a fast optical switch (FOS) is selected as the intra-group connector 11 in this application.

[0043] Fast optical switches offer switching speeds in the nanosecond range and are used in packet switching scenarios. A typical implementation involves integrating optoelectronic materials such as optical waveguides onto a switching chip. Applying an electric field across the waveguides changes the material's refractive index, enabling rapid switching of optical paths. This optical path switching method enables extremely short switching times, on the order of tens of nanoseconds. However, due to the integration technology and control methods, fast optical switches have a relatively low number of ports.

[0044] In some embodiments, each resource pool 3 includes N hardware groups 32 , and each hardware group 32 includes N hardware devices 31 .

[0045] The intra-group connector 11 corresponds to the hardware group 32 on a one-to-one basis, and the N first connection ports of the intra-group connector 11 are connected to the N hardware devices 31 in the hardware group 32 in sequence.

[0046] The N first connection ports of the inter-group connector 12 are connected to one hardware device 31 in each hardware group 32 , and each hardware device 31 is connected to only one inter-group connector 12 .

[0047] In this embodiment, all N first connection ports of the intra-group connector 11 and the inter-group connector 12 are connected to the hardware device 31 , thereby maximizing the utilization of the first connection ports of the intra-pool connector.

[0048] As an optional implementation, all hardware resources can be concentrated in several resource pools 3, so that as many resource pools 3 as possible include N hardware devices 31, and the number of hardware resources in other resource pools 3 is zero, and at most one resource pool 3 with a number of hardware resources between 1 and N is included, thereby canceling the deployment of pool connectors in resource pools 3 with zero resource quantity, thereby reducing the deployment cost of resource pool 3.

[0049] In some embodiments, as Figure 2 As shown, the network architecture includes at most N 2 +1 resource pool 3, each resource pool 3 includes at most N 2 31 hardware devices.

[0050] In this embodiment, in order to facilitate the call of the resource pool 3 contained in the network structure, by setting up to N 2 +1 resource pool 3, and make each resource pool 3 contain at most N 2 The hardware devices 31 are thus numbered in a matrix, which facilitates calling and management.

[0051] In some embodiments, the inter-pool interconnection device 2 includes at most N 2 +1 leaf connector 21 and at most N 2 +1 ridge connector 22.

[0052] Each leaf connector 21 is signal-connected to all hardware devices 31 in one resource pool 3 , and each hardware device 31 is only connected to one leaf connector 21 .

[0053] Each spine connector 22 is signal-connected to all leaf connectors 21 .

[0054] In this embodiment, the inter-pool interconnection device 2 adopts a leaf-spine / Dragonfly hybrid architecture for connection, which can improve the connection efficiency while ensuring that all hardware resources in all resource pools 3 can be interconnected.

[0055] The spine-and-leaf architecture, commonly known as the Clos design, is a two-tier network topology widely used in data centers and enterprise IT environments. Compared to the traditional three-tier network architecture, it brings multiple advantages to data center infrastructure, such as scalability, reduced latency, and improved performance.

[0056] Among them, Dragonfly is the most widely used direct-connect topology network architecture. It is characterized by a small network diameter and low cost. It has been widely used in high-performance computing networks and is also suitable for data center networks with diversified computing power.

[0057] In some embodiments, the inter-pool interconnection device 2 is an optical circuit switch, each of which includes at least N 2 +1 second connection port, the optical circuit switch is connected to the hardware device 31 through the second connection port.

[0058] In this embodiment, in order to ensure the communication bandwidth of the inter-pool connection, an optical circuit switch (OCS) is used as the inter-pool interconnection device 2 in this application.

[0059] Among them, optical circuit switches use a switching method similar to circuit switching. During a communication session, a dedicated optical path is established between the source node and the destination node, and this optical path remains dedicated throughout the session. Its implementation method is mainly through the dynamic configuration of paths in the optical fiber network by integrating optical switches (there are many types, including metal oxide, silicon photonic integrated circuits and other technologies). Through these switches, optical signals can be directed to the destination without optical-to-electrical conversion. The switching time of traditional OCS systems is usually in the millisecond level (about 10-100ms), which is mainly limited by the optical switch technology used. However, with the development of new technologies, research has realized OCS systems with switching times in the microsecond level (such as 11.5μs).

[0060] In some embodiments, the network architecture includes N 2 +1 resource pool 3, each resource pool 3 includes N 2 31 hardware devices.

[0061] The leaf connectors 21 correspond to the resource pools 3 one by one, and the leaf connectors 21 have at least N 2 +1 second connection port and N in resource pool 3 2 The hardware devices 31 are connected in sequence.

[0062] Each spine connector 22 is connected to all leaf connectors 21 .

[0063] In this embodiment, by making the ridge connector 22 and the leaf connector 21 at least N 2 +1 second connection ports are all connected to the hardware device 31, so that all the hardware devices 31 in the network architecture can be connected.

[0064] As an optional implementation, since the number of second connection ports of the optical circuit switch used by the inter-pool connector is usually much greater than the N provided by the fast optical switch, 2 First connection ports, therefore, the network structure provided in this application only needs to be set according to the number of interfaces of the fast optical switch.

[0065] As an optional implementation, in addition to being interconnected through the inter-pool interconnection device 2, the hardware devices 31 contained in each resource pool 3 can also be partially directly interconnected. The interconnection mechanism is as follows:

[0066] Number all resource pools 3, and record the i-th hardware device 31 in the j-th resource pool 3 as H i,j Then H i,j With H j,i+1 Direct interconnection, where i and j must satisfy 1≤j≤i≤N 2 , such as H 1,1 With H 1,2 Direct connection, H 3,2 With H 2,4 Direct connection.

[0067] This application also provides a hardware resource pool 3 communication method, including:

[0068] The resource calling information of the hardware device 31 is calculated according to the preset communication request, wherein the resource calling information is used to indicate the type and quantity of the hardware device 31 that needs to be called by the communication request.

[0069] The connection topology is calculated based on the resource call information, and a connection path is established based on the connection topology.

[0070] It can be seen from the above embodiments of the present application that the hardware resource pool 3 communication network architecture provided by the present application is connected to all hardware devices 31 in the resource pool 3 by setting up an intra-pool interconnection device 1 and an inter-pool interconnection device 2 respectively, thereby reducing the number of forwarding required for communication between different hardware devices 31; and then identifying the communication request by setting a controller, so that the intra-pool connection and the inter-pool connection links are separated, thereby reasonably distributing the communication load.

[0071] It should be noted that the embodiments of the present application can be further described in the following manner:

[0072] A hardware resource pool 3 communication network architecture includes: a controller, an intra-pool interconnection device 1, an inter-pool interconnection device 2 and at least two resource pools 3.

[0073] The controller is signal-connected to the intra-pool interconnection device 1 and the inter-pool interconnection device 2.

[0074] Each resource pool 3 includes at least two hardware devices 31 , and the intra-pool interconnection device 1 is signal-connected to at least two hardware devices 31 in a single resource pool 3 .

[0075] The inter-pool interconnection device 2 is signal-connected to all hardware devices 31 in at least two resource pools 3 .

[0076] The controller is configured to calculate a connection topology according to a preset communication request.

[0077] In response to the connection topology being an intra-pool connection, the intra-pool interconnection device 1 is controlled to open a corresponding path, and establish a connection between the hardware devices 31 that need to be connected.

[0078] In response to the connection topology being an inter-pool connection, the inter-pool interconnection device 2 is controlled to open a corresponding path, and establish a connection between the hardware devices 31 that need to be connected.

[0079] Optionally, each resource pool 3 includes at most N hardware groups 32 , and each hardware group 32 includes at most N hardware devices 31 .

[0080] Optionally, the intra-pool interconnection device 1 includes at most N intra-group connectors 11 and at most N inter-group connectors 12 .

[0081] Each intra-group connector 11 is signal-connected to all hardware devices 31 in one hardware group 32 , and each hardware device 31 is connected to only one intra-group connector 11 .

[0082] Each inter-group connector 12 is signal-connected to one hardware device 31 in all the hardware groups 32 , and each hardware device 31 is connected to only one inter-group connector 12 .

[0083] Optionally, the intra-pool interconnection device 1 is a fast optical switch, which includes N first connection ports. The fast optical switch is connected to the hardware device 31 through the first connection ports.

[0084] Optionally, each resource pool 3 includes N hardware groups 32 , and each hardware group 32 includes N hardware devices 31 .

[0085] The intra-group connector 11 corresponds to the hardware group 32 on a one-to-one basis, and the N first connection ports of the intra-group connector 11 are connected to the N hardware devices 31 in the hardware group 32 in sequence.

[0086] The N first connection ports of the inter-group connector 12 are connected to one hardware device 31 in each hardware group 32 , and each hardware device 31 is connected to only one inter-group connector 12 .

[0087] Optionally, the network architecture includes at most N 2 +1 resource pool 3, each resource pool 3 includes at most N 2 31 hardware devices.

[0088] Optionally, the inter-pool interconnection device 2 includes at most N 2 +1 leaf connector 21 and at most N 2 +1 ridge connector 22.

[0089] Each leaf connector 21 is signal-connected to all hardware devices 31 in one resource pool 3 , and each hardware device 31 is only connected to one leaf connector 21 .

[0090] Each spine connector 22 is signal-connected to all leaf connectors 21 .

[0091] Optionally, the inter-pool interconnection device 2 is an optical circuit switch, each of which includes at least N 2 +1 second connection port, the optical circuit switch is connected to the hardware device 31 through the second connection port.

[0092] Optionally, the network architecture includes N 2 +1 resource pool 3, each resource pool 3 includes N 2 31 hardware devices.

[0093] The leaf connectors 21 correspond to the resource pools 3 one by one, and the leaf connectors 21 have at least N 2 +1 second connection port and N in resource pool 3 2 The hardware devices 31 are connected in sequence.

[0094] Each spine connector 22 is connected to all leaf connectors 21 .

[0095] A hardware resource pool 3 communication method, comprising:

[0096] The resource calling information of the hardware device 31 is calculated according to the preset communication request, wherein the resource calling information is used to indicate the type and quantity of the hardware device 31 that needs to be called by the communication request.

[0097] The connection topology is calculated based on the resource call information, and a connection path is established based on the connection topology.

[0098] Those skilled in the art should understand that the discussion of any of the above embodiments is merely illustrative and is not intended to imply that the scope of the present application (including the claims) is limited to these examples. Within the scope of the present application, the technical features in the above embodiments or different embodiments may be combined, the steps may be implemented in any order, and there are many other variations of the different aspects of the embodiments of the present application as described above, which are not provided in detail for the sake of simplicity.

[0099] In addition, for simplicity of description and discussion, and in order not to make the embodiment of the application difficult to understand, the known power supply / ground connection with integrated circuit (IC) chip and other components may or may not be shown in the accompanying drawings provided. In addition, the device can be shown in the form of a block diagram to avoid making the embodiment of the application difficult to understand, and this also takes into account the following fact, that is, the details of the embodiment of these block diagram devices are highly dependent on the platform to be implemented in the embodiment of the application (that is, these details should be fully within the scope of understanding of those skilled in the art). When specific details (for example, circuit) are set forth to describe exemplary embodiments of the application, it will be apparent to those skilled in the art that the embodiment of the application can be implemented without these specific details or when these specific details are changed. Therefore, these descriptions should be considered to be illustrative rather than restrictive.

[0100] Although the present invention has been described in conjunction with specific embodiments thereof, many alternatives, modifications, and variations of these embodiments will be apparent to those skilled in the art based on the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) may utilize the embodiments discussed.

[0101] The embodiments of the present application are intended to cover all such substitutions, modifications, and variations that fall within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the embodiments of the present application should be included in the scope of protection of this application.

Claims

1. A hardware resource pool communication network architecture, the network architecture comprising: Controller, intra-pool interconnection devices, inter-pool interconnection devices, and at least two resource pools; The controller is signal-connected to the intra-pool interconnection device and the inter-pool interconnection device; Each of the resource pools includes at least two hardware devices, and the intra-pool interconnection device is signal-connected to the at least two hardware devices in a single resource pool; The inter-pool interconnection device is signal-connected to all the hardware devices in the at least two resource pools; The controller is configured to: calculate a connection topology according to a preset communication request; In response to the connection topology being an intra-pool connection, controlling the intra-pool interconnection device to open a corresponding path and establish a connection between the hardware devices that need to be connected; In response to the connection topology being an inter-pool connection, the inter-pool interconnection device is controlled to open a corresponding path, and a connection is established between the hardware devices that need to be connected.

2. The hardware resource pool communication network architecture according to claim 1, wherein: Each of the resource pools includes at most N hardware groups, and each of the hardware groups includes at most N hardware devices.

3. The hardware resource pool communication network architecture according to claim 2, wherein: The intra-pool interconnection equipment includes at most N intra-group connectors and at most N inter-group connectors; Each intra-group connector is signal-connected to all the hardware devices in one hardware group, and each hardware device is connected to only one intra-group connector; Each of the inter-group connectors is signal-connected to one of the hardware devices in all the hardware groups, and each of the hardware devices is only connected to one of the inter-group connectors.

4. The hardware resource pool communication network architecture according to claim 3, wherein: The intra-pool interconnection device is a fast optical switch, which includes N first connection ports. The fast optical switch is connected to the hardware device through the first connection ports.

5. The hardware resource pool communication network architecture according to claim 4, wherein: Each of the resource pools includes N hardware groups, and each of the hardware groups includes N hardware devices; The intra-group connectors correspond to the hardware groups one-to-one, and the N first connection ports of the intra-group connectors are connected to the N hardware devices in the hardware group in sequence; The N first connection ports of the inter-group connector are connected to one hardware device in each hardware group, and each hardware device is connected to only one inter-group connector.

6. The hardware resource pool communication network architecture according to claim 1, wherein: The network architecture includes at most N 2 +1 resource pool, each of which includes at most N 2 The hardware device.

7. The hardware resource pool communication network architecture according to claim 6, wherein: The inter-pool interconnection device includes at most N 2 +1 leaf connector and at most N 2 +1 ridge connector; Each leaf connector is signal-connected to all the hardware devices in one resource pool, and each hardware device is connected to only one leaf connector; Each of the spine connectors is signal-connected to all of the leaf connectors.

8. The hardware resource pool communication network architecture according to claim 7, wherein: The inter-pool interconnection device is an optical circuit switch, each of which includes at least N 2 +1 second connection port, the optical circuit switch is connected to the hardware device through the second connection port.

9. The hardware resource pool communication network architecture according to claim 8, wherein: The network architecture includes N 2 +1 resource pool, each of which includes N 2 Hardware devices; The leaf connectors correspond to the resource pools one by one, and the at least N 2 +1 second connection port and the N in the resource pool 2 Each hardware device is connected in sequence; Each of the spine connectors is connected to all of the leaf connectors.

10. A hardware resource pool communication method, comprising: Calculating hardware device resource call information according to a preset communication request; wherein the resource call information is used to indicate the type and quantity of the hardware devices required to be called by the communication request; A connection topology is calculated according to the resource call information, and a connection path is established according to the connection topology.

Images