Multi-node server architecture

Abstract

The present application relates to the technical field of servers, discloses a multi-node server architecture, and aims to solve the problems of how to reduce the amount of data cabling used and perform independent maintenance of nodes. The multi-node server architecture comprises a server chassis, a central processing unit node which is pluggably mounted in the server chassis, a data switch node which is pluggably mounted in the server chassis, a graphics processing unit node which is pluggably mounted in the server chassis, and a middle backplane module which is mounted in the server chassis. A first connector, a second connector, and a third connector which are in signal interconnection are arranged on the middle backplane module. The central processing unit node is insertedly engaged with the first connector, the data switch node is insertedly engaged with the second connector, and the graphics processing unit node is insertedly engaged with the third connector. According to the present application, the amount of data cabling used can be reduced as much as possible, independent disassembly, assembly, and maintenance operations of each node can be quickly and conveniently carried out, and the heat dissipation performance is ensured while flexible configuration of each node is achieved.

Description

A multi-node server architecture

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims priority to the Chinese patent application filed with the China Patent Office on September 26, 2023, with application number 202311247520.8 and application name “A Multi-node Server Architecture”, the entire contents of which are incorporated herein by reference. Technical Field

[0003] The present application relates to the field of server technology, and in particular to a multi-node server architecture. Background Art

[0004] Servers are an important component of electronic devices, primarily used to provide computing services. Depending on the type of service they provide, servers are categorized as file servers, storage servers, database servers, application servers, web servers, and more.

[0005] In the era of big data, a vast amount of IT (Information Technology) equipment is concentrated in data centers. Data center equipment primarily includes various types of servers, storage, switches, cabinets, and other infrastructure. Each piece of IT equipment is composed of various hardware boards, such as compute modules, storage modules, chassis, and cooling modules.

[0006] Thanks to the advancement of AI (Artificial Intelligence) technology, the demand for computing power is growing across various enterprises and research institutions, leading to the emergence of powerful AI servers. These servers primarily utilize GPU (Graphics Processing Unit) nodes to provide computing services, while also being equipped with CPU (Central Processing Unit) nodes for control and Switch nodes for inter-node data exchange.

[0007] In the prior art, there are two main GPU-based solutions for AI servers: PCIE (Peripheral Component Interconnect Express, a high-speed serial computer expansion bus standard) cards and onboard GPUs. Given the same internal chassis space, onboard GPUs offer higher integration and a relatively smaller footprint than PCIE cards, allowing for more GPU chips and a higher computing power density. Due to the larger overall size of onboard GPUs, AI servers using this approach often adopt a multi-node architecture, placing higher demands on server maintainability.

[0008] However, existing AI servers generally use data cables to interconnect nodes one at a time. This results in a large number of data cables being used, and the cables occupying a large amount of installation space within the server chassis. This makes it difficult to install and remove each node within the server chassis. During maintenance, the data cables at both ends of each node must be plugged and unplugged, which is a large workload and makes it difficult to install and remove each node individually. It also makes it difficult to flexibly configure each node, making it impossible to meet the diverse configuration requirements of different users for AI servers. Furthermore, the large number of data cables occupying ventilation space can also reduce the server's heat dissipation performance.

[0009] Therefore, how to minimize the use of data cables, conveniently and quickly perform separate disassembly and maintenance of each node, achieve flexible configuration of each node, and ensure heat dissipation performance is a technical problem faced by technical personnel in this field.

[0010] Summary of the Invention

[0011] The purpose of this application is to provide a multi-node server architecture that can minimize the use of data cables, facilitate and quickly perform individual disassembly and maintenance of each node, achieve flexible configuration of each node, and ensure heat dissipation performance.

[0012] To solve the above technical problems, the present application provides a multi-node server architecture, including a server chassis, a central processing unit node pluggably installed in the server chassis, a data exchange node pluggably installed in the server chassis, a graphics processing unit node pluggably installed in the server chassis, and a midplane module installed in the server chassis. The midplane module is provided with a first connector, a second connector and a third connector for signal interconnection. The central processing unit node is plugged into the first connector, the data exchange node is plugged into the second connector, and the graphics processing unit node is plugged into the third connector.

[0013] In some specific embodiments, the central processing unit node, the data exchange node, and the graphics processing unit node are vertically layered within the server chassis; and the first connector, the second connector, and the third connector are vertically layered on the midplane module.

[0014] In some specific embodiments, it also includes a power supply module and a heat dissipation module that are pluggable and installed in the server chassis. The mid-backplane module is also provided with a power connector and a heat dissipation control connector. The power supply module is plugged into the power connector, and the heat dissipation module is plugged into the heat dissipation control connector.

[0015] In some specific embodiments, the heat dissipation module includes a heat dissipation fan and a fan backplane, the heat dissipation fan is signal-connected to the fan backplane, and the fan backplane is plugged into and matched with a heat dissipation control connector.

[0016] In some specific embodiments, the server chassis includes a first installation cavity located on the front side of the mid-backplane module and a second installation cavity located on the rear side of the mid-backplane module. The central processing unit node, data exchange node and graphics processing unit node are all installed in the first installation cavity, and the power supply module and heat dissipation module are both installed in the second installation cavity.

[0017] In some specific embodiments, a first mounting rail assembly, a second mounting rail assembly, and a third mounting rail assembly are provided on the side wall of the server chassis in the first mounting cavity, the central processing unit node can be slidably provided on the first mounting rail assembly, the data exchange node can be slidably provided on the second mounting rail assembly, and the graphics processing unit node can be slidably provided on the third mounting rail assembly.

[0018] In some specific embodiments, part or all of the first mounting rail assembly, the second mounting rail assembly, and the third mounting rail assembly are guide rails extending along the depth dimension of the server chassis or a plurality of rollers distributed along the depth dimension of the server chassis.

[0019] In some specific embodiments, a mounting tray connected to both side walls of the server chassis is further provided in the first mounting cavity for supporting the data exchange node.

[0020] In some specific embodiments, multiple power supply compartments are arranged on the side walls of the server chassis in the second installation cavity, and multiple heat dissipation compartments are arranged between the side walls of the server chassis in the second installation cavity. The power supply module can be slidably inserted into the power supply compartment in a vertical posture, and the heat dissipation module can be slidably inserted into the heat dissipation compartment.

[0021] In some specific embodiments, an elastic baffle is connected to the inner wall of each heat dissipation chamber, which is used to elastically reset and close the corresponding heat dissipation chamber when the heat dissipation module is pulled out.

[0022] In some specific embodiments, the CPU node includes a first base and a first mainboard disposed in the first base. A first node connector is disposed at the rear end of the first mainboard, and the first node connector is used to dock with the first connector.

[0023] In some specific embodiments, the CPU node further includes a first front window disposed at the front end of the first base, and a first rotatable operating handle is disposed on a side of the first front window.

[0024] In some specific embodiments, the data exchange node includes a second base and a second mainboard disposed in the second base. A second node connector is disposed at the rear end of the second mainboard, and the second node connector is used to dock with the second connector.

[0025] In some specific embodiments, the data exchange node further includes a function expansion module disposed in the second base, and an expansion connector is disposed at the front end of the second mainboard, and the function expansion module is plugged into and mated with the expansion connector.

[0026] In some specific embodiments, a second slide rail extending along a depth dimension of the second base is provided on an outer side wall of the second base, and the second slide rail is used to cooperate with a second mounting rail assembly provided on a side wall of the server chassis.

[0027] In some specific embodiments, a second transition connecting plate is provided at the front end of the side wall of the second base, and a second operating handle that can be flipped is provided on the second transition connecting plate, as well as a second limiting column for abutting against the second operating handle at a preset flip position.

[0028] In some specific embodiments, the graphics processor node includes a third base, a third mainboard disposed in the third base, a third node connector disposed at a rear end of the third mainboard, and the third node connector is used to dock with the third connector.

[0029] In some specific embodiments, the graphics processor node further includes a third top cover inverted on the third base, with short folded edges erected on both sides of the third base, and side panels on both sides of the third top cover respectively forming a detachable connection with the corresponding short folded edges.

[0030] In some specific embodiments, a third slide rail extending along the depth dimension of the third base is provided on the outer side wall of the third base, and the third slide rail is used to cooperate with a third mounting rail assembly provided on the side wall of the server chassis.

[0031] In some specific embodiments, a third transition connecting plate is provided at the front end of the side wall of the third base, and a third operating handle that can be flipped is provided on the third transition connecting plate, as well as a third limiting column for abutting the third operating handle at a preset flipping position.

[0032] In some specific embodiments, a third front window is provided at the front end of the third base, and a locking member detachably connected to the third front window is provided on the third operating handle for locking the third operating handle on the third front window when the third operating handle is not flipped over.

[0033] In some specific embodiments, an elastic member connected to the rotating shaft of the third operating handle is further provided on the third transition connecting plate, which is used to pre-flip the third operating handle to a preset angle through elastic force when the locking member is unlocked.

[0034] In some specific embodiments, a pressable locking pin is protruded from the outer wall of the side panel of the third top cover, and a limiting block is protruded from the inner wall of the server chassis, and the limiting block is used to abut against the locking pin.

[0035] In some specific embodiments, the inner end of the locking pin is connected to an elastic strip, one end of the elastic strip is connected to the inner wall of the side panel of the third top cover, and a limiting plate is provided on the inner wall of the side panel of the third top cover at the other end of the elastic strip, and the limiting plate is used to form abutment with the other end of the elastic strip.

[0036] In some specific embodiments, the mid-backplane module includes a front backplane and a rear backplane connected as one body, the first connector, the second connector and the third connector are all arranged on the front backplane, and the power connector and the heat dissipation control connector are both arranged on the rear backplane.

[0037] The multi-node server architecture provided in this application mainly includes a server chassis, a central processing unit node, a data exchange node, a graphics processing unit node and a mid-backplane module. Among them, the server chassis is the main component of this server architecture, which is mainly used to install and carry the remaining server components. The central processing unit node is installed in the server chassis and forms a pluggable connection with the server chassis, that is, it can realize the insertion and installation operation and the pull-out and disassembly operation by moving back and forth in a straight line in the server chassis, and can be conveniently performed on the server chassis. The central processing unit node is mainly used to realize functions such as data interaction and global control. The graphics processing unit node is installed in the server chassis and forms a pluggable connection with the server chassis, that is, it can realize the insertion and installation operation and the pull-out and disassembly operation by moving back and forth in a straight line in the server chassis, and can be conveniently performed on the server chassis. The graphics processing unit node is mainly used for AI graphic data calculation. The data exchange node is installed within the server chassis and forms a pluggable connection with the server chassis. This allows for insertion and removal via reciprocating linear movement within the server chassis, making it easy to install and remove from the server chassis. The data exchange node primarily facilitates data exchange and communication between the CPU node and the GPU node. This means the CPU and GPU nodes do not communicate directly, but rather indirectly through the data exchange node. The midplane module is also installed within the server chassis and primarily facilitates unified data transfer between the CPU node and the data exchange node, and between the GPU node and the data exchange node. A first connector, a second connector and a third connector are provided on the mid-backplane module. The first connector, the second connector and the third connector maintain signal interconnection, and the first connector is mainly used to form a plug-in fit with the central processing unit node, thereby realizing the signal connection between the central processing unit node and the mid-backplane module; the second connector is mainly used to form a plug-in fit with the data exchange node, thereby realizing the signal connection between the data exchange node and the mid-backplane module; the third connector is mainly used to form a plug-in fit with the graphics processing unit node, thereby realizing the signal connection between the graphics processing unit node and the mid-backplane module.

[0038] The beneficial effect of the present application is that: since the central processing unit node, the data exchange node and the graphics processing unit node are all plug-in connected to the server chassis, and the central processing unit node, the data exchange node and the graphics processing unit node are respectively connected to the mid-backplane module by plugging and cooperating with the first connector, the second connector and the third connector, indirect data communication between the central processing unit node and the graphics processing unit node is realized by using the mid-backplane module and the data exchange node, so that there is basically no need to use any data cables for interconnection between the nodes. Therefore, on the one hand, the use of data cables can be minimized, thereby overcoming the problem of data cables occupying ventilation area and ensuring heat dissipation performance; on the other hand, the central processing unit node, the data exchange node and the graphics processing unit node can be easily disconnected from the first connector, the second connector and the third connector respectively, so that the central processing unit node, the data exchange node and the graphics processing unit node can be individually unplugged from the server chassis, and each node can be easily and quickly disassembled and maintained, thereby realizing flexible configuration of each node.

[0039] In addition, in some specific embodiments, after the third top cover in the graphics processor node is connected to the short folded edge, the entire graphics processor node forms an inverted structure, and the third motherboard equipped with the GPU chip can be directly installed into the third base from top to bottom vertically during the assembly process, and fixed by fasteners such as screws, without the need for other horizontal sliding movements or other installation operations. Finally, the third top cover is inverted on the short folded edge to complete the assembly; the side wall height of the third base is relatively low, which can shorten the installation stroke and simplify the installation steps of the third motherboard, thereby improving the assembly efficiency of the production line. At the same time, the third motherboard is easy to disassemble and has better maintainability, and also reduces the probability of the third motherboard being bumped during the assembly process.

[0040] In summary, the multi-node server architecture provided by this application can minimize the use of data cables, facilitate and quickly perform separate disassembly and maintenance of each node, achieve flexible configuration of each node, and ensure heat dissipation performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate some embodiments of the present application or technical solutions in the prior art, the following briefly introduces the drawings required for use in some embodiments or descriptions of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without any creative work.

[0042] FIG1 is an exploded view of the overall structure of some specific embodiments provided in this application.

[0043] FIG2 is a schematic diagram of the overall structure of some specific implementation methods provided in this application.

[0044] FIG3 is a schematic diagram of FIG2 from another perspective.

[0045] FIG4 is a schematic diagram of the front structure of the server chassis.

[0046] FIG5 is a schematic diagram of the structure after the installation tray is removed from FIG4 .

[0047] FIG6 is a schematic diagram of the rear end structure of the server chassis.

[0048] FIG7 is a schematic diagram of the specific structure of a central processing unit node.

[0049] FIG8 is an exploded view of the structure of FIG7 .

[0050] FIG9 is an enlarged view of the local structure of FIG7 .

[0051] FIG10 is a schematic diagram of the specific structure of a data exchange node.

[0052] FIG11 is an exploded view of the structure of FIG10 .

[0053] FIG12 is an enlarged view of the local structure of FIG10.

[0054] FIG13 is an exploded diagram of the specific structure of a graphics processor node.

[0055] FIG14 is a schematic diagram of the specific structure of a graphics processor node.

[0056] FIG15 is a schematic diagram of a partial structure of FIG14 .

[0057] FIG16 is a schematic diagram of the specific structure of the third base.

[0058] FIG17 is a schematic diagram of the flipped state of the third operating handle.

[0059] FIG18 is a schematic diagram showing a comparison of front and rear viewing angles of the third front window.

[0060] FIG19 is a schematic diagram of the specific structure of the third top cover.

[0061] FIG20 is a structural diagram of FIG19 from another perspective.

[0062] FIG21 is a schematic diagram of the installation process of a graphics processor node.

[0063] FIG22 is a schematic diagram showing the front and rear perspective comparison of the mid-back panel module.

[0064] In Figures 1-22: first installation cavity a, second installation cavity b; server chassis 1, CPU node 2, data exchange node 3, graphics processor node 4, midplane module 5, power module 6, heat dissipation module 7; first installation track assembly 11, second installation track assembly 12, third installation track assembly 13, installation tray 14, power supply compartment 15, heat dissipation compartment 16, elastic baffle 17, limit block 18, top cover 19, carrying handle 110; first base 21, first motherboard 22, first node connector 23, first front window 24, first operating handle 25, first bottom tray 26, first top cover 27; Second base 31, second mainboard 32, second node connector 33, function expansion module 34, expansion connector 35, second slide rail 36, second transition connecting plate 37, second operating handle 38, second limiting column 39, second top cover 310, second bottom tray 311, hard disk module 312; Third base 41, third mainboard 42, third node connector 43, third top cover 44, short folded edge 45, third slide rail 46, third transition connecting plate 47, third operating handle 48, third limiting column 49, third front window 410, locking member 411, elastic member 412, locking pin 413, elastic strip 414, limiting plate 415, third bottom tray 416, air guide cover 417, positioning column 418, positioning groove 419, avoidance hole 420, reinforcement plate 421, support block 422, reinforcement rib 423, positioning member 424, installation guide plate 425; First connector 51, second connector 52, third connector 53, power connector 54, heat dissipation control connector 55, front back plate 56, rear back plate 57, positioning rod 58, ventilation hole 59; Cooling fan—71, fan back plate—72. DETAILED DESCRIPTION

[0065] The following will be combined with the drawings in some embodiments of the present application to clearly and completely describe the technical solutions in some embodiments of the present application. Obviously, the embodiments described are only some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0066] It should be noted that the descriptions of "first", "second", etc. in this application are only used to distinguish technical features. The descriptions of "first", "second", etc. are only for descriptive purposes and cannot be understood as indicating or implying their relative importance, priority or sequence.

[0067] Please refer to Figures 1 and 2. Figure 1 is a schematic diagram of the overall structure of some specific implementation methods provided in this application, and Figure 2 is a schematic diagram of the overall structure of some specific implementation methods provided in this application.

[0068] In some specific implementations provided in this application, the multi-node server architecture mainly includes a server chassis 1, a central processing unit node 2, a data exchange node 3, a graphics processing unit node 4 and a mid-backplane module 5.

[0069] The server chassis 1 is the main component of the server architecture, and is mainly used to install and carry other server components.

[0070] The CPU node 2 is installed within the server chassis 1 and forms a pluggable connection with the server chassis 1. This allows for easy insertion and removal via reciprocating linear movement within the server chassis 1, making it easy to install and remove from the server chassis 1. The CPU node 2 is primarily used for data exchange and global control.

[0071] The graphics processing unit node 4 is installed in the server chassis 1 and forms a pluggable connection with the server chassis 1. That is, it can be inserted and removed by reciprocating linear movement within the server chassis 1, allowing for convenient installation and removal from the server chassis 1. The graphics processing unit node 4 is primarily used for AI graphics data calculations.

[0072] Data exchange node 3 is installed within server chassis 1 and forms a pluggable connection with the server chassis 1. This means it can be inserted and removed by reciprocating linear movement within the server chassis 1, making it easy to install and remove from the server chassis 1. Data exchange node 3 primarily facilitates data exchange and communication between CPU node 2 and GPU node 4. In other words, CPU node 2 and GPU node 4 do not communicate directly, but rather indirectly through data exchange node 3.

[0073] The midplane module 5 is also installed in the server chassis 1 and is primarily used to implement unified data transfer between the central processing unit node 2 and the data exchange node 3, and between the graphics processing unit node 4 and the data exchange node 3. A first connector 51, a second connector 52, and a third connector 53 are provided on the midplane module 5. These first connector 51, second connector 52, and third connector 53 maintain signal interconnection. The first connector 51 is primarily used to form a plug-in fit with the central processing unit node 2, thereby achieving a signal connection between the central processing unit node 2 and the midplane module 5; the second connector 52 is primarily used to form a plug-in fit with the data exchange node 3, thereby achieving a signal connection between the data exchange node 3 and the midplane module 5; and the third connector 53 is primarily used to form a plug-in fit with the graphics processing unit node 4, thereby achieving a signal connection between the graphics processing unit node 4 and the midplane module 5.

[0074] Thus, in the multi-node server architecture provided by this embodiment, since the CPU node 2, the data exchange node 3, and the graphics processor node 4 are all pluggable connected to the server chassis 1, and the CPU node 2, the data exchange node 3, and the graphics processor node 4 are simultaneously signal-connected to the midplane module 5 by plugging and cooperating with the first connector 51, the second connector 52, and the third connector 53, respectively, the midplane module 5 and the data exchange node 3 are used to implement indirect data communication between the CPU node 2 and the graphics processor node 4, so that there is basically no need to use any data cables to interconnect the nodes. Therefore, on the one hand, the use of data cables can be minimized, thereby overcoming the problem of data cables occupying ventilation area and ensuring heat dissipation performance; on the other hand, the CPU node 2, the data exchange node 3, and the graphics processor node 4 can be easily disconnected from the first connector 51, the second connector 52, and the third connector 53, respectively, so that the CPU node 2, the data exchange node 3, and the graphics processor node 4 can be individually unplugged from the server chassis 1, thereby enabling convenient and quick disassembly and maintenance of each node, thereby achieving flexible configuration of each node.

[0075] In summary, the multi-node server architecture provided by this embodiment can minimize the use of data cables, conveniently and quickly perform individual disassembly and maintenance operations on each node, achieve flexible configuration of each node, and ensure heat dissipation performance.

[0076] To facilitate the insertion and removal of the CPU node 2, data exchange node 3, and graphics processing unit node 4, in this embodiment, the CPU node 2, data exchange node 3, and graphics processing unit node 4 are arranged in a vertical layer within the server chassis 1, that is, they are arranged horizontally and along the height direction. For example, the CPU node 2 is located in the upper layer, the data exchange node 3 is located in the middle layer, and the graphics processing unit node 4 is located in the bottom layer. Correspondingly, the first connector 51, the second connector 52, and the third connector 53 are also arranged in a vertical layer on the midplane module 5, that is, they are distributed along the height direction of the midplane module 5. Moreover, the distribution of the first connector 51, the second connector 52, and the third connector 53 on the midplane module 5 corresponds to the distribution of the CPU node 2, the data exchange node 3, and the graphics processing unit node 4 within the server chassis 1, to ensure a correct one-to-one correspondence.

[0077] Of course, the CPU nodes 2 , data exchange nodes 3 and GPU nodes 4 are not limited to being distributed in a vertical layer within the server chassis 1 . For example, they may also be distributed vertically along the depth direction of the server chassis 1 .

[0078] As shown in FIG3 , FIG3 is a schematic diagram of FIG2 from another perspective.

[0079] In addition, in order to ensure the power supply and heat dissipation performance, a power supply module 6 and a heat dissipation module 7 are added in this embodiment. Among them, the power supply module 6 is installed in the server chassis 1 and forms a pluggable connection with the server chassis 1, that is, it can be inserted and installed and pulled out and disassembled by moving back and forth linearly in the server chassis 1, and can be easily disassembled and assembled on the server chassis 1. The power supply module 6 is mainly used to realize power supply to the remaining server components. The heat dissipation module 7 is installed in the server chassis 1 and forms a pluggable connection with the server chassis 1, that is, it can be inserted and installed and pulled out and disassembled by moving back and forth linearly in the server chassis 1, and can be easily disassembled and assembled on the server chassis 1. The heat dissipation module 7 is mainly used to realize air cooling and heat dissipation of the remaining server components. Of course, the heat dissipation module 7 can also dissipate the heat of the remaining server components by liquid cooling.

[0080] Accordingly, this embodiment adds a power connector 54 and a heat dissipation control connector 55 to the midplane module 5. The power connector 54 is primarily used to form a plug-in connection with the power module 6, thereby establishing an electrical connection between the power module 6 and the midplane module 5. The heat dissipation control connector 55 is primarily used to form a plug-in connection with the heat dissipation module 7, thereby establishing a signal connection between the heat dissipation module 7 and the midplane module 5. This configuration not only eliminates the need for power cables, but also further reduces the number of data cables used.

[0081] In a specific embodiment of the heat dissipation module 7, the module primarily comprises a cooling fan 71 and a fan backplane 72. The cooling fan 71 and the fan backplane 72 are signal-connected, and the fan backplane 72, via a connector, plugs into the heat dissipation control connector 55 on the midplane module 5. With this arrangement, the CPU node 2 can establish a signal connection with the cooling fan 71 via the heat dissipation control connector 55 on the midplane module 5, thereby conveniently controlling the cooling status of the cooling fan 71.

[0082] As shown in FIG4 and FIG5 , FIG4 is a schematic diagram of the front structure of the server chassis 1 , and FIG5 is a schematic diagram of the structure after the installation tray 14 is removed from FIG4 .

[0083] In some specific embodiments of the server chassis 1, the internal space of the server chassis 1 mainly includes a first installation cavity a and a second installation cavity b. The first installation cavity a and the second installation cavity b are separated by the mid-backplane module 5. The first installation cavity a is located in front of the mid-backplane module 5 and is also the front space of the server chassis 1, and is mainly used to install core components such as the central processing unit node 2, the data exchange node 3, and the graphics processing unit node 4; the second installation cavity b is located behind the mid-backplane module 5 and is also the rear space of the server chassis 1, and is mainly used to install other auxiliary server components such as the power supply module 6 and the heat dissipation module 7. In this arrangement, the mid-backplane module 5 is used to divide the internal space of the server chassis 1 into two parts, thereby separating the core components from the other auxiliary server components to avoid installation interference and heat dissipation effects.

[0084] Generally, the server chassis 1 is specifically 6U in height, while the central processing unit node 2 usually occupies 1U in height, the data exchange node 3 usually occupies 2U in height, and the graphics processing unit node 4 usually occupies 3U in height.

[0085] Furthermore, the first installation cavity a of the server chassis 1 specifically includes a first node installation cavity, a second node installation cavity and a third node installation cavity. Among them, the first node installation cavity is mainly used to install the central processing unit node 2, the second node installation cavity is mainly used to install the data exchange node 3, and the third node installation cavity is mainly used to install the graphics processing unit node 4. Generally, the first node installation cavity is located in the upper area of ​​the first installation cavity a, the second node installation cavity is located in the middle area of ​​the first installation cavity a, and the third node installation cavity is located in the lower area of ​​the first installation cavity a. The first node installation cavity, the second node installation cavity and the third node installation cavity form a hierarchical distribution in the vertical direction within the first installation cavity a and do not affect each other. Of course, the vertical distribution form of the first node installation cavity, the second node installation cavity and the third node installation cavity can also be flexibly adjusted, and the relative relationship between the upper and lower positions is not fixed. For example, the first node installation cavity is located in the middle area of ​​the first installation cavity a, the second node installation cavity is located in the upper area of ​​the first installation cavity a, and the third node installation cavity is located in the lower area of ​​the first installation cavity a, etc. Moreover, the first node installation cavity, the second node installation cavity and the third node installation cavity are not limited to forming a vertical hierarchical distribution in the first installation cavity a. For example, they can also be distributed in parallel along the depth direction of the server chassis 1. At this time, the central processing unit node 2, the data exchange node 3 and the graphics processing unit node 4 are respectively installed in the first installation cavity a in a vertical posture along the vertical direction.

[0086] In addition, the first installation cavity a of the server chassis 1 is not limited to including the first node installation cavity, the second node installation cavity and the third node installation cavity. When the number of nodes is greater, the first installation cavity a may also include more node installation cavities.

[0087] Moreover, in order to facilitate the installation of the mid-backplane module 5 in the server chassis 1, the internal space of the server chassis 1 includes, in addition to the first installation cavity a and the second installation cavity b, an intermediate installation cavity (not shown in the figure). Specifically, the intermediate installation cavity is located between the first installation cavity a and the second installation cavity b, separating the two from each other, and is mainly used to install the mid-backplane module 5. Generally, the depth dimension (or width) of the intermediate installation cavity is narrow so as to match the size of the mid-backplane module 5, and vertically extending sliding grooves are provided on the side walls on both sides of the intermediate installation cavity, so that the mid-backplane module 5 can slide vertically into the intermediate installation cavity, thereby realizing the sliding plug-in installation of the mid-backplane module 5 in the server chassis 1. Of course, the mid-backplane module 5 can also be reinforced with fasteners such as screws in the installation cavity, or positioning can be achieved in the installation cavity by cooperating with positioning holes on the mid-backplane module 5 through positioning pins and other components. To facilitate the installation and removal of the CPU node 2, data exchange node 3, and graphics processing unit node 4 within the server chassis 1, a first mounting rail assembly 11, a second mounting rail assembly 12, and a third mounting rail assembly 13 are provided in this embodiment. The CPU node 2 is specifically disposed on the first mounting rail assembly 11 and forms a sliding connection with the first mounting rail assembly 11, enabling it to slide along the first mounting rail assembly 11, thereby enabling the installation and removal of the CPU node 2 within the server chassis 1 along the first mounting rail assembly 11. The data exchange node 3 is specifically disposed on the second mounting rail assembly 12 and forms a sliding connection with the second mounting rail assembly 12, enabling it to slide along the second mounting rail assembly 12, thereby enabling the installation and removal of the data exchange node 3 within the server chassis 1 along the second mounting rail assembly 12. The graphics processing unit node 4 is specifically disposed on the third mounting rail assembly 13 and forms a sliding connection with the third mounting rail assembly 13, enabling it to slide along the third mounting rail assembly 13, thereby enabling the installation and removal of the graphics processing unit node 4 within the server chassis 1 along the third mounting rail assembly 13. With this arrangement, the central processing unit node 2, the data exchange node 3 and the graphics processing unit node 4 each correspond to a mounting track and can slide on their respective mounting tracks, ensuring that the plugging, unplugging and assembly operations of the central processing unit node 2, the data exchange node 3 and the graphics processing unit node 4 are independent of each other and do not affect each other, thereby realizing separate disassembly and maintenance operations and personalized configuration adjustments of the three.

[0088] In a specific embodiment of the first mounting rail assembly 11, the second mounting rail assembly 12, and the third mounting rail assembly 13, the first mounting rail assembly 11, the second mounting rail assembly 12, and the third mounting rail assembly 13 are specifically guide rails extending along the depth dimension of the server chassis 1, or are multiple rollers distributed along the depth dimension of the server chassis 1. For example, the first mounting rail assembly 11 is specifically the aforementioned guide rail, while the second mounting rail assembly 12 and the third mounting rail assembly 13 are specifically multiple rollers. With this arrangement, when the CPU node 2, the data exchange node 3, and the graphics processing unit node 4 are respectively mounted on the first mounting rail assembly 11, the second mounting rail assembly 12, and the third mounting rail assembly 13, they can maintain a horizontal position during insertion and removal, achieving stable insertion and removal operations. Furthermore, when multiple roller structures are used, rolling friction can be used to reduce friction and wear on the CPU node 2, the data exchange node 3, and the graphics processing unit node 4, thereby improving movement smoothness.

[0089] Considering that the data exchange node 3 is usually arranged in the middle layer position of the first installation cavity a in the server chassis 1, and the data exchange node 3 is heavy, if it is supported only by the second installation rail assembly 12, it may cause deformation of the second installation rail assembly 12. In view of this, a mounting tray 14 is added in this embodiment. Specifically, the mounting tray 14 is arranged in the first safety cavity of the server chassis 1, and is specifically connected to the side walls of the server chassis 1. It is mainly used to support the installation of the data exchange node 3 and assist the second installation rail assembly 12 to improve the installation stability of the data exchange node 3. Of course, the mounting tray 14 does not interfere with the movement of the data exchange node 3 on the second installation rail assembly 12.

[0090] As shown in FIG6 , FIG6 is a schematic diagram of the rear end structure of the server chassis 1 .

[0091] In order to facilitate the plugging, unplugging, and assembly / disassembly operations of the power module 6 and the heat dissipation module 7 in the server chassis 1 , a power supply compartment 15 and a heat dissipation compartment 16 are further provided in the server chassis 1 in this embodiment.

[0092] The power supply compartment 15 is disposed within the second mounting cavity b of the server chassis 1, specifically on the side wall of the server chassis 1 and vertically distributed close to the side wall of the server chassis 1. Typically, multiple power supply compartments 15 are provided, such as 3 to 6. Accordingly, the power supply module 6 must be installed vertically within the power supply compartment 15 and form a sliding connection with the bottom plate of the power supply compartment 15, allowing it to slide along the bottom plate of the power supply compartment 15 into or out of the power supply compartment 15. With this arrangement, each power supply module 6 rationally utilizes the side space of the server chassis 1, reduces the space occupied by the central installation space of the server chassis 1, and improves space utilization.

[0093] The heat sink 16 is also installed within the second mounting cavity b of the server chassis 1, specifically located in the middle area of ​​the server chassis 1. Typically, multiple heat sinks are provided, arranged in a matrix along the vertical and width directions according to a specific specification, such as a 3×4 arrangement. Accordingly, the heat sink 7 can be installed within the heat sink 16 and slidably connected to the bottom plate of the heat sink 16, allowing it to slide into and out of the heat sink 16 along the bottom plate. Furthermore, a fan backplane 72 is also installed within the second mounting cavity b of the server chassis 1, specifically located behind the heat sink 16, allowing it to conveniently connect to the cooling fans 71 installed within the heat sink 16. Typically, one fan backplane 72 connects to all cooling fans 71 within the same layer of heat sink 16, for example, one fan backplane 72 connects to three cooling fans 71. This arrangement allows the number of heat sinks 7 to be controlled by inserting and removing heat sinks 7 from the heat sink 16, conveniently adjusting the upper limit of the server's heat dissipation performance.

[0094] Furthermore, considering that after adjusting the configuration of each node, the heat dissipation module 7 in the heat dissipation bin 16 may also be under-equipped. In view of this, the present embodiment also adds an elastic baffle 17 in each heat dissipation bin 16. Specifically, one end of the elastic baffle 17 is connected to the inner wall of the heat dissipation bin 16, such as the top wall of the heat dissipation bin 16, and the other end extends toward the relative inner wall, such as to the bottom wall of the heat dissipation bin 16, and is elastic. When the heat dissipation module 7 is inserted into the heat dissipation bin 16, it will abut against the elastic baffle 17 and elastically deform it to be tightly attached to the inner wall of the heat dissipation bin 16, so that the elastic baffle 17 cannot block the ventilation area of ​​the heat dissipation bin 16; and when the heat dissipation module 7 is pulled out of the heat dissipation bin 16, the elastic baffle 17 rebounds under the action of the elastic force and closes the heat dissipation bin 16, which is equivalent to closing the corresponding heat dissipation bin 16 and reducing the ventilation area to zero. With such an arrangement, when the heat dissipation module 7 is pulled out from a heat dissipation bin 16, the cold air flow originally entering the heat dissipation bin 16 will be blocked by the elastic baffle 17, and thus flow into the remaining nearby heat dissipation bins 16, thereby increasing the cold air flow of the remaining heat dissipation bins 16 and avoiding waste of cooling capacity.

[0095] In addition, this embodiment provides a top cover 19 on top of the server chassis 1. Specifically, the top cover 19 mainly includes a front cover, a middle cover, and a thick cover. The front cover is located in each node area, the middle cover is located in the area of ​​the middle backplane module 5, and the rear cover is located in the area of ​​the power module 6. At the same time, the front and rear covers are fixed to the side walls of the server chassis 1 using fasteners such as rivets. The middle cover is positioned on the side walls of the server chassis 1 using I-shaped nails and then fixed to the side walls of the server chassis 1 using cover locks.

[0096] Furthermore, in order to facilitate the staff to carry the entire server, this embodiment also provides a carrying handle 110 on the side wall of the server chassis 1. Specifically, the carrying handle 110 can be flipped and stored on the side wall of the server chassis 1. When needed, it can be flipped outwards and grasped.

[0097] As shown in FIG7 and FIG8 , FIG7 is a schematic diagram of the specific structure of the central processing unit node 2 , and FIG8 is an exploded diagram of the structure of FIG7 .

[0098] In some specific embodiments of the central processing unit node 2, the central processing unit node 2 primarily includes a first base 21 and a first mainboard 22. The first base 21 is the underlying component of the central processing unit node 2, primarily used to support the remaining components. Server components such as the central processing unit, memory modules, and network cards are mounted on the first mainboard 22. Furthermore, to facilitate signal connection between the first mainboard 22 and the mid-backplane module 5, a first node connector 23 is provided at the rear end of the first mainboard 22. This first node connector 23 is primarily used to interface with the first connector 51 on the mid-backplane module 5, thereby establishing a signal connection between the first mainboard 22 and the mid-backplane module 5. With this arrangement, when disassembling the central processing unit node 2, the central processing unit node 2 can be disconnected from the mid-backplane module 5 by simply unplugging the first node connector 23 from the first connector 51.

[0099] As shown in FIG9 , FIG9 is an enlarged view of the local structure of FIG7 .

[0100] Furthermore, to facilitate insertion, removal, and removal of the CPU node 2 by personnel, in this embodiment, the CPU node 2 further includes a first front window 24 and a first operating handle 25. The first front window 24 is located at the front end of the first base 21, while the first operating handle 25 is located to the side of the first front window 24, for example, on both sides of the first front window 24, or only on one side of the first front window 24. Specifically, the first operating handle 25 is pivotally connected to the side of the first front window 24, enabling horizontal rotation relative to the first front window 24. In a normal state, the first operating handle 25 rotates to rest against the outer surface of the first front window 24, allowing for storage and securing with fasteners such as hand-tightened bolts. During insertion and removal, the first operating handle 25 rotates away from the outer surface of the first front window 24 and toward the front end of the first base 21, facilitating manual gripping and force application to insert or remove the entire CPU node 2 into or from the server chassis 1.

[0101] In addition, a first bottom tray 26 is provided within the first base 21 for supporting and installing the first motherboard 22. A plurality of heat dissipation holes may also be provided on the first front window 24 to enhance ventilation and heat dissipation. A first top cover 27 is also provided on the top of the first base 21. The first top cover 27 is positioned by inserting it into the top slot of the side wall of the first base 21 through components such as I-pins, and then fixed to the first base 21 through a push-type top cover lock. A connector positioning hole is also provided at the rear end of the first motherboard 22 to position it with the positioning column 418 provided on the mid-backplane module 5. A positioning notch is also provided at the rear end of the first base 21 to position it with components such as I-pins provided inside the server chassis 1, thereby achieving the installation and positioning of the entire central processing unit node 2 within the server chassis 1.

[0102] As shown in FIG10 and FIG11 , FIG10 is a schematic diagram of a specific structure of the data exchange node 3 , and FIG11 is an exploded diagram of the structure of FIG10 .

[0103] In some specific embodiments of the data exchange node 3, the data exchange node 3 mainly includes a second base 31 and a second mainboard 32. The second base 31 is the bottom component of the data exchange node 3 and is mainly used to carry the remaining components. Server components such as the data processing module are installed on the second mainboard 32, and in order to facilitate the signal connection between the second mainboard 32 and the mid-backplane module 5, a second node connector 33 is provided at the rear end of the second mainboard 32. The second node connector 33 is mainly used to dock with the second connector 52 on the mid-backplane module 5, thereby achieving a signal connection between the second mainboard 32 and the mid-backplane module 5. With such a configuration, when disassembling the data exchange node 3, the data exchange node 3 can be disconnected from the mid-backplane module 5 by simply unplugging the second node connector 33 from the second connector 52.

[0104] To improve the server's scalability, this embodiment adds a function expansion module 34 to the data exchange node 3. The function expansion module 34 is disposed within the second base 31, for example, at the front end of the second base 31. Specifically, it can be a PCIE module or another type of expansion module, such as one capable of mounting a PCIE card equipped with a GPU chip or a PCIE card equipped with a network chip. Generally, the function expansion module 34 is disposed at the front end of the second base 31, while the second mainboard 32 is disposed at the rear end of the second base 31. To facilitate signal connection between the function expansion module 34 and the second mainboard 32, this embodiment also includes an expansion connector 35 at the front end of the second mainboard 32. Specifically, the expansion connector 35 is primarily configured to plug into and mate with the connector on the function expansion module 34, thereby enabling signal connection between the function expansion module 34 and the second mainboard 32 and facilitating assembly and disassembly between the function expansion module 34 and the second mainboard 32, thereby achieving flexible configuration.

[0105] In order to facilitate the sliding and plugging movement of the data exchange node 3 on the second mounting rail assembly 12, a second slide rail 36 is added in this embodiment. Specifically, the second slide rail 36 is arranged on the outer wall of the second base 31, and can be arranged on only one side or on both sides at the same time. Specifically, it extends a preset length along the depth dimension direction (usually the length direction) of the second base 31, and is mainly used to cooperate with the second mounting rail assembly 12 set on the side wall of the server chassis 1. In this way, through the cooperation between the second slide rail 36 and the second mounting rail assembly 12, when the data exchange node 3 is plugged in and out, the second slide rail 36 can slide along the second mounting rail assembly 12, thereby guiding the plugging and unplugging movement of the data exchange node 3. For example, when the second mounting rail assembly 12 specifically adopts multiple rollers, the second slide rail 36 can simultaneously form rolling friction with each roller.

[0106] As shown in FIG12 , FIG12 is an enlarged view of the local structure of FIG10 .

[0107] To facilitate the insertion, removal, and assembly of the data exchange node 3, a second transition plate 37 and a second operating handle 38 are provided in this embodiment. The second transition plate 37 is positioned at the front end of the sidewall of the second base 31, for example, on both sides of the second base 31, or only on one side. The second operating handle 38 is disposed on the second transition plate 37 and pivotally connected thereto, enabling it to rotate relative to the second transition plate 37. In normal operation, the second operating handle 38 is tilted to a vertical position and housed on the outer surface of the second transition plate 37. It can also be secured with fasteners such as hand-tightened bolts. During insertion and removal, the second operating handle 38 tilts outward, away from the second transition plate 37, at a predetermined angle, and faces the front of the second base 31, making it easier for personnel to grasp and apply force to insert or remove the entire data exchange node 3 into or from the server chassis 1.

[0108] Furthermore, considering that the second operating handle 38 should not be tilted too far, a second limiting post 39 is added in this embodiment. Specifically, the second limiting post 39 is disposed on the second transition connecting plate 37, specifically in the rear region of the second operating handle 38. This second limiting post 39 is primarily used to abut the rear end of the second operating handle 38 when the second operating handle 38 is tilted outward to a predetermined angle, preventing the second operating handle 38 from tilting outward, thereby limiting the maximum tilt angle of the second operating handle 38. Generally, the maximum tilt angle of the second operating handle 38 is between 30° and 60°.

[0109] In addition, this embodiment also provides heat dissipation holes on the side walls of the second base 31, and an air guide cover is provided inside the second base 31 to enhance ventilation and heat dissipation performance. At the same time, a reinforcing beam is provided at the rear end of the second base 31 to improve structural strength. A second top cover 310 is provided on the top of the second base 31. The second top cover 310 is positioned and matched with the corresponding slots on the top of the side walls of the second base 31 by components such as I-pins, and then the second top cover 310 is fixed to the second base 31 by the top cover lock. A second bottom tray 311 is also provided inside the second base 31 to support and install the second motherboard 32. A hard disk module 312 can also be provided at the front end of the second base 31. The hard disk module 312 can generally accommodate 4 to 8 vertically inserted 2.5-inch hard disks. After the hard disks are plugged into the hard disk backplane, they are connected to the second motherboard 32 via data cables.

[0110] As shown in Figures 13, 14 and 15, Figure 13 is an exploded view of the specific structure of the graphics processor node 4, Figure 14 is a schematic diagram of the specific structure of the graphics processor node 4, and Figure 15 is a schematic diagram of the partial structure of Figure 14.

[0111] In some specific embodiments of the graphics processor node 4, the graphics processor node 4 primarily includes a third base 41 and a third mainboard 42. The third base 41 is the underlying component of the graphics processor node 4, primarily used to support the remaining components. Server components such as a GPU chip are mounted on the third mainboard 42. Furthermore, to facilitate signal connection between the third mainboard 42 and the mid-backplane module 5, a third node connector 43 is provided at the rear end of the third mainboard 42. This third node connector 43 is primarily used to interface with a third connector 53 on the mid-backplane module 5, thereby achieving signal connection between the third mainboard 42 and the mid-backplane module 5. With this arrangement, when disassembling the graphics processor node 4, the graphics processor node 4 can be disconnected from the mid-backplane module 5 by simply unplugging the third node connector 43 from the third connector 53.

[0112] As shown in FIG. 16 and FIG. 21 , FIG. 16 is a schematic diagram of the specific structure of the third base 41 , and FIG. 21 is a schematic diagram of the installation process of the graphics processor node 4 .

[0113] Considering the large size of the GPU motherboard with the GPU chip installed in the prior art, the GPU motherboard assembly process requires the use of specialized jigs to complete multiple translation steps. This makes the GPU board installation process cumbersome, limiting the production efficiency of the graphics processor node 4. Furthermore, the more assembly steps there are, the higher the probability of the GPU motherboard colliding with the chassis due to operator errors. To address this issue, and to address the issue of the GPU motherboard assembly process impacting production line assembly efficiency due to the large number of assembly steps, this embodiment adds a third top cover 44 and a short hem 45 to the graphics processor node 4. The third top cover 44, similar to the first top cover 2719 and second top cover 310, serves as the cover structure for the third base 41. However, the third top cover 44 has side panels that are significantly taller than the short hem 45, allowing it to be mounted upside down on the third base 41. The short hem 45 is positioned vertically on either side of the third base 41. Its low height, equivalent to the side hem of the third base 41, makes the third base 41 more similar to a flat plate or disk. At the same time, the side panels of the third top cover 44 are respectively connected to the short folded edges 45 on both sides to form a detachable connection. With this arrangement, after the third top cover 44 is connected to the short folded edges 45, the entire graphics processor node 4 forms an inverted structure, and during the assembly process, the third mainboard 42 equipped with the GPU chip can be directly installed vertically from top to bottom into the third base 41 and fixed with fasteners such as screws, without the need for other horizontal sliding movements or other installation operations. Finally, the third top cover 44 is inverted on the short folded edges 45 to complete the assembly; the side wall height of the third base 41 is relatively low, which can shorten the installation stroke and simplify the installation steps of the third mainboard 42, thereby improving the assembly efficiency of the production line. At the same time, the third mainboard 42 is easy to disassemble, has better maintainability, and also reduces the probability of the third mainboard being bumped during the assembly process.

[0114] Furthermore, to facilitate the detachable connection between the third top cover 44 and the short hem 45, in this embodiment, a positioning slot 419 is provided at the top of the short hem 45, and a component such as an I-nail is provided at the bottom of the third top cover 44 to cooperate with the positioning slot 419 for installation and positioning. Furthermore, a plurality of threaded holes are provided on the side walls of the short hem 45 to form a threaded connection with the side panels of the third top cover 44.

[0115] In order to facilitate the sliding and plugging movement of the graphics processor node 4 on the third mounting rail assembly 13, a third slide rail 46 is added in this embodiment. Specifically, the third slide rail 46 is set on the outer wall of the third base 41, and can be set on only one side or on both sides at the same time. It extends a preset length along the depth dimension direction (usually the length direction) of the third base 41, and is mainly used to cooperate with the third mounting rail assembly 13 set on the side wall of the server chassis 1. With such a setting, through the cooperation between the third slide rail 46 and the third mounting rail assembly 13, when the graphics processor node 4 is plugged in and out, the third slide rail 46 can slide along the third mounting rail assembly 13, thereby guiding the plugging and unplugging movement of the graphics processor node 4. For example, when the third mounting rail assembly 13 specifically adopts multiple rollers, the third slide rail 46 can simultaneously form rolling friction with each roller.

[0116] As shown in FIG. 17 , FIG. 17 is a schematic diagram of the flipped state of the third operating handle 48 .

[0117] To facilitate the insertion, removal, and assembly of the graphics processing unit node 4, this embodiment includes a third transition plate 47 and a third operating handle 48. The third transition plate 47 is positioned at the front end of the sidewall of the third base 41, for example, on both sides of the third base 41, or only on one side. The third transition plate 47 also facilitates the installation of the third front window 410 at the front end of the third base 41. The third operating handle 48 is mounted on the third transition plate 47 and pivotally connected to the third transition plate 47, enabling flipping relative to the third transition plate 47. In the normal state, the third operating handle 48 is flipped to a vertical position and is stored on the outer side of the third transition connecting plate 47. It can also be fixed by a locking member 411 such as a hand-tightened bolt; in the plugging and unplugging operation state, the third operating handle 48 is flipped outward to move away from the third transition connecting plate 47 and is in a preset tilt angle position, and the whole is facing the front of the third base 41, so that the staff can grasp and apply force with their hands, thereby inserting the entire graphics processor node 4 into the server chassis 1, or pulling the entire graphics processor node 4 out of the server chassis 1.

[0118] Furthermore, considering that the third operating handle 48 should not be tilted too far, a third limiting post 49 is added in this embodiment. Specifically, the third limiting post 49 is disposed on the third transition connecting plate 47, specifically in the rear region of the third operating handle 48. This post is primarily used to abut the rear end of the third operating handle 48 when the third operating handle 48 is tilted outward to a predetermined angle, preventing the third operating handle 48 from tilting outward, thereby limiting the maximum tilt angle of the third operating handle 48. Generally, the maximum tilt angle of the third operating handle 48 is between 30° and 60°.

[0119] Furthermore, this embodiment further includes an elastic member 412. Specifically, one end of the elastic member 412 is connected to the rotating shaft of the third operating handle 48, and the other end is connected to the outer wall of the third transition connecting plate 47. When the third operating handle 48 is flipped inward to the storage state, the elastic member 412 is compressed and accumulates elastic potential energy. At the same time, because the locking member 411 locks the third operating handle 48 in the storage state, the elastic member 412 can remain in the compressed state. When the locking member 411 is unlocked by the staff, the elastic member 412 immediately rebounds, releasing the elastic potential energy and driving the third operating handle 48 to flip outward. In other words, the elastic force drives the third operating handle 48 to automatically pre-flip to a preset angle, thereby facilitating the staff's grasping and eliminating the need to flip the third operating handle 48 outward from the storage state. This not only saves effort but also improves operating efficiency.

[0120] As shown in Figures 19 and 20, Figure 19 is a schematic diagram of the specific structure of the third top cover 44, and Figure 20 is a structural diagram of Figure 19 from another perspective.

[0121] In addition, considering the heavy weight of the graphics processor node 4, if the staff does not pay attention to the force control during the process of pulling out and disassembling the graphics processor node 4, the graphics processor node 4 may fall directly from the third base 41 due to the inertia of the graphics processor node 4, causing damage. To address this, a locking pin 413 and a limit block 18 are added in this embodiment. Specifically, the locking pin 413 is set on the outer wall of the side panel of the third top cover 44, and can be set on both side panels of the third top cover 44 at the same time, or can be set on only one side panel of the third top cover 44. At the same time, the locking pin 413 protrudes from the outer wall of the side panel of the third top cover 44 and can be pressed by the staff. Accordingly, the limit block 18 is specifically set on the inner side wall of the server chassis 1 and protrudes from the inner side wall of the server chassis 1, mainly used to form abutment with the locking pin 413. With this arrangement, when the staff pulls the graphics processor node 4 out of the server chassis 1, they cannot directly pull the graphics processor node 4 out of the third base 41, but are forced to pause halfway due to the abutment of the limit block 18 on the locking pin 413. Only after the staff presses and unlocks the locking pin 413 can they continue to pull the graphics processor node 4 out, thereby effectively preventing the graphics processor node 4 from accidentally falling.

[0122] Furthermore, to prevent the staff from pressing the lock pin 413 too deeply and causing it to fail, an elastic strip 414 and a limiting plate 415 are added in this embodiment. The elastic strip 414 is arranged at the inner end of the lock pin 413, and one end of the lock pin 413 is connected to the inner wall of the side panel of the third top cover 44, while the other end of the lock pin 413 extends along the length direction of the side panel of the third top cover 44. When the lock pin 413 is pressed, the elastic strip 414 elastically deforms and deflects inward. The limiting plate 415 is arranged on the inner wall of the side panel of the third top cover 44, specifically at the other end of the elastic strip 414, and is mainly used to form abutment with the other end of the elastic strip 414, limiting the maximum deflection of the elastic strip 414, thereby preventing the lock pin 413 from being pressed too deeply by the staff.

[0123] In addition, this embodiment also includes a third bottom tray 416 within the third base 41 to support and mount the third mainboard 42. Multiple gourd-shaped holes are defined in the third bottom tray 416, and multiple positioning posts 418 are located on the bottom surface of the third base 41. The gourd-shaped holes and positioning posts 418 cooperate to position the third bottom tray 416 on the third base 41. Positioning slots 419 are defined at the top of the short hem 45, and positioning members 424, such as I-pins, are located at the bottom of the third top cover 44. The positioning slots 419 and positioning members 424 cooperate to position the third top cover 44 on the third base 41. To improve heat dissipation, an air scoop 417 is added to the third base 41. Furthermore, the top of the short hem 45 includes overlapping mounting slots for mounting the air scoop 417, as well as retaining holes 420 for retaining the elastic strip 414 and the limiting plate 415. At the point where the front end of the third base 41 connects to the third transition connecting plate 47, an L-shaped reinforcement plate 421 is formed by raising a portion of the short folded edge 45 to enhance the connection strength. A positioning notch is provided at the rear end of the third base 41 to facilitate its positioning and installation within the server chassis 1. A slanted installation guide plate 425 is also provided at the rear end of the third base 41. This installation guide plate 425 is primarily used for installing the graphics processor node 4. The contact between the installation guide plate 425 and the server chassis 1 guides the installation movement of the third base 41, allowing for smoother installation within the server chassis 1.

[0124] As shown in FIG. 18 , FIG. 18 is a schematic diagram comparing front and rear viewing angles of the third front window 410 .

[0125] Furthermore, this embodiment also features multiple heat dissipation holes on the third front window 410 to enhance ventilation and heat dissipation. A support block 422 is also provided at the top of the third front window 410 to support the front end of the third top cover 44 and prevent it from sinking. To enhance the structural strength of the third front window 410, this embodiment also features multiple reinforcing ribs 423 on the inner sidewalls.

[0126] As shown in FIG22 , FIG22 is a schematic diagram comparing the front and rear viewing angles of the middle back panel module 5 .

[0127] In some specific embodiments of the mid-backplane module 5, the mid-backplane module 5 primarily includes a front backplane 56 and a rear backplane 57. The front backplane 56 and the rear backplane 57 are integrally connected, with the front backplane 56 facing the front of the server chassis 1, while the rear backplane 57 faces the rear of the server chassis 1. The first connector 51, the second connector 52, and the third connector 53 are all located on the front backplane 56, while the power connector 54 and the heat dissipation control connector 55 are both located on the rear backplane 57. This arrangement isolates the first connector 51, the second connector 52, and the third connector 53, as well as the power connector 54 and the heat dissipation control connector 55, from each other via the front backplane 56 and the rear backplane 57, respectively, thereby preventing heat concentration and connection confusion.

[0128] In addition, multiple ventilation holes 59 can be provided on both the front back plate 56 and the rear back plate 57 to enhance ventilation and heat dissipation performance. Furthermore, considering that the first node connector 23, the second node connector 33, and the third node connector 43 need to be docked with the first connector 51, the second connector 52, and the third connector 53 on the front back plate 56, respectively, and the power module 6 and the heat dissipation module 7 need to be docked with the power connector 54 and the heat dissipation control connector 55 on the rear back plate 57, respectively, to ensure accurate docking, this embodiment further provides multiple positioning rods 58 on the front back plate 56 and the rear back plate 57, so that the positioning rods 58 can form a positioning fit with components such as the positioning countersunk holes on the connectors at each node or on the power module 6 and the heat dissipation module 7.

[0129] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present application. Various modifications to these embodiments will be readily apparent to one 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 application. Therefore, the present application is not limited to the embodiments shown herein, but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A multi-node server architecture, characterized in that: The invention comprises a server chassis (1), a central processing unit node (2) pluggably installed in the server chassis (1), a data exchange node (3) pluggably installed in the server chassis (1), a graphics processing unit node (4) pluggably installed in the server chassis (1), and a mid-backplane module (5) installed in the server chassis (1); the mid-backplane module (5) is provided with a first connector (51), a second connector (52) and a third connector (53) for signal interconnection; the central processing unit node (2) is pluggably matched with the first connector (51), the data exchange node (3) is pluggably matched with the second connector (52), and the graphics processing unit node (4) is pluggably matched with the third connector (53).

2. The multi-node server architecture according to claim 1, characterized in that: The central processing unit node (2), the data exchange node (3) and the graphics processing unit node (4) are vertically layered in the server chassis (1); and the first connector (51), the second connector (52) and the third connector (53) are vertically layered on the mid-backplane module (5).

3. The multi-node server architecture according to claim 1, characterized in that: It also comprises a power module (6) and a heat dissipation module (7) which are pluggably installed in the server chassis (1); the mid-backplane module (5) is also provided with a power connector (54) and a heat dissipation control connector (55); the power module (6) is pluggably matched with the power connector (54); and the heat dissipation module (7) is pluggably matched with the heat dissipation control connector (55).

4. The multi-node server architecture according to claim 3, characterized in that: The heat dissipation module (7) comprises a heat dissipation fan (71) and a fan back plate (72); the heat dissipation fan (71) is signal-connected to the fan back plate (72), and the fan back plate (72) is plug-fitted to the heat dissipation control connector (55).

5. The multi-node server architecture according to claim 3, characterized in that: The server chassis (1) comprises a first installation cavity (a) located at the front side of the mid-backplane module (5) and a second installation cavity (b) located at the rear side of the mid-backplane module (5); the central processing unit node (2), the data exchange node (3) and the graphics processing unit node (4) are all installed in the first installation cavity (a); and the power supply module (6) and the heat dissipation module (7) are both installed in the second installation cavity (b).

6. The multi-node server architecture according to claim 5, characterized in that: A first mounting rail assembly (11), a second mounting rail assembly (12) and a third mounting rail assembly (13) are arranged on the side wall of the server chassis (1) in the first mounting cavity (a); the central processing unit node (2) can be slidably arranged on the first mounting rail assembly (11); the data exchange node (3) can be slidably arranged on the second mounting rail assembly (12); and the graphics processing unit node (4) can be slidably arranged on the third mounting rail assembly (13).

7. The multi-node server architecture according to claim 6, characterized in that: Part or all of the first mounting rail assembly (11), the second mounting rail assembly (12) and the third mounting rail assembly (13) are guide rails extending along the depth dimension direction of the server chassis (1) or include a plurality of rollers distributed along the depth dimension direction of the server chassis (1).

8. The multi-node server architecture according to claim 6, characterized in that: The first installation cavity (a) is also provided with an installation tray (14) connected to the side walls on both sides of the server chassis (1) for supporting the data exchange node (3).

9. The multi-node server architecture according to claim 5, characterized in that: A plurality of power supply compartments (15) are arranged on the side wall of the server chassis (1) in the second installation cavity (b), and a plurality of heat dissipation compartments (16) are arranged between the side walls of the server chassis (1) in the second installation cavity (b); the power supply module (6) is slidably inserted into the power supply compartment (15) in a vertical posture, and the heat dissipation module (7) is slidably inserted into the heat dissipation compartment (16).

10. The multi-node server architecture according to claim 9, characterized in that: An elastic baffle (17) is connected to the inner wall of each heat dissipation bin (16) and is used to elastically reset and close the corresponding heat dissipation bin (16) when the heat dissipation module (7) is pulled out.

11. The multi-node server architecture according to claim 1, characterized in that: The central processing unit node (2) comprises a first base (21) and a first mainboard (22) arranged in the first base (21); a first node connector (23) is arranged at the rear end of the first mainboard (22); and the first node connector (23) is used for docking with the first connector (51).

12. The multi-node server architecture according to claim 11, characterized in that: The central processing unit node (2) further comprises a first front window (24) arranged at the front end of the first base (21), and a first rotatable operating handle (25) is arranged on the side of the first front window (24).

13. The multi-node server architecture according to claim 1, characterized in that: The data exchange node (3) comprises a second base (31) and a second mainboard (32) arranged in the second base (31); a second node connector (33) is arranged at the rear end of the second mainboard (32); and the second node connector (33) is used for docking with the second connector (52).

14. The multi-node server architecture according to claim 13, characterized in that: The data exchange node (3) further comprises a function expansion module (34) arranged in the second base (31); an expansion connector (35) is arranged at the front end of the second mainboard (32); the function expansion module (34) is plugged into and matched with the expansion connector (35).

15. The multi-node server architecture according to claim 13, characterized in that: A second slide rail (36) extending in the depth direction of the second base (31) is provided on the outer side wall thereof, and the second slide rail (36) is used to cooperate with a second mounting rail assembly (12) provided on the side wall of the server chassis (1).

16. The multi-node server architecture according to claim 13, characterized in that: A second transition connecting plate (37) is provided at the front end of the side wall of the second base (31), and a second operating handle (38) capable of flipping movement and a second limiting column (39) for abutting against the second operating handle (38) at a preset flipping position are provided on the second transition connecting plate (37).

17. The multi-node server architecture according to any one of claims 1 to 16, characterized in that: The graphics processor node (4) comprises a third base (41), a third mainboard (42) arranged in the third base (41), a third node connector (43) being arranged at the rear end of the third mainboard (42), and the third node connector (43) being used for docking with the third connector (53).

18. The multi-node server architecture according to claim 17, characterized in that: The graphics processor node (4) further comprises a third top cover (44) which is inverted on the third base (41); short folded edges (45) are vertically arranged on both sides of the third base (41); and side panels on both sides of the third top cover (44) are respectively detachably connected to the corresponding short folded edges (45).

19. The multi-node server architecture according to claim 17, characterized in that: A third slide rail (46) extending in the depth direction of the third base (41) is arranged on the outer side wall thereof, and the third slide rail (46) is used to cooperate with a third mounting rail assembly (13) arranged on the side wall of the server chassis (1).

20. The multi-node server architecture according to claim 17, characterized in that: A third transition connecting plate (47) is provided at the front end of the side wall of the third base (41), and a third operating handle (48) capable of flipping movement is provided on the third transition connecting plate (47), as well as a third limiting column (49) for abutting against the third operating handle (48) at a preset flipping position.

21. The multi-node server architecture according to claim 20, characterized in that: A third front window (410) is arranged at the front end of the third base (41), and a locking piece (411) detachably connected to the third front window (410) is arranged on the third operating handle (48) for locking the third operating handle (48) on the third front window (410) when the third operating handle (48) is not turned over.

22. The multi-node server architecture according to claim 21, characterized in that: The third transition connecting plate (47) is also provided with an elastic member (412) connected to the rotating shaft of the third operating handle (48), which is used to pre-flip the third operating handle (48) to a preset angle by elastic force when the locking member (411) is unlocked.

23. The multi-node server architecture according to claim 18, characterized in that: A pressable locking pin (413) is protrudingly provided on the outer wall of the side plate of the third top cover (44), and a limiting block (18) is protrudingly provided on the inner wall of the server chassis (1), wherein the limiting block (18) is used to form abutment with the locking pin (413).

24. The multi-node server architecture according to claim 23, characterized in that: The inner end of the locking pin (413) is connected to an elastic strip (414), one end of the elastic strip (414) is connected to the inner wall of the side plate of the third top cover (44), and a limiting plate (415) is provided on the inner wall of the side plate of the third top cover (44) at the other end of the elastic strip (414), and the limiting plate (415) is used to form abutment with the other end of the elastic strip (414).

25. The multi-node server architecture according to claim 3, characterized in that: The middle backplane module (5) comprises a front backplane (56) and a rear backplane (57) which are connected as one body, the first connector (51), the second connector (52) and the third connector (53) are all arranged on the front backplane (56), and the power connector (54) and the heat dissipation control connector (55) are both arranged on the rear backplane (57).

Images