Racks and server systems

Abstract

This utility model provides a server rack and server system, relating to the field of artificial intelligence technology, particularly large models and chips. The specific implementation is as follows: The rack includes a frame, an orthogonal backplane, and a cable tray; the frame has a first mounting area and a second mounting area, and the orthogonal backplane and cable tray are both fixed to the frame; wherein, the first mounting area is used to install a first device, the first device including a first node, the first node being electrically connected to the orthogonal backplane for data transmission through the orthogonal backplane; the second mounting area is used to install a second device, the second device including a second node, the second node being electrically connected to the cable tray for data transmission through the cable tray.

Description

Technical Field

[0001] This utility model relates to the field of artificial intelligence technology, and in particular to the fields of large models and chips. It can be applied to data centers and servers. More specifically, this utility model provides a rack and server system. Background Technology

[0002] A supernode server can integrate multiple nodes to meet the needs of application scenarios that require a lot of computing power, such as large models. Utility Model Content

[0003] This utility model provides a server rack and server system.

[0004] According to one aspect of the present invention, a cabinet is provided, including a rack, a first device, and a second device. The rack has a first mounting area and a second mounting area. The first device is disposed in the first mounting area and includes a first node and an orthogonal backplane. The first node is electrically connected to the orthogonal backplane to transmit data through the orthogonal backplane. The second device is disposed in the second mounting area and includes a second node and a cable tray. The second node is electrically connected to the cable tray to transmit data through the cable tray.

[0005] According to one aspect of the present invention, a server system is provided, comprising: the aforementioned server rack, a first device, and a second device, wherein the first device is installed in a first installation area of ​​the server rack and includes a first node, the first node being electrically connected to an orthogonal backplane in the server rack to transmit data through the orthogonal backplane; and the second device is installed in a second installation area of ​​the server rack and includes a second node, the second node being electrically connected to a cable tray in the server rack to transmit data through the cable tray.

[0006] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this utility model, nor is it intended to limit the scope of this utility model. Other features of this utility model will become readily apparent from the following description. Attached Figure Description

[0007] The accompanying drawings are provided for a better understanding of this solution and do not constitute a limitation on this utility model. Wherein:

[0008] Figure 1A This is a schematic structural diagram of the back of a server system according to an embodiment of the present utility model;

[0009] Figure 1B This is a schematic structural diagram of the front of the server system according to an embodiment of the present utility model;

[0010] Figure 2 This is an exploded view of a server system according to another embodiment of the present invention; and

[0011] Figure 3 This is a schematic structural diagram of the first device and the second device according to an embodiment of the present utility model. Detailed Implementation

[0012] The following description, in conjunction with the accompanying drawings, illustrates exemplary embodiments of the present invention, including various details to aid understanding. These embodiments should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present invention. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0013] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein are to be interpreted in a manner consistent with the context of this specification, and not in an idealized or overly rigid way.

[0014] When using expressions such as "at least one of A, B, and C," the expression should generally be interpreted in accordance with the meaning commonly understood by a person skilled in the art (e.g., "a system having at least one of A, B, and C" should include, but is not limited to, systems having only A, only B, only C, A and B, A and C, B and C, and / or systems having A, B, and C, etc.). Two components / pipelines are connected, either directly or indirectly through other components / pipelines.

[0015] A supernode server can integrate multiple nodes, which may include multiple compute nodes and multiple switching nodes. The compute nodes and switching nodes need to be interconnected to communicate data.

[0016] In one technical solution, computing nodes and switching nodes can be interconnected through a cable tray, and the signals between the computing nodes and switching nodes are transmitted through copper cables in the cable tray to ensure transmission efficiency.

[0017] In another technical solution, compute nodes and switching nodes are interconnected via an orthogonal backplane. Both compute nodes and switching nodes are electrically connected to the interfaces of the orthogonal backplane via plug-in connections to achieve signal transmission. Orthogonality is a connection method that indicates that the surface of the node and the surface of the backplane are orthogonally arranged.

[0018] It should be noted that cable trays and orthogonal backplanes are two different architectures. Supernode servers interconnected using cable trays and those interconnected using orthogonal backplanes have different structures. Therefore, supernode servers of these two architectures cannot be mixed and matched in the same rack; for example, compute nodes in the two types of supernode servers cannot be used interchangeably.

[0019] However, in practical applications, some users use supernode servers configured with cable trays, some use supernode servers configured with orthogonal backplanes, and some use both types of supernode servers simultaneously. Therefore, for both supernode server vendors and users, a single rack can only deploy one type of supernode server, resulting in poor deployment and delivery flexibility.

[0020] This utility model aims to provide a server rack and a server system including the rack, which can be a rack for supernode servers. The rack includes a frame, an orthogonal backplane, and a cable tray. The frame has a first mounting area and a second mounting area. A first device is installed in the first mounting area, and the first device adopts an orthogonal backplane architecture. Simultaneously, a second device is installed in the second mounting area, and the second device adopts a cable tray architecture. Therefore, this rack can simultaneously accommodate devices with both orthogonal backplane and cable tray architectures, solving the problem of incompatibility between devices with different architectures deployed within the same rack.

[0021] The technical solutions provided in this disclosure will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0022] Figure 1A This is a schematic structural diagram of the back of the server system according to an embodiment of the present utility model. Figure 1B This is a schematic structural diagram of the front of the server system according to an embodiment of the present utility model.

[0023] like Figure 1A and Figure 1B As shown, in this embodiment, the cabinet includes: a rack 110, an orthogonal backplane 120 and a cable tray 130. The rack 110 is used to install the first device 210 and the second device 220.

[0024] The rack 110 has a first mounting area and a second mounting area. The rack 110 can adopt a frame structure or other structures. The rack 110 is mainly used to fix the components in the cabinet, such as the first device 210 and the second device 220, and other components can also be arranged according to actual needs.

[0025] The first device 210 is disposed in the first installation area. For example, the first device 210 can be placed in the first installation area or fixedly connected to the rack 110 by bolts or other components. The first device 210 includes a first node, and the number of first nodes can be one or more. A single first node can be a computing node or a switching node. The computing node can be a CPU (Central Processing Unit) or a GPU (Graphics Processing Unit), and the switching node can be a switch. The first node is electrically connected to the orthogonal backplane 120 to transmit data through the orthogonal backplane 120. For example, the interface of the first node and the socket of the orthogonal backplane 120 are electrically connected by a connector plug-in mating. In this embodiment, the orthogonal backplane 120 may have a protective housing.

[0026] The second device 220 is disposed in the second installation area. For example, the second device 220 can be placed in the second installation area or fixedly connected to the rack 110 by bolts or other components. The second device 220 includes a second node, and the number of second nodes can be one or more. A single second node can be a computing node or a switching node, wherein the computing node can be a CPU or a GPU, and the switching node can be a switch. The second node is electrically connected to the cable tray 130 to transmit data through the cable tray 130. For example, the interface of the second node and the socket of the cable tray 130 are electrically connected by a connector plug-in mating. In this embodiment, the cable tray 130 may have a protective housing.

[0027] According to the cabinet provided in this embodiment, during use, a first device 210 can be installed in the first installation area of ​​the rack 110, and the first device 210 is connected to the orthogonal backplane 120. A second device 220 can be installed in the second installation area of ​​the rack 110, and the second device 220 is connected to the cable tray 130. Therefore, this cabinet can simultaneously accommodate devices suitable for both the orthogonal backplane 120 and the cable tray 130 architectures, solving the problem of incompatible deployment of devices with different architectures within the same cabinet.

[0028] Figure 2 This is an exploded view of a cabinet according to another embodiment of the present invention.

[0029] like Figure 2 As shown, in this embodiment, the server system includes a server rack, a first device 210, and a second device 220. The server rack includes a rack 110, an orthogonal backplane 120, a cable tray 130, a conductive device 140, a first power supply device 150, and a second power supply device 160.

[0030] The rack 110 has a first mounting area 111, a second mounting area 112, a first power supply area 113, and a second power supply area 114. A first device 210 is mounted in the first mounting area 111, a second device 220 is mounted in the second mounting area 112, a first power supply device 150 is mounted in the first power supply area 113, and a second power supply device 160 is mounted in the second power supply area 114. The rack 110 can be divided into these areas by using plates or beams.

[0031] The conductive device 140 is fixedly connected to the frame 110 by bolts, clips, or other means. The conductive device 140 is a strip-shaped structure extending along a predetermined direction. The conductive device 140 is made of a suitable conductive material such as copper, for example, the conductive device 140 is a strip-shaped conductive copper busbar. The predetermined direction can be from top to bottom, from left to right, etc., and this embodiment does not limit the predetermined direction. The conductive device 140 includes a first conductive part, an insulating part, and a second conductive part arranged sequentially along the predetermined direction. The insulating part isolates the first conductive part and the second conductive part, thereby preventing mutual interference between the two power supply circuits.

[0032] The first node 211 in the first device 210 has a first power interface, and the first power supply device 150 is electrically connected to the first power interface of the first node 211 through a first conductive part, so that the first power supply device 150 supplies power to the first node 211 through the first conductive part. Similarly, the second node 221 in the second device 220 has a second power interface, and the second power supply device 160 is electrically connected to the second power interface of the second node 221 through a second conductive part, so that the second power supply device 160 supplies power to the second node 221 through the second conductive part.

[0033] In this embodiment, the first power supply device 150 and the second power supply device 160 supply power to the first device 210 and the second device 220, respectively, and the first power supply device 150 and the second power supply device 160 share a single conductive device 140. By integrating the first conductive part and the second conductive part into a strip structure and using an insulating part to achieve electrical isolation, the number of independent conductive components in the cabinet can be reduced, the wiring structure can be simplified, and the installation complexity of each component in the cabinet can be reduced.

[0034] It should be noted that in other embodiments, the first power interface of the first node 211 and the second power interface of the second node 221 can be electrically connected to the same power supply device through the conductive device 140, so that the first node 211 and the second node 221 can be powered by the same power supply device.

[0035] like Figure 2As shown, in this embodiment, the first power supply area 113, the first mounting area 111, the second mounting area 112, and the second power supply area 114 are arranged sequentially along a predetermined direction. After the rack is assembled, the first power supply device 150, the first device 210, the second device 220, and the second power supply device 160 are arranged sequentially along the predetermined direction. The predetermined direction can be from top to bottom or other directions; this embodiment does not limit the predetermined direction. The first power supply device 150 can supply power to the first device 210 nearby, and the second power supply device 160 can supply power to the second device 220 nearby, thereby shortening the power path and reducing power transmission loss. It is understood that in other examples, the areas in the rack 110 can be arranged in other ways, for example, the first power supply area 113, the first mounting area 111, the second power supply area 114, and the second mounting area 112 can be arranged sequentially along the predetermined direction.

[0036] like Figure 2 As shown, in this embodiment, the cabinet includes a rack 110 and a cooling system. The cooling system includes an inlet pipe 170 and an outlet pipe 180. The first device 210 and the second device 220 employ liquid cooling. The first node 211 in the first device 210 has a first inlet 2111 and a first outlet 2112. The first node 211 may have a cavity inside or outside for supplying coolant flow. The second node 221 in the second device 220 has a second inlet 2211 and a second outlet 2212.

[0037] The liquid inlet pipe 170 has a first pipe outlet and a second pipe outlet. The first pipe outlet is sealed to the first liquid inlet 2111, and the second pipe outlet is sealed to the second liquid inlet 2211. Similarly, the liquid outlet pipe 180 has a first pipe inlet and a second pipe inlet. The first pipe inlet is sealed to the first liquid outlet 2112, and the second pipe inlet is sealed to the second liquid outlet 2212. A liquid cooling connector can be used to achieve a sealed connection between the pipes. Furthermore, the liquid inlet pipe 170 and the liquid outlet pipe 180 are connected to the secondary liquid supply pipes in the computer room to dissipate heat from each first node 211 and each second node 221.

[0038] During the heat exchange process, the low-temperature coolant from the secondary side of the computer room enters the inlet pipe 170. A portion of the coolant in the inlet pipe 170 enters the first inlet 2111 through the first pipe outlet, then flows through the interior of the first node 211 to cool it down. The high-temperature coolant then enters the outlet pipe 180 through the first outlet 2112 and the first pipe inlet. Another portion of the coolant in the inlet pipe 170 enters the second inlet 2211 through the second pipe outlet, then flows through the second node 221 to cool it down. The high-temperature coolant then enters the outlet pipe 180 through the second outlet 2212 and the second pipe inlet. The two portions of coolant merge and return to the secondary side of the computer room for further heat exchange and cooling to become low-temperature coolant.

[0039] In this embodiment, the first device 210 and the second device 220 share a single cooling system. Since there are no separate inlet pipes 170 and outlet pipes 180 for the first device 210 and the second device 220, the number of pipes and connecting components can be reduced, lowering the overall cost of the cabinet and simplifying installation. Furthermore, the coolant in the first device 210 and the second device 220 is distributed to both low-temperature coolant and recovered to high-temperature coolant through the same cooling system, ensuring uniform heat dissipation and avoiding the uneven distribution of cooling capacity that could result from multiple independent cooling systems.

[0040] like Figure 2 As shown, according to another embodiment of the present invention, the first device 210 includes a plurality of first nodes 211, and the second device 220 includes a plurality of second nodes 221. The plurality of first nodes 211 and the plurality of second nodes 221 are arranged sequentially along a predetermined direction, which may be a top-to-bottom direction or other directions.

[0041] The inlet pipe 170 is a pipe extending in a predetermined direction. It can be seen that when multiple first nodes 211 and multiple second nodes 221 are arranged sequentially from top to bottom, the inlet pipe 170 also extends in a top-to-bottom direction. The inlet pipe 170 has multiple first pipe outlets and multiple second pipe outlets. Each of the multiple first pipe outlets corresponds one-to-one with and is sealed to the first inlet 2111 of each of the multiple first nodes 211. Similarly, each of the multiple second pipe outlets corresponds one-to-one with and is sealed to the second inlet 2211 of each of the multiple second nodes 221.

[0042] The liquid outlet pipe 180 is a pipe extending in a predetermined direction, for example, the liquid outlet pipe 180 extends in a downward direction. The liquid outlet pipe 180 has multiple first pipe inlets and multiple second pipe inlets. The multiple first pipe inlets correspond one-to-one with and are sealed to the first liquid outlet 2112 of each of the multiple second nodes 221. The multiple second pipe inlets correspond one-to-one with and are sealed to the second liquid outlet 2212 of each of the multiple second nodes 221.

[0043] This embodiment arranges multiple first nodes 211 and second nodes 221 sequentially along a predetermined direction, and uses inlet pipes 170 and outlet pipes 180 extending in the same direction for unified liquid supply and return. This simplifies and tidies up the cooling pipe layout, reduces complex bends and branches, thereby lowering flow resistance and improving coolant delivery efficiency. Furthermore, this pipe arrangement improves rack space utilization and is suitable for use in high-density servers.

[0044] like Figure 2 As shown, according to another embodiment of the present invention, the conductive device 140 is a strip-shaped structure extending in a predetermined direction. The inlet pipe 170 and the outlet pipe 180 are pipes extending in the predetermined direction, for example, the conductive device 140, the inlet pipe 170, and the outlet pipe 180 all extend from top to bottom. The conductive device 140 is located between the inlet pipe 170 and the outlet pipe 180. In this embodiment, the conductive device 140, the inlet pipe 170, and the outlet pipe 180 are arranged parallel to each other in the same direction, and the conductive device 140 is placed between the inlet pipe 170 and the outlet pipe 180. In this way, the interfaces for connecting to the pipes and the interfaces for connecting to the conductive device 140 in the first device 210 and the second device 220 are all located on the same side of the device, thus making assembly more convenient.

[0045] like Figure 2 As shown, according to another embodiment of the present invention, the frame 110 has a first side and a second side opposite to the first side. The first side is, for example, the front side of the frame 110, and the second side is, for example, the rear side of the frame 110. A first mounting area 111 and a second mounting area 112 both extend from the first side to the second side. Alternatively, the first mounting area 111 and the second mounting area 112 may also extend from the first side to the second side. An orthogonal backplate 120, a cable tray 130, a conductive device 140, an inlet pipe 170, and an outlet pipe 180 are all disposed on the second side.

[0046] Taking the first side and the second side as the front and rear sides respectively as examples, the assembly process is explained as follows: The first power supply device 150 and the second power supply device 160 can be installed from the front side into the two power supply areas 113 and 114 in the frame 110, and the first power supply device 150 and the second power supply device 160 are respectively connected to the first conductive part and the second conductive part in the conductive device 140 on the rear side of the frame 110 to achieve a power supply effect. Furthermore, the first device 210 and the second device 220 can also be installed from the front side into the first mounting area 111 and the second mounting area 112 of the frame 110, and the first device 210 and the second device 220 are respectively connected to the first conductive part and the second conductive part in the conductive device 140 to achieve a power receiving effect. Simultaneously, the first device 210 and the second device 220 are connected to the cooling pipes to achieve a pipe connection effect. The cooling pipes include the aforementioned liquid inlet pipe 170 and liquid outlet pipe 180.

[0047] In this embodiment, the conductive device 140 and liquid cooling components are centrally located at the rear of the rack 110, while the installation areas for the first device 210, the second device 220, the first power supply device 150, and the second power supply device 160 are located at the front of the rack 110. This achieves front-to-back functional partitioning, improving the maintainability and assembly efficiency of the rack. Furthermore, this embodiment employs a front-side device plug-and-play design, allowing maintenance personnel to complete the assembly of components without having to go around to the rear of the rack 110, reducing operational complexity. The centralized rear-side layout of the conductive device 140 and cooling pipes also optimizes cable and pipe routing, reducing space occupancy.

[0048] Figure 3 This is a schematic structural diagram of the first device and the second device according to an embodiment of the present utility model.

[0049] like Figure 3 As shown, in this embodiment, the first device 210 includes multiple first nodes 211, which include a first number of computing nodes and a second number of switching nodes. For example, the multiple first nodes 211 may include 8 computing nodes and 4 switching nodes. In actual deployment, 4 computing nodes can be arranged above and below the 4 switching nodes respectively. A computing node may include multiple processors, which may be a central processing unit or a graphics processing unit. For example, a computing node may include 4 or 8 graphics processors. In practical applications, each first node 211 in the first device 210 can be electrically connected to the orthogonal backplane 120 through a first connector 212, thereby realizing the interconnection between nodes. For example, the computing nodes in the first device 210 can transmit data with other computing nodes or switching nodes through the orthogonal backplane 120. In this embodiment, the first device 210 integrates multiple computing nodes and multiple switching nodes, which can meet the needs of scenarios requiring a large amount of computing power, such as large models.

[0050] like Figure 3 As shown, in this embodiment, the second device 220 includes multiple second nodes 221, which include a third number of computing nodes and a fourth number of switching nodes. For example, the multiple second nodes 221 may include 8 computing nodes and 4 switching nodes. In actual deployment, 4 computing nodes can be arranged above and below the 4 switching nodes respectively. A computing node may include multiple processors, which may be a central processing unit or a graphics processing unit. For example, a computing node may include 4 or 8 graphics processors. In practical applications, each second node 221 in the second device 220 can be electrically connected to the cable tray 130 through a second connector 222, thereby realizing the interconnection between nodes. For example, the computing nodes in the second device 220 can transmit data with other computing nodes or switching nodes through an orthogonal backplane 120. In this embodiment, the second device 220 integrates multiple computing nodes and multiple switching nodes, which can meet the needs of scenarios requiring a large amount of computing power, such as large models.

[0051] It's important to note that currently, to increase deployment density and shorten interconnection distances, supernode servers are deployed in rack-wide configurations, meaning one supernode server is deployed per rack. A single rack typically uses 64 or 128 cards as the minimum deployment unit, with 64 cards referring to 64 compute nodes per supernode server. However, while this type of rack is suitable for data centers or other large-scale deployment scenarios, edge devices have lower computational demands. Therefore, using 64 or 128 cards per rack as the minimum deployment unit presents issues of excessive computing power, wasted resources, and excessively high costs for edge devices or other users with lower computational demands, failing to meet the deployment requirements of the edge.

[0052] like Figure 3 As shown, according to another embodiment of the present invention, the first number of computing nodes in the first device 210 includes 32 graphics processors. For example, the first device 210 includes 8 computing nodes and 4 switching nodes, and each computing node includes 4 graphics processors, thus the first device 210 includes a total of 32 graphics processors. In this embodiment, the first device 210 deploys 32 graphics processors, that is, the first device 210 uses 32 cards as the minimum deployment unit. Compared with the method of using 64 cards or 128 cards as the minimum deployment unit, this can be applied to edge devices or other scenarios with smaller computing loads, alleviating the problems of excessive computing power, waste of resources, and excessive cost.

[0053] like Figure 3As shown, according to another embodiment of the present invention, the third number of computing nodes in the second device 220 includes 32 graphics processors. For example, the second device 220 includes 8 computing nodes and 4 switching nodes, and each computing node includes 4 graphics processors. Thus, the second device 220 includes a total of 32 graphics processors, thereby alleviating the problems of excessive computing power, waste of resources, and high costs.

[0054] In practical applications, a single server rack can deploy two independent devices, namely the first device 210 and the second device 220. The first device 210 and the second device 220 can form a 64-card computing node, meeting the needs of scenarios requiring high computing power, such as data centers. If the user's computing power requirement is lower, only the first device (210) or the second device (220) needs to be installed or run, without needing to install or run the other device, thus meeting the needs of edge devices, small-scale models, and other scenarios with lower computing power requirements.

[0055] In one example, the rack uses a 46U (46Unit, where Unit is a unit of server rack height) design, with a height of 2300mm, a depth of 1200mm, and a width of 600mm. The dimensions of the first and second devices within the rack can be 21 inches.

[0056] Those skilled in the art will understand that the features described in the various embodiments and / or claims of this utility model can be combined or combined in various ways, even if such combinations or combinations are not explicitly described in this utility model. In particular, the features described in the various embodiments and / or claims of this utility model can be combined or combined in various ways without departing from the spirit and teachings of this utility model. All such combinations and / or combinations fall within the scope of this utility model.

[0057] The specific embodiments described above do not constitute a limitation on the scope of protection of this utility model. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this utility model should be included within the scope of protection of this utility model.

Claims

1. A server rack, characterized in that, include: The rack has a first mounting area and a second mounting area; An orthogonal backplate is fixed to the frame. Cable tray, fixed to the frame; The first installation area is used to install a first device, the first device including a first node, the first node being used to be electrically connected to the orthogonal backplane to transmit data through the orthogonal backplane; The second installation area is used to install a second device, the second device including a second node, the second node being used to electrically connect to the cable tray to transmit data through the cable tray.

2. The cabinet according to claim 1, characterized in that, The first node has a first power interface, and the second node has a second power interface; the rack further includes a first power area and a second power area; the cabinet further includes: A conductive device, wherein the conductive device is a strip-shaped structure extending along a predetermined direction, and the conductive device includes a first conductive part, an insulating part, and a second conductive part arranged sequentially along the predetermined direction; A first power supply device is disposed in the first power area, and the first power supply device is electrically connected to the first power interface of the first node through the first conductive part. The second power supply device is disposed in the second power area and is electrically connected to the second power interface of the second node through the second conductive part.

3. The cabinet according to claim 2, characterized in that, The first power supply area, the first mounting area, the second mounting area, and the second power supply area are arranged sequentially along the predetermined direction.

4. The cabinet according to claim 1, characterized in that, The first node has a first liquid inlet and a first liquid outlet, and the second node has a second liquid inlet and a second liquid outlet; The cabinet also includes: The liquid inlet pipeline has a first pipeline outlet and a second pipeline outlet, wherein the first pipeline outlet is used for a sealed connection with the first liquid inlet, and the second pipeline outlet is used for a sealed connection with the second liquid inlet. The liquid outlet pipeline has a first pipeline inlet and a second pipeline inlet, the first pipeline inlet being used for a sealed connection with the first liquid outlet, and the second pipeline inlet being used for a sealed connection with the second liquid outlet.

5. The cabinet according to claim 4, characterized in that, The number of first nodes is multiple, the number of second nodes is multiple, and the multiple first nodes and multiple second nodes are arranged sequentially along a predetermined direction; The liquid inlet pipeline is a pipeline extending along the predetermined direction. The liquid inlet pipeline has multiple first pipeline outlets and multiple second pipeline outlets. The multiple first pipeline outlets are used to correspond one-to-one with and be sealed to the first liquid inlet of each of the multiple first nodes. The multiple second pipeline outlets are used to correspond one-to-one with and be sealed to the second liquid inlet of each of the multiple second nodes. The liquid outlet pipeline is a pipeline extending along the predetermined direction. The liquid outlet pipeline has multiple first pipeline inlets and multiple second pipeline inlets. The multiple first pipeline inlets are used to correspond one-to-one with and be sealed to the first liquid outlet of each of the multiple second nodes. The multiple second pipeline inlets are used to correspond one-to-one with and be sealed to the second liquid outlet of each of the multiple second nodes.

6. The cabinet according to claim 4, characterized in that, Also includes: A conductive device is provided, wherein the first power interface of the first node and the second power interface of the second node are respectively electrically connected to the power supply equipment through the conductive device. The conductive device is a strip-shaped structure extending in a predetermined direction. The inlet pipe and the outlet pipe are pipes extending in the predetermined direction, and the conductive device is located between the inlet pipe and the outlet pipe.

7. The cabinet according to claim 6, characterized in that, The frame has a first side and a second side opposite to the first side, and both the first mounting area and the second mounting area extend from the first side to the second side; the orthogonal back plate, the cable tray, the conductive device, the liquid inlet pipe and the liquid outlet pipe are all disposed on the second side.

8. A server system, characterized in that, include: The cabinet as described in any one of claims 1 to 7; A first device is installed in a first installation area of ​​the cabinet. The first device includes a first node, which is electrically connected to an orthogonal backplane in the cabinet to transmit data through the orthogonal backplane. as well as The second device is installed in a second mounting area of ​​the cabinet. The second device includes a second node that is electrically connected to a cable tray in the cabinet to transmit data through the cable tray.

9. The server system according to claim 8, characterized in that, The number of the first nodes is multiple, and the multiple first nodes include a first number of computing nodes and a second number of exchange nodes; The number of the second nodes is multiple, including a third number of computing nodes and a fourth number of exchange nodes.

10. The server system according to claim 9, characterized in that, The first number of computing nodes includes 32 graphics processors, and the third number of computing nodes includes 32 graphics processors.

Images