Server system and computing cluster

Abstract

The invention discloses a server system and a computing cluster, and relates to the field of servers, in order to solve the problem of low adaptation efficiency caused by the fact that a current server adopts a fixed topology, the server system comprises a carrier plate, at least one computing device and at least one core plate provided with a processor, and the core plate is detachably connected with the carrier plate. The connection board is respectively connected with the core board and the computing equipment and is used for loading and running the corresponding configuration firmware according to the current computing scene corresponding to the form of the computing equipment. Different numbers of processor configurations can be provided, so that different numbers of processing cores and interface resources are provided, the carrier plate, the core plate, the connecting plate and the like are designed separately, connection of different topologies is achieved, the connecting plate can flexibly switch running configuration firmware according to a current computing scene corresponding to the form of the computing device, and the computing device configuration efficiency is improved. Therefore, different forms of computing equipment can be adapted, and the adaptation efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the field of servers, and in particular to a server system and a computing cluster. Background Art

[0002] With the rapid development of the artificial intelligence market, processor and GPU (Graphics Processing Unit) technologies have also made rapid progress, bringing greater adjustments to the adaptation and debugging of AI servers. Currently, AI servers generally use one or two fixed high-speed connection topologies. This design is insufficient in flexibility and adjustability. When adapting to new processor platforms or GPUs, it is often restricted by topological links, resulting in a series of adaptation problems such as performance bottlenecks and poor interconnection.

[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 server system and a computing cluster that can provide different numbers of processor configurations, thereby providing different numbers of processing cores and interface resources. The discrete designs of the carrier board, core board and connection board can realize connections of different topologies. The connection board can flexibly switch the running configuration firmware according to the current computing scenario corresponding to the form of the computing device, thereby adapting to computing devices of different forms and improving adaptation efficiency.

[0005] To solve the above technical problems, the present invention provides a server system, including a carrier board, at least one computing device, and at least one core board provided with a processor, wherein the core board and the carrier board are detachably connected, and further comprising: a connecting board respectively connected to the core board and the computing device, for loading and running corresponding configuration firmware according to the current computing scenario corresponding to the form of the computing device.

[0006] Among them, a plurality of switching chips are arranged on the connection board, and the switching chips are used to load and run corresponding configuration firmware according to the current computing scenario corresponding to the form of the computing device.

[0007] Among them, the computing device includes a first computing device in a first form, and the connecting board is also provided with at least one slot connected to the switching chip, and the first computing device is connected to the switching chip on the connecting board through the slot; the switching chips on the connecting board are interconnected so that the first computing devices connected to the switching chips are interconnected.

[0008] Among them, the computing device includes a second computing device of a second form, the server system includes a universal substrate for mounting the second computing device, any two of the second computing devices on the universal substrate are interconnected, and a first connector is provided on the universal substrate; a second connector connected to the switching chip is provided on the connecting board, and the first connector is connected to all the second connectors.

[0009] Among them, the connecting board is also provided with at least one slot connected to the switching chip and at least one communication interface connected to the switching chip, and the server system also includes a network card device and a storage device, the network card device is connected to the connecting board through the slot, and the storage device is connected to the connecting board through the communication interface; the network card device is used to receive and send network data; the storage device is used to store the data to be calculated and the calculation result data of the computing device in the current computing task.

[0010] Among them, the connecting board is also provided with a plurality of slots connected to the switching chip, and the server system also includes a network card device and a storage device, and the network card device and the storage device are connected to the connecting board through their respective corresponding slots; the network card device is used to receive and send network data; the storage device is used to store the data to be calculated and the calculation result data of the computing device in the current computing task.

[0011] Wherein, the storage device includes a hard disk and / or a non-volatile storage medium.

[0012] The computing device is used to determine the storage requirements of the computing result data, and store the computing result data in the hard disk and / or the non-volatile storage medium corresponding to the storage requirements.

[0013] The carrier board includes a plurality of interconnected third connectors, and the core board is detachably connected to the third connectors.

[0014] Among them, the core board is also provided with at least one memory device, and the processor on the core board is used to access the local memory device and / or the remote memory device according to the requirements of the current computing task in the current computing scenario. The local memory device is a memory device provided on the same core board as the processor, and the remote memory device is a memory device provided on a different core board from the processor.

[0015] There are multiple core boards, and processors on any two core boards transmit task data in the current computing scenario to each other through the third connector.

[0016] Wherein, the carrier board includes a power supply device for supplying power to the core board connected to the carrier board.

[0017] The processor is further configured to allocate communication resources to at least one of the computing devices according to the current computing scenario.

[0018] Among them, the connecting board also includes a control component, a configuration firmware switching circuit and multiple flash memory modules, the control component includes an interactive part with adjustable position, wherein: the first end of the configuration firmware switching circuit is connected to the switching chip, the multiple second ends of the configuration firmware switching circuit are connected to the multiple flash memory modules in a one-to-one correspondence, the configuration end of the configuration firmware switching circuit is connected to the control component, and the multiple flash memory modules store configuration firmware corresponding to multiple forms in a one-to-one correspondence; the control component is used to output a configuration selection signal corresponding to the current computing scenario according to the current position of the interactive part; the configuration firmware switching circuit is used to respond to the configuration selection signal, determine the target flash memory module corresponding to the configuration selection signal, and control the first end of the configuration firmware switching circuit to be connected to the second end connected to the target flash memory module, so that the switching chip loads the configuration firmware in the target flash memory module corresponding to the configuration selection signal during the power-on startup process; the connecting board also has multiple reserved resistor positions arranged between the multiple flash memory modules and the switching chip, which are used to install configuration resistors after determining the current computing scenario.

[0019] In order to solve the above technical problems, the present invention further provides a computing cluster, comprising at least one server system as described in any one of the above.

[0020] The present invention provides a server system, which is provided with a core board and a carrier board of a processor that are detachably connected to provide different numbers of processor configurations, thereby providing different numbers of processing cores and interface resources. The carrier board, the core board and the connection board are discretely designed to achieve connections of different topologies. The connection board can flexibly switch the running configuration firmware according to the current computing scenario corresponding to the form of the computing device, thereby adapting to computing devices of different forms and improving the adaptation efficiency. The present invention also provides a computing cluster, which has the same beneficial effects as the above-mentioned server system. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] 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.

[0022] Figure 1 This is a schematic diagram of the structure of the first server system provided by the present invention.

[0023] Figure 2 This is a schematic structural diagram of a connecting plate provided by the present invention.

[0024] Figure 3 This is a schematic diagram of the structure of a second server system provided by the present invention.

[0025] Figure 4 This is a schematic diagram of the structure of the third server system provided by the present invention. DETAILED DESCRIPTION

[0026] The core of the present invention is to provide a server system and a computing cluster that can provide different numbers of processor configurations, thereby providing different numbers of processing cores and interface resources. The discrete designs of the carrier board, core board and connection board can realize connections of different topologies. The connection board can flexibly switch the running configuration firmware according to the current computing scenario corresponding to the form of the computing device, thereby adapting to computing devices of different forms and improving adaptation efficiency.

[0027] 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.

[0028] First, please refer to Figure 1 The present invention provides a server system, including a carrier board 1, at least one computing device 2, and at least one core board 3 provided with a processor 31, wherein the core board 3 and the carrier board 1 are detachably connected, and further comprising: a connecting board 4 respectively connected to the core board 3 and the computing device 2, for loading and running corresponding configuration firmware according to the current computing scenario corresponding to the form of the computing device 2.

[0029] Considering that in the server, the processor 31 is generally used as the RC (Root Complex), that is, the source of the PCIe (Peripheral Component Interconnect Express, a standard high-speed serial bus for interconnecting external devices) high-speed signal, the number of processors 31 determines the total amount of available PCIe resources. The more processors 31 there are, the higher the overall PCIe bandwidth that can be provided. In actual products, the number of processors 31 to be used needs to be determined according to the actual computing scenario. For example, a dual-way server uses two processors 31, and a four-way server uses four processors 31. In order to make the PCIe resources on the host side adjustable, this embodiment adopts a combination of the core board 3 and the carrier board 1 to modularize the PCIe resources corresponding to each processor 31, and determines the number of core boards 3 and the type of carrier board 1 to be used according to actual usage requirements. Specifically, the core board 3 can be detachably connected to the carrier board 1 so that the server system can be flexibly configured and upgraded according to different computing needs. The processor 31 provided on each core board 3 corresponds to a PCIe resource, so that the PCIe bandwidth of each processor 31 can be modularly configured according to actual needs, allowing the server system to adjust the number of processors 31 and PCIe bandwidth according to dual-way, quad-way or other multi-way configurations.

[0030] In addition, for different types of computing devices 2, such as GPU (Graphics Processing Unit), FPGA (Field-Programmable Gate Array), etc., the types of data transmission used in practical applications are different. For example, in some application scenarios, it is necessary to interconnect each PCIe switch (switch chip) to realize P2P (Peer-to-Peer) interconnection between each GPU, while in some application scenarios, it is not necessary to interconnect PCIe switches, and the PCIe resources of the switch chip can be used for other devices to achieve higher performance. Therefore, the connection board 4 in this embodiment realizes the switching of different types of interconnection by loading and running the configuration firmware corresponding to the current computing scene under the form of the computing device 2. Specifically, the connection board 4 plays a bridging role between the core board 3 and the computing device 2. It can load and run the corresponding configuration firmware according to the current computing scene corresponding to the type of the computing device 2, ensuring that the server system can adapt to different data processing requirements. At the same time, the server system can adapt to different types of computing devices 2, such as GPU, FPGA, etc. These devices are different in data transmission types, and the server system adapts to these differences through the configuration firmware of the connection board 4.

[0031] By adopting the solution of this embodiment, the number of core boards 3 can be increased or decreased according to computing needs, and the type of carrier board 1 can be adjusted to quickly adapt to different computing scenarios. The changing needs can be met by replacing or upgrading the core board 3 and the computing device 2. The efficiency and overall performance of data transmission are improved by optimizing the allocation of PCIe resources. Resources can also be configured according to actual needs to avoid cost waste caused by over-configuration.

[0032] It can be seen that in this embodiment, the core board 3 provided with the processor 31 and the carrier board 1 are detachably connected to provide different numbers of processor 31 configurations, thereby providing different numbers of processing cores and interface resources. The carrier board 1, the core board 3 and the connecting board 4 are discretely designed to realize connections of different topologies. The connecting board 4 can flexibly switch the running configuration firmware according to the current computing scenario corresponding to the form of the computing device 2, thereby adapting to computing devices of different forms and improving the adaptation efficiency.

[0033] Based on the above embodiments:

[0034] In an exemplary embodiment, referring to Figure 2 A plurality of switching chips 41 are provided on the connection board 4, and the switching chips 41 are used to load and run corresponding configuration firmware according to the current computing scenario corresponding to the form of the computing device 2.

[0035] In this embodiment, a plurality of switching chips 41 are provided on the connection board 4, and the switching chips 41 are used for chip management and allocation of PCIe channels, allowing efficient data communication between different computing devices 2 (such as GPU, FPGA, etc.) in the server. According to the needs of the actual application scenario, the working mode of the switching chip 41 can be adjusted by loading different configuration firmware, thereby adjusting the interconnection state between the various switching chips 41 on the connection board 4. For example, in a scenario where P2P interconnection between GPUs is required, the interconnection between the switching chips 41 can be enabled, so that the GPUs can communicate directly with each other without passing through the server's processor 31, thereby reducing the delay in data transmission and improving communication efficiency. In a scenario where interconnection between the switching chips 41 is not required, the PCIe resources originally used for interconnection between the switching chips 41 can be reallocated to other devices, such as additional storage devices or network adapters, thereby improving the performance and resource utilization of the overall system. This firmware switching mechanism enables the server to quickly adapt to different data processing requirements. The switching chip 41 in this embodiment includes but is not limited to PCIeswitch.

[0036] In an exemplary embodiment, the computing device 2 includes a first computing device in a first form, which is connected to a switching chip 41 on a connecting board 4 via a slot; the switching chips 41 on the connecting board 4 are interconnected to interconnect the computing devices 2 connected to the switching chips 41.

[0037] In this embodiment, taking the computing device 2 as a GPU as an example, GPUs usually have two forms: OAM (Open Accelerator Module) and AIC (Accelerator Interface Card). For GPUs of different forms, the link topology from CPU to GPU may be different, but all require PCIe switch to expand PCIe resources and implement functions such as P2P. The connection board 4 can be connected to the MCIO of the core board 3 through the MCIO (Mini Cool Edge O, full name multi-channel input / output)_U interface to realize the connection of the PCIe link from the RC end to the switch uplink port. The other ports of the connection board 4 can be connected to the UBB board (such as the GPU in the OAM form) through the BTB (Board-to-Board) connector, the AIC device can be installed through the PCIe slot, and the hard disk backplane can be connected through the MCIO_D interface or the interconnection between the switching chips 41 can be realized.

[0038] When a first computing device of a first form is selected, such as Figure 3 In the topology shown, different numbers of PCIe resources can be provided by selecting different numbers of core boards 3, and AIC-shaped GPUs can be installed on the PCIe slots of the switch board. The switching chips 41 can be connected to each other through MCIO_D interface cables to achieve interconnection between different switching chips 41, so as to realize functions such as P2P interconnection between GPUs.

[0039] In an exemplary embodiment, the computing device 2 includes a second computing device of a second form, and the server system includes a universal substrate for installing the second computing device, and any two second computing devices on the universal substrate are interconnected, and a first connector is provided on the universal substrate; a second connector connected to the switching chip 41 is provided on the connecting board, and the first connector is connected to all the second connectors.

[0040] When a second computing device of a second form is selected, such as Figure 4In the topology shown, the PCIe resources of the processor 31 are connected to the MCIO_U connector of the connecting board 4 by the MCIO connector on the core board 3 through a cable, and are connected to the downstream devices, including the network card, the hard disk, and the OAM on the UBB board, etc. through the expansion of the PCIe switch chip. The advantage of the core board 3 here is that when adapting to different models of OAM, different numbers of PCIe resources can be provided by selecting different numbers of core boards 3 to meet the requirements of performance or peripheral configuration. Some MCIOs on the core board 3 can be connected to the connecting board 4, and some can be connected to other peripheral devices, which can be determined according to the actual configuration scheme, and this embodiment is not limited here.

[0041] In an exemplary embodiment, referring to Figure 4 The connection board 4 is also provided with at least one slot connected to the switching chip 41 and at least one communication interface connected to the switching chip 41. The server system also includes a network card device and a storage device. The network card device is connected to the connection board 4 through the slot, and the storage device is connected to the connection board 4 through the communication interface. The network card device is used to receive and send network data; the storage device is used to store the data to be calculated and the calculation result data of the computing device 2 in the current calculation task. Among them, the slot is a PCIe slot.

[0042] In this embodiment, the connecting board 4 is provided with at least one slot, which is directly connected to the switching chip 41. This slot is used to connect a network card device (such as Figure 4 The connection board 4 is also provided with at least one communication interface, which is also connected to the switching chip 41 and is used to connect a storage device (such as a network card). Figure 4 The hard disk shown in the figure). Such a design allows the storage device to quickly exchange data with other components in the server system. The network card device is connected to the connection board 4 through a slot, and its main function is to receive and send network data. It ensures that the server can communicate with the external network, whether it is Internet data exchange or network communication in the internal data center. The storage device is connected to the connection board 4 through a communication interface, and is used to store the data to be calculated and the calculation result data of the computing device 2 in the current computing task. The storage device can be a hard disk drive, a solid-state drive, or other types of storage media, which provide the necessary data access speed and capacity to support the computing tasks of the server. By using slots and communication interfaces, the server system can easily add or replace network cards and storage devices, improving the maintainability and upgradeability of the system. The switching chip 41 is directly connected to the network card and storage device, reducing the delay in data transmission and improving the efficiency of network and storage operations. The system can select suitable network cards and storage solutions according to different application requirements to optimize performance and cost.

[0043] In an exemplary embodiment, referring to Figure 3 The connecting board 4 is also provided with at least one slot connected to the switching chip 41. The server system also includes a network card device and a storage device, which are connected to the connecting board 4 through the slot; the network card device is used to receive and send network data; the storage device is used to store the data to be calculated and the calculation result data of the computing device 2 in the current computing task.

[0044] In this embodiment, the slot on the connection board 4 is directly connected to the switching chip 41, providing an interface so that the network card device (such as Figure 3 NICs as shown) and storage devices such as Figure 3 The hard disk shown in the figure) can exchange data with other parts of the server system through this slot. By integrating the network card and storage device into the connection board 4, the complexity of wiring is reduced, and the neatness and reliability of the system are improved. The design of the slot allows the network card and storage device to be upgraded or replaced to adapt to changing technology and performance requirements. The network card and storage device directly connected to the switch chip 41 can use the high-speed PCIe channel, thereby providing higher data transmission rates and lower latency.

[0045] In an exemplary embodiment, the storage device includes a hard disk and / or a non-volatile storage medium.

[0046] In an exemplary embodiment, the computing device 2 is used to determine the storage requirements of the computing result data, and store the computing result data in a hard disk and / or a non-volatile storage medium corresponding to the storage requirements.

[0047] In this embodiment, the computing device 2 (such as a CPU or GPU) is responsible for determining the storage requirements of the calculation result data. This includes evaluating factors such as the size of the data, access frequency, storage life, and performance requirements. Based on these storage requirements, the computing device 2 stores the calculation result data on the most suitable storage medium. For example, for large, infrequently accessed data, you may choose to store it on a HDD (Hard Disk Drive) to take advantage of its high capacity and low cost. For small data that needs to be frequently accessed or read and written quickly, you may choose to store it on an SSD (Solid State Disk) or NVMe (Non-Volatile Memory Express) SSD to take advantage of its high-speed performance. Storing data on a storage medium that best suits its characteristics can optimize the utilization of storage resources and improve overall system performance. Selecting the appropriate storage medium based on the access mode and performance requirements of the data can reduce costs without reducing performance. The system can flexibly adjust storage policies according to different application requirements to adapt to changing workloads.

[0048] In an exemplary embodiment, the carrier board 1 includes a plurality of interconnected third connectors, and the core board 3 is detachably connected to the third connector. There are multiple core boards 3, and the processors 31 on any two core boards 3 transmit task data in the current computing scenario to each other through the third connector.

[0049] In this embodiment, multiple core boards 3 are interconnected through multiple interconnected third connectors, thereby allowing system expansion and adding additional functions.

[0050] In an exemplary embodiment, at least one memory device is also provided on the core board 3. The processor 31 on the core board 3 is used to access the local memory device and / or the remote memory device according to the requirements of the current computing task in the current computing scenario. The local memory device is a memory device provided on the same core board 3 as the processor 31, and the remote memory device is a memory device provided on a different core board 3 than the processor 31.

[0051] In this embodiment, the local memory device is a memory device located on the same core board 3 as the processor 31. Since it is located on the same board as the processor 31, the access latency is low, and it is usually used to store data and instructions currently being processed by the processor 31. The remote memory device is a memory device located on a different core board 3. The processor 31 can access the remote memory through the interconnection structure of the system, which is convenient for processing large amounts of data or computing tasks that require collaboration across multiple processors 31.

[0052] As an optional embodiment, the processor 31 can flexibly allocate memory resources according to task requirements to optimize performance and efficiency. By supporting remote memory devices, the system can expand its memory capacity without being affected by the memory limitations on a single core board 3. In a system including multiple core boards 3, if a memory device on a core board 3 fails, memory devices on other core boards 3 can take over the task, thereby improving the reliability of the system.

[0053] In an exemplary embodiment, the carrier board 1 includes a power supply device for supplying power to the core board 3 connected to the carrier board 1 .

[0054] In this embodiment, the power supply device on the carrier board 1 is responsible for converting the external power supply (such as AC or DC) into a voltage and current suitable for the core board 3 and other connected devices to work, and the power supply device also needs to effectively distribute the converted power to various components on the carrier board 1, including the core board 3 and other connected sub-boards. By integrating the power supply device on the carrier board 1, a stable and reliable power supply can be provided for the entire system, ensuring the normal operation of the core board 3 and other components.

[0055] In an exemplary embodiment, the processor 31 is further configured to allocate communication resources to at least one computing device 2 according to a current computing scenario.

[0056] In this embodiment, the processor 31 first needs to identify the current computing scenario, which may involve analyzing factors such as system load, user needs, application type, network conditions, etc. The processor 31 evaluates the needs of each computing device 2 for communication resources. For example, some applications may require higher bandwidth or lower latency, and the processor 31 formulates a resource allocation strategy based on the evaluation results. In this way, the processor 31 ensures that each computing device 2 in the system can obtain appropriate communication resources according to its needs and the current computing scenario, thereby improving the performance of the entire system and user experience.

[0057] In an exemplary embodiment, the connection board 4 also includes a control component, a configuration firmware switching circuit and multiple flash memory modules, the control component includes an interactive part with adjustable position, wherein: the first end of the configuration firmware switching circuit is connected to the switching chip 41, multiple second ends of the configuration firmware switching circuit are connected one-to-one with multiple flash memory modules, the configuration end of the configuration firmware switching circuit is connected to the control component, and multiple flash memory modules store configuration firmware corresponding to multiple forms one-to-one; the control component is used to output a configuration selection signal corresponding to the current computing scenario according to the current position of the interactive part; the configuration firmware switching circuit is used to respond to the configuration selection signal, determine the target flash memory module corresponding to the configuration selection signal, and control the first end of the configuration firmware switching circuit to be connected to the second end connected to the target flash memory module, so that the switching chip 41 loads the configuration firmware in the target flash memory module corresponding to the configuration selection signal during the power-on startup process; the connection board 4 also has multiple reserved resistor positions arranged between the multiple flash memory modules and the switching chip 41, which are used to install configuration resistors after determining the current computing scenario.

[0058] In this embodiment, the control component includes an adjustable interaction part, which can adjust the position according to external operation or system requirements. The function of the control component is to output a configuration selection signal according to the current position of the interaction part, which will instruct the configuration firmware switching circuit to select which flash memory module.

[0059] The first end of the configuration firmware switching circuit is connected to the switching chip 41, and the multiple second ends are respectively connected to the multiple flash memory modules, and the configuration end is connected to the control component for receiving the configuration selection signal. When the configuration selection signal is received, the configuration firmware switching circuit determines the target flash memory module, and the first end of the control circuit is connected to the second end of the target flash memory module, so that the switching chip 41 can load the configuration firmware in the target flash memory module during the power-on startup process. Each flash memory module stores a configuration firmware corresponding to a form, which is pre-programmed for different application scenarios. By switching the circuit, different flash memory modules can be selected, thereby changing the configuration and function of the switching chip 41. In addition, a plurality of reserved resistor positions are provided on the connecting board 4, which are located between the flash memory module and the switching chip 41. After determining the current application scenario, configuration resistors can be installed as needed. These resistors may be used to adjust circuit characteristics, optimize signal transmission, or provide additional protection functions. By simply adjusting the position of the interactive part, the configuration of the switching chip 41 can be quickly changed to adapt to different application scenarios. The reserved resistor position and the switchable flash memory module allow the system to be upgraded or modified in the future without replacing the entire connection board 4. By using the configuration firmware to switch the circuit, the risk of hardware failure can be reduced because the function can be changed without physically replacing the component. It can be understood that this connection board 4 design provides a highly flexible and configurable solution for server systems, especially in data centers and high-performance computing environments that need to quickly adapt to different computing scenarios.

[0060] In summary, the present invention proposes an AI server solution that is easy to adapt and debug. With a relatively low development cost, it can realize changes in different numbers of CPUs, changes in different PCIe link topologies, and flexible switching of different switch configurations, providing great convenience for the adaptation of each processor 31 platform and GPU, especially when conducting performance surveys and problem location, avoiding performance bottlenecks or problems that cannot be realized due to fixed board design.

[0061] In a second aspect, the present invention further provides a computing cluster, comprising at least one server system as described in any one of the above embodiments.

[0062] For an introduction to a computing cluster provided by the present invention, please refer to the above embodiments, and the present invention will not be described in detail here.

[0063] 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.

[0064] 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, It is characterized in that It includes a carrier board, at least one computing device, and at least one core board with a processor, wherein the core board and the carrier board are detachably connected, and further includes: The connection boards respectively connected to the core board and the computing device are used to load and run corresponding configuration firmware according to the current computing scenario corresponding to the form of the computing device.

2. The server system according to claim 1, It is characterized in that The connection board is provided with a plurality of switching chips, and the switching chips are used to load and run corresponding configuration firmware according to the current computing scenario corresponding to the form of the computing device.

3. The server system according to claim 2, It is characterized in that The computing device comprises a first computing device of a first form, the connecting board is further provided with at least one slot connected to the switching chip, and the first computing device is connected to the switching chip on the connecting board through the slot; The switching chips on the connection board are interconnected so that the first computing devices connected to the switching chips are interconnected.

4. The server system according to claim 2, It is characterized in that The computing device includes a second computing device of a second form, the server system includes a universal substrate for mounting the second computing device, any two second computing devices on the universal substrate are interconnected, and a first connector is provided on the universal substrate; The connection board is provided with a second connector connected to the switching chip, and the first connector is connected to all the second connectors.

5. The server system according to claim 2, It is characterized in that The connection board is also provided with at least one slot connected to the switching chip and at least one communication interface connected to the switching chip, the server system further includes a network card device and a storage device, the network card device is connected to the connection board through the slot, and the storage device is connected to the connection board through the communication interface; The network card device is used to receive and send network data; The storage device is used to store the data to be calculated and the calculation result data of the computing device in the current calculation task.

6. The server system according to claim 2, It is characterized in that The connection board is also provided with a plurality of slots connected to the switching chip, and the server system further comprises a network card device and a storage device, and the network card device and the storage device are connected to the connection board through the corresponding slots; The network card device is used to receive and send network data; The storage device is used to store the data to be calculated and the calculation result data of the computing device in the current calculation task.

7. The server system according to claim 6, It is characterized in that The storage device includes a hard disk and / or a non-volatile storage medium.

8. The server system according to claim 7, It is characterized in that The computing device is used to determine the storage requirements of the calculation result data, and store the calculation result data in the hard disk and / or the non-volatile storage medium corresponding to the storage requirements.

9. The server system according to claim 1, It is characterized in that The carrier board includes a plurality of interconnected third connectors, and the core board is detachably connected to the third connectors.

10. The server system according to claim 9, It is characterized in that The core board is also provided with at least one memory device. The processor on the core board is used to access local memory devices and / or remote memory devices according to the requirements of the current computing task in the current computing scenario. The local memory device is a memory device provided on the same core board as the processor, and the remote memory device is a memory device provided on a different core board from the processor.

11. The server system according to claim 9, It is characterized in that There are multiple core boards, and processors on any two core boards transmit task data in the current computing scenario to each other through the third connector.

12. The server system according to claim 1, It is characterized in that The carrier board includes a power supply device for supplying power to the core board connected to the carrier board.

13. The server system according to claim 1, It is characterized in that The processor is further configured to allocate communication resources to at least one of the computing devices according to the current computing scenario.

14. The server system according to any one of claims 1 to 13, It is characterized in that The connection board further includes a control component, a configuration firmware switching circuit and a plurality of flash memory modules, wherein the control component includes an interactive part with adjustable position, wherein: The first end of the configuration firmware switching circuit is connected to the switching chip on the connection board, the multiple second ends of the configuration firmware switching circuit are connected to the multiple flash memory modules in a one-to-one correspondence, the configuration end of the configuration firmware switching circuit is connected to the control component, and the multiple flash memory modules store the configuration firmware corresponding to the multiple forms in a one-to-one correspondence; The control component is used to output a configuration selection signal corresponding to the current computing scenario according to the current position of the interaction part; The configuration firmware switching circuit is used to respond to the configuration selection signal, determine the target flash memory module corresponding to the configuration selection signal, and control the first end of the configuration firmware switching circuit to be connected to the second end connected to the target flash memory module, so that the switching chip loads the configuration firmware in the target flash memory module corresponding to the configuration selection signal during the power-on startup process; The connecting board also has a plurality of reserved resistor positions disposed between the plurality of flash memory modules and the switching chip, for installing configuration resistors after determining the current computing scenario.

15. A computing cluster, It is characterized in that Comprising at least one server system according to any one of claims 1-14.

Images