Data center, server communication method and system, product and medium

Abstract

The invention discloses a data center, a server communication method, a system, a product and a medium, relates to the field of servers, and aims to solve the problems of idle resources and difficult signal transmission in a multi-host server system, the server system comprises a plurality of computing groups, each computing group comprises a plurality of computing devices, and the computing devices are arranged in the computing groups. Any two computing devices in the computing group are connected with each other; the switching chips are configured to identify each computing device connected with the switching chips; and the first management component is connected with the first transmission interface on the switching chip, and the first management component is configured to acquire all the computing devices identified by the switching chip, distribute independent identifiers to all the computing devices, and configure routing tables of all the optical ports on the switching chip based on the independent identifiers, so that the switching chip forwards data according to the routing tables. According to the invention, the communication efficiency of the multi-host server system is improved, the resource idling phenomenon is reduced, and the overall performance of the data center is improved.

Description

Technical Field

[0001] The present invention relates to the field of servers, and in particular to a data center, a server communication method, a system, a product and a medium. Background Art

[0002] With the rapid development of cloud computing and informatization, the scale of data centers continues to expand, and the demand for servers is growing. Graphics Processing Units (GPUs) are often used in servers for accelerated computing. In a multi-host server system, a chassis contains multiple hosts, each of which is connected to multiple GPUs. Each host can usually only communicate with the GPU connected to it. Communication between GPUs under the host requires data forwarding through the host, which can easily cause idle resources and difficulty in signal transmission. This limitation not only affects the scalability of data centers, but also limits the potential of high-performance computing.

[0003] Therefore, how to provide a solution to the above technical problems is a problem that those skilled in the art need to solve at present. Summary of the invention

[0004] The purpose of the present invention is to provide a data center, server communication method, system, product and medium to improve the communication efficiency of a multi-host server system, reduce idle resources, and improve the overall performance of a data center.

[0005] In order to solve the above technical problems, the present invention provides a server system, comprising:

[0006] A plurality of computing groups, each of which includes a plurality of computing devices, and any two computing devices in the computing groups are interconnected; n switching chips, the switching chip being configured to identify each computing device connected to itself, the plurality of optical ports on the ith switching chip correspondingly connecting to the ith computing device in the plurality of computing groups, i=1, 2, ..., n; a first management component connected to a first transmission interface on the switching chip, the first management component being configured to obtain all the computing devices identified by the switching chip, assign independent identifiers to all the computing devices, and configure a routing table of all the optical ports on the switching chip based on the independent identifiers, so that the switching chip forwards data according to the routing table; the routing table including a correspondence between each optical port and the independent identifier of the computing device connected thereto.

[0007] Optionally, the server system also includes multiple hosts, each of which is connected to at least two corresponding computing groups. The host is configured to determine a first target computing device among all the computing devices, determine a second target computing device among all the computing devices connected to itself, and access the first target computing device through the second target computing device and a switching chip connected to the second target computing device.

[0008] Optionally, the first management component is configured to obtain all the computing devices identified by the switching chip, allocate address space to all the computing devices, allocate independent identifiers to all the computing devices within the address space, and configure the routing table of all the optical ports on the switching chip based on the independent identifier, so that the switching chip forwards data according to the routing table.

[0009] Optionally, the first management component is further configured to control n of the switching chips to perform a reset operation.

[0010] Optionally, the first management component is also configured to identify transmission information of each of the optical ports, and adjust the interface specifications and / or interface protocols of each of the optical ports based on the transmission information; the transmission information includes the device type of the computing device connected to the optical port and the data transmission rate of the optical port.

[0011] Optionally, the server system further includes a switching board, on which at least one switching chip is disposed, and a plurality of optical modules corresponding to the plurality of optical ports on the switching chip are also disposed on the switching board, and the computing device is connected to the optical port via the optical module.

[0012] Optionally, the server system also includes a first programmable device, which is disposed on the switch board, and the first programmable device is configured to reset the optical module disposed on the switch board where the first programmable device is located, so that the switching chip disposed on the switch board where the first programmable device is located can identify the computing device connected to the optical module.

[0013] Optionally, the first programmable device is configured to reset each of the optical modules connected to the switching chip on the switching board where the first programmable device is located when the switching chip on the switching board where the first programmable device is located is powered on, and / or, when receiving a reset signal sent by the first management component, reset each of the optical modules connected to the switching chip on the switching board where the first programmable device is located.

[0014] Optionally, the first programmable device is also configured to output a mode control signal; the mode control signal includes a first control signal corresponding to a host software control mode and a second control signal corresponding to a module hardware control mode; the optical module is configured to adjust the current power mode to a low power mode in response to the first control signal, and to adjust the current power mode to a high power mode in response to the second control signal.

[0015] Optionally, the optical module is further configured to adjust the current power mode to the high power mode in response to receiving a host switching signal when in the low power mode.

[0016] Optionally, the first programmable device is further configured to respond to an alarm signal output by the optical module and perform a corresponding protection operation.

[0017] Optionally, the server system also includes a second programmable device, which is configured to control the power-on of the first management component and output a power-on permission signal when receiving a power-on signal, and output a trigger signal to the first management component after receiving a power-on completion signal; the first programmable device is also configured to control the power-on of the switching board where it is located after receiving the power-on permission signal, and output the power-on completion signal when the power-on of the switching board where it is located is completed; the first management component is also configured to obtain all the computing devices identified by the switching chip in response to receiving the trigger signal, assign independent identifiers to all the computing devices, and configure the routing table of all the optical ports on the switching chip based on the independent identifier, so that the switching chip forwards data according to the routing table.

[0018] Optionally, the second programmable device is further configured to identify the presence information of each of the switch boards, and output a logic determination instruction according to the presence information; the first management component is further configured to execute a corresponding management operation in response to the logic determination instruction.

[0019] Optionally, the server system further includes an interface indicator board, which includes a plurality of status indicator lights; the second programmable device is further configured to control the plurality of status indicator lights to perform indication operations according to their respective corresponding current status information.

[0020] Optionally, the server system also includes: multiple fan modules; multiple first temperature acquisition modules, the first temperature acquisition modules are configured to collect first temperature information of the corresponding switching chip; multiple second temperature acquisition modules, the second temperature acquisition modules are configured to collect second temperature information of the corresponding computing group; a second management component, configured to regulate the rotation speed of the fan module based on all the first temperature information and all the second temperature information.

[0021] In order to solve the above technical problems, the present invention also provides a data center, comprising a plurality of server systems as described in any one of the above.

[0022] In order to solve the above technical problems, the present invention also provides a server communication method, which is applied to a server system as described in any one of the above texts, and the server communication method includes: identifying each computing device connected to itself through a switching chip in the server system; obtaining all the computing devices identified by the switching chip through a first management component in the server system, assigning independent identifiers to all the computing devices, and configuring routing tables of all optical ports on the switching chip based on the independent identifiers; after receiving data sent by a first computing device through any optical port of the switching chip in the server system, transmitting the data to a second computing device through a target optical port according to the routing table.

[0023] In order to solve the above technical problems, the present invention also provides a server communication system, which is applied to a server system as described in any one of the above texts, and the server communication system includes: an identification module, which is used to identify each computing device connected to itself through a switching chip in the server system; an establishment module, which is used to obtain all the computing devices identified by the switching chip through a first management component in the server system, assign independent identifiers to all the computing devices, and configure the routing table of all optical ports on the switching chip based on the independent identifier; a routing module, which is used to transmit the data sent by the first computing device through any optical port of the switching chip in the server system to the second computing device through the target optical port according to the routing table.

[0024] In order to solve the above technical problem, the present invention also provides a computer program product, including a computer program / instruction, which implements the server communication method described above when executed by a processor.

[0025] In order to solve the above technical problem, the present invention further provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the server communication method described above is implemented.

[0026] The present invention provides a server system, which first divides multiple computing devices connected to each host into device groups, and any two computing devices in the device group are interconnected to achieve communication efficiency between computing devices under the same host, and then sets n switching chips, and multiple optical ports on the i-th switching chip are correspondingly connected to the i-th computing device in multiple computing groups, and the interconnection between each device group is achieved through the switching chip. The first management component is responsible for obtaining all computing devices identified by the switching chip and assigning independent identifiers to these devices. This allocation method ensures the uniqueness of each computing device in the network. Based on the independent identifier of the computing device, the first management component configures the routing table of all optical ports on the switching chip. The routing table contains the corresponding relationship between each optical port and the independent identifier of the computing device connected to it, so that the switching chip can efficiently forward data according to the routing table. Through direct connection and routing forwarding of the switching chip, the host can access any computing device in the server system, making the communication of the multi-host server system more efficient, reducing the phenomenon of idle resources, and improving the overall performance of the data center.

[0027] The present invention also provides a data center, a server communication method, a server communication system, a computer program product, and a computer-readable storage medium. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0029] Figure 1 A schematic diagram of the structure of a server system provided by the present invention.

[0030] Figure 2 A schematic diagram of the interconnection of multiple computing devices provided by the present invention.

[0031] Figure 3 This is a structural schematic diagram of a switching board provided by the present invention.

[0032] Figure 4 This is a schematic diagram of the structure of another server system provided by the present invention. DETAILED DESCRIPTION

[0033] The core of the present invention is to provide a data center, server communication method, system, product and medium to improve the communication efficiency of multi-host server systems, reduce idle resources, and improve the overall performance of the data center.

[0034] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0035] First, please refer to Figure 1 The present invention provides a server system, comprising:

[0036] There are multiple computing groups 1, each computing group 1 includes multiple computing devices 11, and any two computing devices 11 in the computing group 1 are interconnected.

[0037] There are n switching chips SW, and the switching chip SW is configured to identify each computing device 11 connected to itself. The multiple optical ports on the i-th switching chip SW correspond to the i-th computing device 11 in multiple computing groups 1, i=1, 2, ..., n.

[0038] The first management component 21 is connected to the first transmission interface on the switching chip SW. The first management component 21 is configured to obtain all computing devices 11 identified by the switching chip SW, assign independent identifiers to all computing devices 11, and configure the routing table of all optical ports on the switching chip SW based on the independent identifiers, so that the switching chip SW forwards data according to the routing table; the routing table includes the corresponding relationship between each optical port and the independent identifier of the computing device 11 connected thereto.

[0039] In this embodiment, the multi-host server system includes multiple computing motherboards, each computing motherboard can be connected to multiple computing devices 11, and the multiple computing devices 11 connected to each computing motherboard are divided into at least one computing group 1. The multiple computing devices 11 in the same computing group 1 are interconnected in pairs, so that the computing devices 11 in the computing group 1 do not need to go through the host on the computing motherboard when interacting with each other. The computing devices 11 in the computing group 1 can be interconnected through a first communication bus, and the first communication bus includes but is not limited to a PCIe (Peripheral Component Interconnect Express, a standard high-speed serial bus for external device interconnection) bus.

[0040] The server system also includes n switching chips SW, each of which includes multiple optical ports, and multiple computing devices 11 in the computing group 1 are connected to the optical ports on the multiple switching chips SW in a one-to-one correspondence. In order to improve the utilization rate of the optical port resources on the switching chip SW, as an optional embodiment, when dividing the computing devices 11 connected to the same computing motherboard, they can be divided according to the number of switching chips SW. For example, if there are 4 switching chips SW in total, the 4 computing devices 11 connected to the same computing motherboard can be divided into a computing group 1, and the first computing device 11 in the computing group 1 is connected to any optical port of the first switching chip, the second computing device 11 in the computing group 1 is connected to any optical port of the second switching chip, the third computing device 11 in the computing group 1 is connected to any optical port of the third switching chip, and the fourth computing device 11 in the computing group 1 is connected to any optical port of the fourth switching chip, so that all computing devices 11 connected to the computing motherboards can be connected together through the switching chips SW. It can be understood that the computing device 11 and the switching chip SW can be connected via a second communication bus, which includes but is not limited to a PCIe bus. The first communication bus and the second communication bus can be the same or different, and can be selected according to actual engineering needs. This embodiment is not limited here. Figure 1 Only four computing groups 1 and two switching chips SW are shown, and each computing group is illustrated by two computing devices 11. The specific architecture can be arranged according to actual engineering needs.

[0041] Reference Figure 2 As shown, S0-S7 refers to 8 GPUs connected to a single computing motherboard (i.e., computing device 11 in this embodiment). There are a total of 8 computing motherboards, so there are a total of 8 groups of S0-S7 in the figure. Among them, the 4 GPUs S0, S3, S4 and S7 are interconnected inside the host. These 4 GPUs can be regarded as a whole. The 64 GPUs in the entire architecture can be divided into 16 groups, each with 4 GPUs. The data and information transmission between them can be transmitted through interconnected cables. Each of these 4 GPUs is connected to a different switching chip SW, such as Figure 2 As shown, S0 in the upper left corner is connected to the first switch chip, S4 is connected to the second switch chip, S3 is connected to the third switch chip, and S7 is connected to the fourth switch chip. Thus, each switch chip SW has 16 external interfaces, which are connected to a GPU of each GPU group (i.e., computing group 1 in this embodiment), so that all GPUs connected to the motherboard can be connected together through the switch chip SW. The switch chip SW is configured to be able to identify the computing device 11 connected to itself.

[0042] The first management component 21 is connected to the first transmission interface of each switching chip SW through a third communication bus. The third communication bus can be an I2C (Inter-Integrated Circuit) bus, etc. The transmission rate of the third communication bus can be lower than the transmission rate of the first communication bus and the transmission rate of the second communication bus. The first management component 21 can be an mCPU (micro Central Processing Unit). The first management component 21 can obtain the computing device 11 identified by each switching chip SW through the third communication bus, and then assign an independent identifier to each computing device 11. The independent identifier can be an independent number or an assigned independent address, and it is sufficient to ensure that the independent identifiers of each computing device 11 are different. Then, the routing table of all optical ports on each switching chip SW is configured based on the independent identifier, so as to determine which computing device 11 is connected to each optical port on each switching chip SW. When data is forwarded through the switching chip SW, the destination computing device can be determined through the routing table, so as to select the corresponding optical port for data transmission, thereby improving the efficiency and accuracy of data transmission.

[0043] As an optional embodiment, the optical port on the switch chip SW for connecting the computing device 11 is configured as PCIeEP (Endpoint) mode.

[0044] For example, it is assumed that there are 4 switch chips SW, namely the first switch chip, the second switch chip, the third switch chip and the fourth switch chip. The 8 GPUs (S0-S7) under each computing motherboard are divided into two computing groups 1, each group containing 4 GPUs. For example, the first group is S0, S1, S2, S3, and the second group is S4, S5, S6, S7. S0 in the first group is connected to an optical port of the first switch chip, S1 is connected to an optical port of the second switch chip, S2 is connected to an optical port of the third switch chip, and S3 is connected to an optical port of the fourth switch chip. S4 in the second group is connected to an optical port of the first switch chip, S5 is connected to an optical port of the second switch chip, S6 is connected to an optical port of the third switch chip, and S7 is connected to an optical port of the fourth switch chip. In each computing group 1, GPUs are interconnected through a first communication bus (such as a PCIe bus). For example, S0, S1, S2, and S3 are interconnected through a PCIe bus. The first management component 21 is connected to the switching chip SW through a third communication bus (such as an I2C bus) to obtain information about the computing device 11 identified by the switching chip SW. The first management component 21 assigns an independent identifier to each GPU, for example: the independent identifier of S0 is G1, the independent identifier of S1 is G2, and so on. The first management component 21 configures the routing table of all optical ports on the switching chip SW according to the independent identifier, including one optical port on the switching chip connected to G1 (S0) and another optical port connected to G5 (S4). When G1 (S0) needs to transmit data with G5 (S4), the data is first transmitted to the switching chip through the first communication bus (PCIe bus) in the computing group 1, and then transmitted to G5 (S4) through the optical port of the switching chip according to the routing table.

[0045] The multi-host server system architecture in this embodiment can achieve efficient and accurate data transmission, while improving the utilization rate of the optical port resources on the switching chip SW.

[0046] It can be seen that in this embodiment, first, multiple computing devices 11 connected to each host are divided into device groups, and any two computing devices 11 in the device group are interconnected to achieve communication efficiency between computing devices 11 under the same host. Then, n switching chips SW are set, and multiple optical ports on the i-th switching chip SW are correspondingly connected to the i-th computing device 11 in multiple computing groups 1. The interconnection between each device group is achieved through the switching chip SW. The first management component 21 is responsible for obtaining all computing devices 11 identified by the switching chip SW and assigning independent identifiers to these devices. This allocation method ensures the uniqueness of each computing device 11 in the network. Based on the independent identifier of the computing device 11, the first management component 21 configures the routing table of all optical ports on the switching chip SW. The routing table contains the corresponding relationship between each optical port and the independent identifier of the computing device 11 connected to it, so that the switching chip SW can efficiently forward data according to the routing table. Through direct connection and routing forwarding of the switching chip SW, the host can access any computing device 11 in the server system, making the communication of the multi-host server system more efficient, reducing resource idleness, and improving the overall performance of the data center.

[0047] Based on the above embodiments:

[0048] In an exemplary embodiment, the server system also includes multiple hosts, which are connected to corresponding at least two computing groups 1. The hosts are configured to determine a first target computing device among all computing devices 11, determine a second target computing device among all computing devices 11 connected to themselves, and access the first target computing device through the second target computing device and a switching chip SW connected to the second target computing device.

[0049] In this embodiment, each computing motherboard includes a host, and the host can access the computing devices 11 connected to other hosts through the computing devices 11 connected thereto, so that each host can read the information of all computing devices 11.

[0050] Assume that there are two hosts, host A is connected to the first computing group 1C1 (including computing devices 11 independently identified as G1~G4) and the second computing group 1C2 (including computing devices 11 independently identified as G5~G8), host B is connected to the third computing group 1C3 (including computing devices 11 independently identified as G9~G12) and the fourth computing group 1C4 (including computing devices 11 independently identified as G13~G16), the computing devices 11 in each computing group 1 are interconnected through a first communication bus (such as a PCIe bus), and the computing devices 11 of each computing group 1 are connected to the corresponding switching chip SW.

[0051] If host A needs to access G10 in computing group 1C3, host A first determines G5 in computing group 1C2 directly connected to it as the second target computing device. Host A sends a request to the switch chip SW connected to computing group 1C2 through G5. The switch chip SW forwards the request to the switch chip SW connected to computing group 1C3 according to the routing table. The switch chip SW that receives the request then forwards the request to G10. Host B accesses computing device 11 under host A: Assuming that host B needs to access G2 in computing group 1C1, host B first determines G13 in computing group 1C4 directly connected to it as the second target computing device. Host B sends a request to the switch chip SW connected to computing group 1C4 through G13. The switch chip SW forwards the request to the switch chip SW connected to computing group 1C1 according to the routing table. The switch chip SW that receives the request then forwards the request to G2. When G10 receives a request from host A, it can perform computing tasks and return the results to host A through the switching chip SW network. Similarly, when G2 receives a request from host B, it can also perform computing tasks and return the results to host B through the switching chip SW network.

[0052] In this way, each host can access and read the information of all computing devices 11 in the server system through its directly connected computing group 1 and switching chip SW network, realizing resource sharing and efficient data transmission.

[0053] In an exemplary embodiment, the first management component 21 is configured to obtain all computing devices 11 identified by the switching chip SW, allocate address space to all computing devices 11, allocate independent identifiers to all computing devices 11 within the address space, and configure the routing tables of all optical ports on the switching chip SW based on the independent identifiers, so that the switching chip SW forwards data according to the routing table.

[0054] In this embodiment, the first management component 21 allocates a unified address space for all computing devices 11 in the server system. This address space can be virtual or physical, but it must be large enough to allocate independent identifiers to all devices. Within the address space, the first management component 21 allocates a unique independent identifier to each computing device 11. This identifier can be a MAC (Media Access Control Address) address, an IP (Internet Protocol Address) address, or a custom number to ensure that the identifier of each device is unique.

[0055] For example, the first computing device 11 is assigned an identifier of 0x01, the second computing device 11 is assigned an identifier of 0x02, the sixteenth computing device 11 is assigned an identifier of 0x10, and so on, until each computing device 11 is assigned an independent identifier.

[0056] The first management component 21 configures the routing table of all optical ports on the switching chip SW according to the assigned independent identifier. The routing table will instruct each switching chip SW how to forward data according to the destination identifier of the data packet. Taking the computing device 11 as a GPU as an example, the optical port 0 of the first switching chip is connected to GPU 01 (identified as 0x01), and the optical port 1 of the first switching chip is connected to GPU 05 (identified as 0x05). The routing table entries may be as follows: destination identifier 0x01→optical port 0, destination identifier 0x05→optical port 1, and so on. When the switching chip SW receives a data packet, it will check the destination identifier of the data packet and determine which optical port should be used to forward the data according to the routing table. For example, if the first switching chip receives a data packet with a destination identifier of 0x02, it will look up the routing table and find that the data should be forwarded through the optical port connected to GPU2.

[0057] In this way, the first management component 21 ensures that the communication between all computing devices 11 in the server system is efficient and accurate, and each switching chip SW is able to quickly forward data according to the configured routing table without passing through the host processor, thereby realizing P2P (Peer-to-Peer) communication between GPU EP devices, reducing latency and improving the overall performance of the system.

[0058] In an exemplary embodiment, the first management component 21 is further configured to control the n switch chips SW to perform a reset operation.

[0059] In this embodiment, the first management component 21 first initializes a reset command, which will be sent to all switching chips SW to instruct them to perform a reset operation. Specifically, the first management component 21 sends a reset command to each switching chip SW through a control interface (such as I2C, SPI (Serial Peripheral Interface), network interface, etc.). This process can be performed in parallel or in sequence, and can be set according to actual engineering needs. This embodiment does not make specific limitations here.

[0060] After receiving the reset command, each switch chip SW sends a confirmation signal to the first management component 21, indicating that it has received the reset command. After receiving the reset command, each switch chip SW starts to perform a reset operation. The reset operation includes but is not limited to the following operations: disconnecting all current data connections and sessions; resetting internal states, including routing tables, caches, counters, etc.; resetting hardware registers to default states; and may also include hardware-level resets, such as power cycles.

[0061] The first management component 21 monitors the reset process of each switching chip SW to ensure that the reset operation is performed correctly. If an error occurs during the reset process or a switching chip SW does not respond, the first management component 21 can take further measures, such as resending the reset command or recording error information. Once the switching chip SW completes the reset operation, it sends a reset completion signal to the first management component 21. After the reset is completed, the first management component 21 may need to reconfigure the switching chip SW, including reallocating address space, independent identification, and configuring routing tables to ensure that the switching chip SW can work normally again.

[0062] In an exemplary embodiment, the first management component 21 is also configured to identify transmission information of each optical port and adjust the interface specification and / or interface protocol of each optical port based on the transmission information; the transmission information includes the device type of the computing device 11 connected to the optical port and the data transmission rate of the optical port.

[0063] In this embodiment, the first management component 21 first monitors the device type and data transmission rate of the computing device 11 connected to each optical port. This information can be obtained by querying the device or through the switching chip SW report. Based on the monitored transmission information, the first management component 21 analyzes the data transmission requirements of each optical port. For example, if a certain optical port is connected to a high-performance GPU, it may require a higher data transmission rate and a specific interface protocol. The first management component 21 determines which optical port interface specifications and / or interface protocols (PCIe or CXL (Compute Express Link, Compute Express Link), etc.) need to be adjusted based on the transmission requirements. This may include adjusting the data transmission rate, modifying the signal encoding method, changing the error checking mechanism, etc. The first management component 21 sends a command to the switching chip SW to perform specific adjustment operations, which may involve modifying the hardware register settings of the switching chip SW or updating the firmware configuration. This process ensures that each optical port can be optimized according to the characteristics of the computing device 11 to which it is connected, thereby improving the performance and efficiency of the entire system.

[0064] In an exemplary embodiment, the server system also includes a switching board, which is provided with at least one switching chip SW. The switching board is also provided with multiple optical modules 32 corresponding to the multiple optical ports on the switching chip SW, and the computing device 11 is connected to the optical port through the optical module 32.

[0065] In this embodiment, the hardware topology of the switching chip SW is as follows: Figure 3 As shown, the switch chip SW includes but is not limited to 18 X8 PCIe interfaces, of which 16 X8 PCIe interfaces are used as optical ports. Figure 3 PE0-PE7 are port numbers, each optical port is connected to a QSFP-DD connector, and each QSFP-DD connector is connected to the computing device 11 through an external optical module 32. The remaining two X8 PCIe interfaces are used to communicate with other switching chips SW as reserved interfaces. The clock signal of the switching chip SW comes from the clock generator clk generator on the board.

[0066] Among them, at least one switching chip SW is arranged on each switching board, and they are arranged according to a certain layout. The 16 optical ports on each switching chip SW are respectively connected to the QSFP-DD (Quad Small Form Factor Pluggable-Double Density) connector on the switching board, and the QSFP-DD connector is connected to the computing device 11 (such as GPU) through the optical module 32. The optical module 32 is responsible for converting electrical signals into optical signals to achieve high-speed data transmission.

[0067] In an exemplary embodiment, referring to Figure 3 The server system also includes a first programmable device 31, which is disposed on the switching board. The first programmable device 31 is configured to reset the optical module 32 disposed on the switching board where the first programmable device 31 is located, so that the switching chip SW disposed on the switching board where the first programmable device 31 is located can identify the computing device 11 connected to the optical module 32.

[0068] In this embodiment, the switching board may also be provided with a first programmable device 31, which includes but is not limited to a CPLD (Complex Programmable Logic Device). The first programmable device 31 controls the optical module 32 on the switching board where it is located to reset. Once the optical module 32 is reset, the first programmable device 31 notifies the switching chip SW to re-detect the computing device 11 connected to the optical module 32.

[0069] It can be understood that after the system is started or the optical module 32 is replaced, the reset operation can ensure that the optical module 32 returns to a known initial state. If the optical module 32 encountered an error or anomaly in the previous operation, the reset operation can help clear these error states so that the optical module 32 can resume normal operation. In some cases, the connection between the optical module 32 and the computing device 11 may be disconnected due to software or hardware problems, and the reset operation can re-establish these connections. The reset operation allows the switch chip SW to re-detect the existence and status of the optical module 32 so that the system can correctly identify all hardware components.

[0070] In an exemplary embodiment, the first programmable device 31 is configured to reset each optical module 32 connected to the switching chip SW on the switching board where the first programmable device 31 is located when the switching chip SW on the switching board where the first programmable device 31 is located is powered on, and / or, when receiving a reset signal sent by the first management component 21, reset each optical module 32 connected to the switching chip SW on the switching board where the first programmable device 31 is located.

[0071] In this embodiment, the management board equipped with the first management component 21 sends a reset signal (SW_PERST) to the switching chip SW for resetting the switching chip SW. In addition, 16 groups of RESET signals (reset signals) are connected to the first programmable device 31. After receiving the RESET signal, the first programmable device 31 sends it to 16 QSFP-DD optical modules. There are two mechanisms for the first programmable device 31 to send the RESET signal. One is self-control. After the switching chip SW is powered on and the clock is stable, it sends the RESET signal to the QSFP-DD optical module. This process ensures that the optical module 32 is in the correct initial state before the switching chip SW is ready. The other is to wait for the RESET signal sent by the management board and then send it to the 16 QSFP-DD optical modules. This mechanism allows the system administrator or the first management component 21 to remotely control the reset of the optical module 32.

[0072] In an exemplary embodiment, the first programmable device 31 is also configured to output a mode control signal; the mode control signal includes a first control signal corresponding to the host software control mode and a second control signal corresponding to the module hardware control mode; the optical module 32 is configured to adjust the current power mode to a low power mode in response to the first control signal, and to adjust the current power mode to a high power mode in response to the second control signal.

[0073] In an exemplary embodiment, the optical module 32 is further configured to adjust the current power mode to the high power mode in response to receiving a host switching signal when in the low power mode.

[0074] In this embodiment, the signal sent to the optical module 32 also includes an LPM (Low Power Mode) mode signal (i.e., a mode control signal in this embodiment), and the LPM signal allows the host to define whether the QSFP-DD optical module is initialized in the host software control mode (LPM is set to high) or the module hardware control mode (LPM is set to low). Under the control of the host software, the optical module 32 should remain in the low power mode until the software can transition to the high power mode defined in the QSFP-DD management interface specification. In the module hardware control mode, the optical module 32 can be converted to the high power mode immediately after the management interface is initialized.

[0075] Among them, in low power mode, the optical module 32 consumes less power and is usually used for initialization or standby state. In high power mode, the optical module 32 reaches full power operation and is suitable for data transmission. The host software control mode allows the system administrator to dynamically adjust the power mode of the optical module 32 according to actual needs, providing higher system flexibility and manageability. When it is necessary to switch from low power mode to high power mode for data transmission, the optical module 32 can quickly respond to the host switching signal to ensure the timeliness and efficiency of data transmission. In the module hardware control mode, the optical module 32 can be converted to high power mode immediately after the management interface is initialized, which helps to ensure the stability and security of the system. This power mode control mechanism provides more efficient energy management, better equipment performance and longer equipment life for data centers and network equipment, while enhancing the overall reliability and flexibility of the system.

[0076] In an exemplary embodiment, the first programmable device 31 is further configured to respond to the alarm signal output by the optical module 32 and perform a corresponding protection operation.

[0077] In this embodiment, the optical module 32 and the first programmable device 31 also transmit the in-position signal (PRSNT) of the optical module 32 and the alarm signal (INTL) of the optical module 32. When the first programmable device 31 receives the alarm signal of the optical module 32, it performs the corresponding protection operation. Specifically, when the optical module 32 detects an internal fault or abnormality, it triggers an alarm signal (INTL), which is sent to the first programmable device 31. After receiving the alarm signal, the first programmable device 31 performs a series of protection operations according to a preset program, including but not limited to resetting the optical module 32 to try to restore its function, removing the affected optical module 32 from service to avoid further data transmission errors, notifying the management system or host about the alarm event for further diagnosis and maintenance, and switching to a backup optical module 32 or path to maintain the continuity of data transmission.

[0078] Figure 3The illustrated switch board also includes a power connector, a management connector, a sensor, a replaceable switch, a bidirectional conversion switch, an arbitration module, etc.

[0079] In an exemplary embodiment, referring to Figure 4 , the server system further comprises:

[0080] The second programmable device 22 is configured to control the power-on of the first management component 21 and output a power-on permission signal when receiving a power-on signal, and output a trigger signal to the first management component 21 after receiving a power-on completion signal; the first programmable device 31 is also configured to control the power-on of the switching board where it is located after receiving the power-on permission signal, and output a power-on completion signal when the power-on of the switching board where it is located is completed; the first management component 21 is also configured to obtain all computing devices 11 recognized by the switching chip SW in response to receiving the trigger signal, assign independent identifiers to all computing devices 11, and configure the routing tables of all optical ports on the switching chip SW based on the independent identifiers, so that the switching chip SW forwards data according to the routing tables.

[0081] In this embodiment, a second programmable device 22 is also deployed on the management board. The second programmable device 22 includes but is not limited to CPLD. The second programmable device 22, the first programmable device 31 and the first management component 21 work together to ensure that the switching board and the switching chip SW are correctly powered on and configured for data forwarding.

[0082] Specifically, when the system is powered on, the second programmable device 22 receives a power-on signal, controls the power-on of other components on the management board, such as the power-on of the first management component 21, and outputs a power-on permission signal. The first programmable logic device receives the power-on permission signal, controls the power-on of the switch board where it is located, and when the power-on of the switch board is completed, the first programmable logic device outputs a power-on completion signal. After the second programmable device 22 receives the power-on completion signal, the second programmable device 22 outputs a trigger signal to the first management component 21. The first management component 21 receives the trigger signal output by the second programmable device 22, and executes the operation of obtaining all computing devices 11 identified by the switching chip SW. In this way, it can be ensured that the system components are powered on in the correct order to avoid hardware damage caused by improper power-on order. The first management component 21 is responsible for configuration and management, which simplifies the complexity of the system. The collaborative work between multiple programmable devices improves the reliability and stability of the system. After the switching chip SW is powered on, the first management component 21 obtains the computing device 11 identified by the switching chip SW to ensure that the acquired device information is accurate and correct. After the switching chip SW is fully powered on and running stably, device identification is performed to avoid identification errors or device damage caused by unstable power supply. Moreover, by performing device identification after power-on, it can be ensured that all components in the system are in a known and synchronized state. Device identification is not performed before the switching chip SW is fully powered on, which can avoid operations during chip initialization and reduce operational conflicts and potential errors. The power-on logic of this embodiment ensures the stability and reliability of the system during startup and configuration, reduces the possibility of errors, and improves the overall performance and user experience of the system.

[0083] In an exemplary embodiment, the second programmable device 22 is further configured to identify the presence information of each switch board and output a logic determination instruction according to the presence information; the first management component 21 is further configured to execute a corresponding management operation in response to the logic determination instruction.

[0084] In this embodiment, the second programmable device 22 is configured to perform logic judgment, specifically, the in-place information of each switch board can be judged. The second programmable device 22 identifies whether each switch board is in place (i.e., whether it has been correctly inserted and connected to the system) through a hardware detection mechanism (such as a switch, a sensor, or an EEPROM (Electrically Erasable Programmable Read-Only Memory) on the I2C bus). According to the in-place information of the switch board, the second programmable device 22 outputs a logic judgment instruction. These instructions may be a simple in-place / out-of-place state, or a more complex logical combination, indicating how the system responds to different configuration situations. The first management component 21 receives the logic judgment instruction from the second programmable device 22 and performs management operations, which include but are not limited to performing normal configuration and management operations if the switch board is in place; if the switch board is not in place, one of the following operations may be performed: ignoring the switch board, not configuring it, recording error information, and notifying the system administrator, trying to re-detect the in-place state of the switch board, and adjusting the system configuration to adapt to the absence of the switch board. This embodiment improves the flexibility and robustness of the multi-host server system.

[0085] In an exemplary embodiment, the server system further includes: a plurality of fan modules; a plurality of first temperature acquisition modules, the first temperature acquisition modules are configured to acquire first temperature information of the corresponding switch chip SW; a plurality of second temperature acquisition modules, the second temperature acquisition modules are configured to acquire second temperature information of the corresponding computing group 1; the second management component 23 is configured to control the speed of the fan module based on all the first temperature information and all the second temperature information, referring to Figure 4 shown.

[0086] In this embodiment, the second management component 23 may be specifically a BMC (Baseboard Management Controller), the second management component 23, the first management component 21 and the second programmable device 22 are all arranged on the same management board, the fan module may be arranged on the same fan board, each fan module may be independently controlled, each switching chip SW may correspond to a first temperature acquisition module, or a first temperature acquisition module may be arranged on a switching board, each computing group 1 may correspond to a second temperature acquisition module, or a first temperature acquisition module may be arranged on a computing motherboard, the first temperature acquisition module regularly collects temperature information of the switching chip SW, and the second temperature acquisition module regularly collects temperature information of the computing group 1. The second management component 23 collects temperature data from all first and second temperature acquisition modules, analyzes all temperature information, and adjusts the speed of the fan module according to a preset strategy. For example, if the temperature of a switching chip SW or computing group 1 exceeds a threshold, the second management component 23 increases the speed of the corresponding fan module to enhance the cooling effect.

[0087] This embodiment can dynamically adjust the cooling system according to the real-time heat load of the server, thereby improving energy efficiency and extending the life of the hardware.

[0088] In an exemplary embodiment, referring to Figure 4 The server system also includes an interface indicator board, which includes a plurality of status indicator lights; the second programmable device 22 is also configured to control the plurality of status indicator lights to perform indication operations according to their respective corresponding current status information.

[0089] In this embodiment, the IO (Input / Output) board (i.e., the interface indicator board in this embodiment) provides an external interface for the multi-host server system, including but not limited to 1 Power button, 1 UID (Unique Identifier) ​​button, 4 LED (Light Emitting Diode) lights for indicating the system status (respectively, system status indicator, BMC status indicator, power status indicator, fan status indicator), 2 network RJ45 (Registered Jack-45) interfaces, 1 debug RJ45 interface, and 1 USB (Universal Serial Bus) interface.

[0090] Among them, the Power button is used to start or shut down the server system, and the UID button is used to identify the server so as to quickly locate it among multiple servers in the rack. The LED lights include a system status indicator, a BMC status indicator, a power status indicator and a fan status indicator, which are used to indicate the status of different components. The network RJ45 interface is used to connect to the network, which may include a management network interface and a data network interface, and the Debug RJ45 interface is used for diagnostic and debugging purposes. The USB interface is used to connect external devices such as a keyboard, a mouse or a USB storage device. Specifically, the second programmable device 22 controls the on and off and flashing of the LED light according to the status information of each part of the system to indicate different system states. The second programmable device 22 obtains status information from various components of the server system (such as power module, fan module, BMC, etc.).

[0091] Figure 4 In the present invention, the server system includes a management board, two switch boards, an interface indicator board and a fan board. Each switch board has 16 external QSFP-DD interfaces, with a total of 32 QSFP-DD interfaces. The management board includes a first management component 21 (mCPU), a second management component 23 (BMC) and a second programmable device 22 (CPLD-B). The switch board is composed of a switch chip SW, a first programmable device 31 (CPLD-S), 2 MCIO (Mini Cool Edge IO, mini cooling edge input / output) interfaces and 16 QSFP-DD connectors. The switch chip SW is responsible for the interconnection and data transmission of the external computing device 11. The CPLD-S is dedicated to the configuration of the QSFP-DD optical module, and the MCIO module is reserved for data transmission between the two switch boards. Among them, the mCPU provides switch chip SW management; the BMC is used for out-of-band management, and the CPLD-B is used for power-on and power-off timing control of the entire machine, logical judgment, and light indication. Combined with Figure 3The management board transmits management signals to the switch board through the Manage Connector, including UART (Universal Asynchronous Receiver / Transmitter) signals, I2C signals, PERST signals, etc. The source of the UART signal is the BMC and mCPU on the management board, which are connected to the switch chip SW for chip debugging and information reading. The I2C signal includes 1 I2C signal from the mCPU and 2 I2C signals from the BMC. All three groups of I2C signals are connected to the CPLD for communication. In addition, one group of I2C signals from the BMC is connected to the switch chip SW for management and communication, and another group of I2C signals is used for ID identification. This group of I2C signals is connected to the switch chip SW, and 16 groups of signals are separated and connected to 16 QSFP-DD optical modules. In addition, the management board supplies power to the switch board through the power connector, provides 3.3V and 12V power to the switch board, and controls the power-on timing of other components through the CPLD-S. The server system is deployed in a server chassis.

[0092] In a second aspect, the present invention further provides a data center, comprising a plurality of server systems as described above.

[0093] The data center in this embodiment includes at least two chassis, each chassis has a management board, 2 switch boards, an IO board and a fan board. The entire architecture has a total of 4 switch boards, each of which has 16 external QSFP-DD interfaces, a total of 64 QSFP-DD interfaces, and a switch chip is selected. The GPUs belonging to different motherboards are connected to the same switch chip using optical modules. The interconnection topology of the system enables each host to read the information of all GPUs.

[0094] In a third aspect, the present invention also provides a server communication method, which is applied to a server system as any one of the above items, and the server communication method includes: identifying each computing device connected to itself through a switching chip in the server system; obtaining all computing devices identified by the switching chip through a first management component in the server system, assigning independent identifiers to all computing devices, and configuring routing tables of all optical ports on the switching chip based on the independent identifiers; after receiving data sent by the first computing device through any optical port of the switching chip in the server system, transmitting the data to the second computing device through the target optical port according to the routing table.

[0095] In an exemplary embodiment, the server system further includes a plurality of hosts, the hosts are connected to the corresponding at least two computing groups, and the server communication method further includes:

[0096] The host is configured to determine a first target computing device among all computing devices, determine a second target computing device among all computing devices connected to the host, and access the first target computing device using the second target computing device and a switching chip connected to the second target computing device.

[0097] In an exemplary embodiment, the server communication method further includes:

[0098] All computing devices identified by the switching chip are obtained through the first management component, address space is allocated to all computing devices, independent identifiers are allocated to all computing devices in the address space, and routing tables of all optical ports on the switching chip are configured based on the independent identifiers so that the switching chip forwards data according to the routing table.

[0099] In an exemplary embodiment, the server communication method further includes:

[0100] The first management component is used to control n switching chips to perform a reset operation.

[0101] In an exemplary embodiment, the server communication method further includes:

[0102] The transmission information of each optical port is identified through the first management component, and the interface specification and / or interface protocol of each optical port is adjusted based on the transmission information; the transmission information includes the device type of the computing device connected to the optical port and the data transmission rate of the optical port.

[0103] In an exemplary embodiment, the server system also includes a switching board, which is provided with at least one switching chip, and the switching board is also provided with multiple optical modules corresponding to the multiple optical ports on the switching chip, and the computing device is connected to the optical port through the optical module.

[0104] In an exemplary embodiment, the server system also includes a first programmable device, which is disposed on a switching board; the server communication method also includes: resetting the optical module disposed on the switching board where the first programmable device is located, so that the switching chip disposed on the switching board where the first programmable device is located can identify the computing device connected to the optical module.

[0105] In an exemplary embodiment, the server communication method further includes:

[0106] The first programmable device powers on the switching chip on the switching board where it is located, performs a reset operation on each optical module connected to the switching chip on the switching board where it is located, and / or, upon receiving a reset signal sent by the first management component, performs a reset operation on each optical module connected to the switching chip on the switching board where it is located.

[0107] In an exemplary embodiment, the server communication method further includes:

[0108] A mode control signal is output through a first programmable device; the mode control signal includes a first control signal corresponding to a host software control mode and a second control signal corresponding to a module hardware control mode; the optical module responds to the first control signal to adjust the current power mode to a low power mode, and responds to the second control signal to adjust the current power mode to a high power mode.

[0109] In an exemplary embodiment, the server communication method further includes:

[0110] When the optical module is in the low power mode, in response to receiving a host switching signal, the current power mode is adjusted to the high power mode.

[0111] In an exemplary embodiment, the server communication method further includes:

[0112] The first programmable device responds to the alarm signal output by the optical module and executes a corresponding protection operation.

[0113] In an exemplary embodiment, the server system further includes a second programmable device, and the server communication method further includes:

[0114] When receiving a power-on signal, the second programmable device controls the first management component to power on and outputs a power-on permission signal. After receiving a power-on completion signal, the second programmable device outputs a trigger signal to the first management component. After receiving a power-on permission signal, the first programmable device controls the switching board where the first management component is located to power on. When the power-on of the switching board where the first management component is located is completed, the second programmable device outputs a power-on completion signal. In response to receiving the trigger signal, the first management component obtains all computing devices identified by the switching chip, assigns independent identifiers to all computing devices, and configures the routing tables of all optical ports on the switching chip based on the independent identifiers so that the switching chip forwards data according to the routing tables.

[0115] In an exemplary embodiment, the server communication method further includes:

[0116] The second programmable device identifies the presence information of each switching board, and outputs a logic determination instruction according to the presence information; the first management component responds to the logic determination instruction and executes a corresponding management operation.

[0117] In an exemplary embodiment, the server system further includes an interface indicator board, which includes a plurality of status indicator lights; the server communication method further includes: controlling the plurality of status indicator lights to perform indication operations according to their respective corresponding current status information through a second programmable device.

[0118] In an exemplary embodiment, the server system also includes: a second management component, multiple fan modules, multiple first temperature acquisition modules, and multiple second temperature acquisition modules; the server communication method also includes: collecting the first temperature information of the corresponding switching chip through the first temperature acquisition module; collecting the second temperature information of the corresponding computing group through the second temperature acquisition module; and regulating the rotation speed of the fan module based on all the first temperature information and all the second temperature information through the second management component.

[0119] In a fourth aspect, the present invention further provides a server communication system, which is applied to a server system as any one of the above items, and the server communication system includes: an identification module, which is used to identify each computing device connected to itself through a switching chip in the server system; an establishment module, which is used to obtain all computing devices identified by the switching chip through a first management component in the server system, assign independent identifiers to all computing devices, and configure routing tables of all optical ports on the switching chip based on the independent identifiers; a routing module, which is used to transmit the data sent by the first computing device through any optical port of the switching chip in the server system to the second computing device through the target optical port according to the routing table.

[0120] In an exemplary embodiment, the server system also includes multiple hosts, which are connected to corresponding at least two computing groups. The server communication system is also used to: configure the host to determine a first target computing device among all computing devices, determine a second target computing device among all computing devices connected to itself, and access the first target computing device using the second target computing device and a switching chip connected to the second target computing device.

[0121] In an exemplary embodiment, the server communication system is also used to: obtain all computing devices identified by the switching chip through the first management component, allocate address space to all computing devices, allocate independent identifiers to all computing devices within the address space, and configure the routing tables of all optical ports on the switching chip based on the independent identifiers, so that the switching chip forwards data according to the routing table.

[0122] In an exemplary embodiment, the server communication system is further used to: perform a reset operation for controlling n switch chips through the first management component.

[0123] In an exemplary embodiment, the server communication system is also used to: identify the transmission information of each optical port through a first management component, and adjust the interface specifications and / or interface protocols of each optical port based on the transmission information; the transmission information includes the device type of the computing device connected to the optical port and the data transmission rate of the optical port.

[0124] In an exemplary embodiment, the server system also includes a switching board, which is provided with at least one switching chip, and the switching board is also provided with multiple optical modules corresponding to the multiple optical ports on the switching chip, and the computing device is connected to the optical port through the optical module.

[0125] In an exemplary embodiment, the server system also includes a first programmable device, which is disposed on a switching board; the server communication system is also used to: reset the optical module disposed on the switching board where the server communication system is located through the first programmable device, so that the switching chip disposed on the switching board where the server communication system is located can identify the computing device connected to the optical module.

[0126] In an exemplary embodiment, the server communication system is also used to: power on the switching chip on the switching board where the server communication system is located through the first programmable device, and perform a reset operation on each optical module connected to the switching chip on the switching board where the server communication system is located; and / or, upon receiving a reset signal sent by the first management component, perform a reset operation on each optical module connected to the switching chip on the switching board where the server communication system is located.

[0127] In an exemplary embodiment, the server communication system is also used to: output a mode control signal through a first programmable device; the mode control signal includes a first control signal corresponding to a host software control mode and a second control signal corresponding to a module hardware control mode; adjust the current power mode to a low power mode in response to the first control signal through the optical module, and adjust the current power mode to a high power mode in response to the second control signal.

[0128] In an exemplary embodiment, the server communication system is further configured to: when the optical module is in a low power mode, in response to receiving a host switching signal, adjust the current power mode to a high power mode.

[0129] In an exemplary embodiment, the server communication system is further configured to: respond to an alarm signal output by the optical module through the first programmable device and execute a corresponding protection operation.

[0130] In an exemplary embodiment, the server system also includes a second programmable device, and the server communication system is also used to: when receiving a power-on signal, control the first management component to power on and output a power-on permission signal through the second programmable device, and after receiving a power-on completion signal, output a trigger signal to the first management component; after receiving the power-on permission signal through the first programmable device, control the power-on of the switching board where the switch is located, and when the power-on of the switching board where the switch is located is completed, output a power-on completion signal; in response to receiving the trigger signal, obtain all computing devices identified by the switching chip through the first management component, assign independent identifiers to all computing devices, and configure the routing tables of all optical ports on the switching chip based on the independent identifiers, so that the switching chip forwards data according to the routing table.

[0131] In an exemplary embodiment, the server communication system is also used to: identify the presence information of each switching board through the second programmable device, and output a logic determination instruction according to the presence information; and execute a corresponding management operation in response to the logic determination instruction through the first management component.

[0132] In an exemplary embodiment, the server system further includes an interface indicator board, which includes a plurality of status indicator lights; the server communication system is further used to: control the plurality of status indicator lights through a second programmable device to perform indication operations according to their respective corresponding current status information.

[0133] In an exemplary embodiment, the server system also includes: a second management component, multiple fan modules, multiple first temperature acquisition modules, and multiple second temperature acquisition modules; the server communication system is also used to: collect the first temperature information of the corresponding switching chip through the first temperature acquisition module; collect the second temperature information of the corresponding computing group through the second temperature acquisition module; and regulate the rotation speed of the fan module based on all the first temperature information and all the second temperature information through the second management component.

[0134] In a fifth aspect, the present invention further provides a computer program product, comprising a computer program / instruction, which implements the above server communication method when executed by a processor.

[0135] For an introduction to a computer program product provided by the present invention, please refer to the above embodiments, and the present invention will not be described in detail here.

[0136] A computer program product provided by the present invention has the same beneficial effects as the above-mentioned server communication method.

[0137] In a sixth aspect, the present invention further provides a computer-readable storage medium having a computer program stored thereon, and the computer program implements the above-mentioned server communication method when executed by a processor.

[0138] The computer-readable storage medium may include: a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and other media that can store program codes.

[0139] For an introduction to a computer-readable storage medium provided by the present invention, please refer to the above-mentioned embodiments, and the present invention will not be described in detail here.

[0140] A computer-readable storage medium provided by the present invention has the same beneficial effects as the above-mentioned server communication method.

[0141] It should also be noted that, in this specification, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprises", "comprising" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the statement "comprising a ..." does not exclude the presence of other identical elements in the process, method, article or device including the element.

[0142] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A server system, characterized in that: include: A plurality of computing groups, each of the computing groups comprising a plurality of computing devices, any two computing devices in the computing groups being interconnected; n switching chips, each of which is configured to identify each of the computing devices connected to itself, and a plurality of optical ports on the i-th switching chip are correspondingly connected to an i-th computing device in a plurality of the computing groups, i=1, 2, ..., n; A first management component is connected to the first transmission interface on the switching chip, and the first management component is configured to obtain all the computing devices identified by the switching chip, assign independent identifiers to all the computing devices, and configure the routing table of all the optical ports on the switching chip based on the independent identifiers, so that the switching chip forwards data according to the routing table; the routing table includes a correspondence between each optical port and the independent identifier of the computing device connected thereto.

2. The server system according to claim 1, characterized in that: The server system also includes multiple hosts, which are connected to corresponding at least two computing groups. The hosts are configured to determine a first target computing device among all the computing devices, determine a second target computing device among all the computing devices connected to themselves, and access the first target computing device through the second target computing device and a switching chip connected to the second target computing device.

3. The server system according to claim 1, characterized in that: The first management component is configured to obtain all the computing devices identified by the switching chip, allocate address space to all the computing devices, allocate independent identifiers to all the computing devices within the address space, and configure the routing tables of all the optical ports on the switching chip based on the independent identifiers so that the switching chip forwards data according to the routing table.

4. The server system according to claim 1, characterized in that: The first management component is further configured to control the n switching chips to perform a reset operation.

5. The server system according to claim 1, characterized in that: The first management component is also configured to identify transmission information of each of the optical ports and adjust the interface specifications and / or interface protocols of each of the optical ports based on the transmission information; the transmission information includes the device type of the computing device connected to the optical port and the data transmission rate of the optical port.

6. The server system according to claim 1, characterized in that: The server system also includes a switching board, on which at least one switching chip is provided, and on which a plurality of optical modules corresponding to the plurality of optical ports on the switching chip are also provided, and the computing device is connected to the optical port via the optical module.

7. The server system according to claim 6, characterized in that: The server system also includes a first programmable device, which is disposed on the switch board. The first programmable device is configured to reset the optical module disposed on the switch board where the first programmable device is located, so that the switch chip disposed on the switch board where the first programmable device is located can identify the computing device connected to the optical module.

8. The server system according to claim 7, characterized in that: The first programmable device is configured to reset each of the optical modules connected to the switching chip on the switching board where the first programmable device is located when the switching chip on the switching board where the first programmable device is located is powered on, and / or, when receiving a reset signal sent by the first management component, reset each of the optical modules connected to the switching chip on the switching board where the first programmable device is located.

9. The server system according to claim 7, characterized in that: The first programmable device is further configured to output a mode control signal; the mode control signal includes a first control signal corresponding to a host software control mode and a second control signal corresponding to a module hardware control mode; The optical module is configured to adjust a current power mode to a low power mode in response to the first control signal, and to adjust a current power mode to a high power mode in response to the second control signal.

10. The server system according to claim 9, characterized in that: The optical module is further configured to adjust the current power mode to the high power mode in response to receiving a host switching signal when in the low power mode.

11. The server system according to claim 7, characterized in that: The first programmable device is also configured to respond to the alarm signal output by the optical module and perform a corresponding protection operation.

12. The server system according to claim 7, characterized in that: The server system also includes: A second programmable device is configured to control the first management component to power on when receiving a power-on signal, and output a power-on permission signal, and output a trigger signal to the first management component after receiving a power-on completion signal; The first programmable device is further configured to control the switching board where the first programmable device is located to power on after receiving the power-on permission signal, and output the power-on completion signal when the switching board where the first programmable device is located is powered on; The first management component is also configured to, in response to receiving the trigger signal, obtain all the computing devices identified by the switching chip, assign independent identifiers to all the computing devices, and configure the routing tables of all the optical ports on the switching chip based on the independent identifiers so that the switching chip forwards data according to the routing tables.

13. The server system according to claim 12, characterized in that: The second programmable device is further configured to identify the presence information of each of the switch boards and output a logic determination instruction according to the presence information; The first management component is further configured to execute a corresponding management operation in response to the logic determination instruction.

14. The server system according to claim 12, characterized in that: The server system further comprises an interface indicator board, wherein the interface indicator board comprises a plurality of status indicator lights; The second programmable device is also configured to control the plurality of status indicator lights to perform indication operations according to their respective corresponding current status information.

15. The server system according to any one of claims 1 to 14, characterized in that: The server system also includes: Multiple fan modules; A plurality of first temperature acquisition modules, wherein the first temperature acquisition modules are configured to acquire first temperature information of the corresponding switching chip; A plurality of second temperature acquisition modules, wherein the second temperature acquisition modules are configured to acquire second temperature information of the corresponding computing group; The second management component is configured to control the rotation speed of the fan module based on all the first temperature information and all the second temperature information.

16. A data center, characterized in that: It comprises a plurality of server systems as claimed in any one of claims 1 to 15.

17. A server communication method, characterized in that: Applied to the server system according to any one of claims 1 to 15, the server communication method comprises: Identify each computing device connected to the server system through a switch chip in the server system; Acquire all the computing devices identified by the switching chip through the first management component in the server system, assign independent identifiers to all the computing devices, and configure routing tables of all optical ports on the switching chip based on the independent identifiers; After receiving the data sent by the first computing device through any optical port of the switching chip in the server system, the data is transmitted to the second computing device through the target optical port according to the routing table.

18. A server communication system, characterized in that: Applied to the server system according to any one of claims 1 to 15, the server communication system comprises: An identification module, used to identify each computing device connected to itself through a switching chip in the server system; Establishing a module, used for acquiring all the computing devices identified by the switching chip through the first management component in the server system, assigning independent identifiers to all the computing devices, and configuring routing tables of all optical ports on the switching chip based on the independent identifiers; The routing module is used to transmit the data sent by the first computing device through a target optical port according to the routing table to the second computing device after receiving the data through any optical port of the switching chip in the server system.

19. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the server communication method of claim 17 is implemented.

20. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by the processor, the server communication method according to claim 17 is implemented.

Images