CN117311467A

CN117311467A - Computing system, control device, heat dissipation method and storage medium

Info

Publication number: CN117311467A
Application number: CN202211406326.5A
Authority: CN
Inventors: 汪昌祺; 刘洪广; 许慧锋; 郭中天
Original assignee: Huawei Technologies Co Ltd
Current assignee: Huawei Technologies Co Ltd
Priority date: 2022-06-29
Filing date: 2022-11-10
Publication date: 2023-12-29
Also published as: WO2024002000A1

Abstract

The application provides a computing system, a control device, a heat dissipation method and a storage medium, which can correspondingly control the opening of a flow regulating valve based on the condition that a computing load executes a service so as to regulate the flow of a liquid cooling channel of the computing load, thereby ensuring timely and accurate heat dissipation to the computing load. The computing system comprises a first control device, a detection device, a flow regulating valve and a computing load, wherein the computing load is used for executing a service; the detection device is used for detecting the service load information of the calculation load and sending the service load information to the first control device; the first control device is used for determining the target flow of the liquid cooling channel for calculating the load according to the service load information and determining the opening of the flow regulating valve according to the target flow; the first control device is also used for sending a control instruction, and the control instruction is used for adjusting the flow regulating valve to the opening degree so as to adjust the flow of the liquid cooling channel of the calculation load to the target flow.

Description

Computing system, control device, heat dissipation method and storage medium

The present application claims priority from chinese patent application No. 202210754960.1 entitled "a flow regulated heat dissipating system" filed on 29 of 2022, 06, the entire contents of which are incorporated herein by reference.

Technical Field

The present disclosure relates to the field of computing technologies, and in particular, to a computing system, a control device, a heat dissipation method, and a storage medium.

Background

As shown in fig. 1, a heat dissipation system of a conventional computing system, such as a data center, includes a cooling tower 101, a cooling liquid distribution unit (coolant distribution unit, CDU) 102, and a cabinet 103. The cooling tower 101 is connected to the CDU102 via a primary water supply pipe 111. The CDU102 is connected to the cabinet 103 through a secondary side water supply pipe 112. The CDU102 includes a heat exchanger 104. The secondary water supply pipe 112 takes away heat of the cabinet 103, and the heat flows into the heat exchanger 104 along with the flow direction of liquid cooling in the secondary water supply pipe 112. The cooling tower 101 exchanges heat with the secondary water supply pipe 112 by flowing through the primary water supply pipe 111 of the heat exchanger 104 to dissipate heat from the cabinet 103.

The primary water supply pipe 111 is connected to a flow rate regulating valve 113. When the control system of the machine room detects the temperature change of the heat dissipating liquid in the secondary water supply pipe 112 and indicates the change of the primary cooling capacity demand, the control system can adjust the opening of the flow rate adjusting valve 113, thereby adjusting the flow rate of the heat dissipating liquid in the primary water supply pipe 111. Thus, the flow rate adjusting valve is adjusted according to the temperature change of the heat dissipation liquid of the secondary side water supply pipe 112, the adjustment is not timely enough, and the heat dissipation efficiency is low.

Disclosure of Invention

The embodiment of the application provides a computing system, a control device, a heat dissipation method and a storage medium, which can correspondingly control the opening of a flow regulating valve based on the condition that a computing load executes a service so as to regulate the flow of a liquid cooling channel of the computing load, thereby ensuring timely and accurate heat dissipation to the computing load.

The first aspect of the application provides a computing system, which comprises a first control device, a detection device, a flow regulating valve and a computing load, wherein the computing load is used for executing a service; the detecting device is used for detecting the business load information of the calculation load and sending the business load information to the first control device; the first control device is used for determining the target flow of the liquid cooling channel of the calculation load according to the service load information and determining the opening of the flow regulating valve according to the target flow; the first control device is further used for sending a control instruction, and the control instruction is used for adjusting the flow regulating valve to the opening degree so as to adjust the flow of the liquid cooling channel of the calculation load to the target flow.

According to the scheme, the first control device obtains a control instruction according to service load information related to the service execution of the computing load, the second control device adjusts the flow regulating valve to the opening degree included in the control instruction so as to regulate the flow of the liquid cooling channel of the computing load to the target flow, and the cooling capacity generated by the target flow is matched with the heat dissipation requirement of the service execution of the computing load, so that timely and accurate heat dissipation of the computing load is realized.

Based on the first aspect, in an optional implementation manner, the computing load is one computing node, a plurality of computing nodes, or a computing node in one cabinet.

By adopting the implementation mode, if the calculation load is one calculation node, the control signaling can independently adjust the flow of the liquid cooling channel of the single calculation node, so that the accurate heat dissipation of the single calculation node is realized. If the computing load is a plurality of computing nodes, the control signaling can synchronously regulate the flow of the liquid cooling channels of the plurality of computing nodes, so that the efficiency of regulating the flow of the liquid cooling channels is improved. If the computational load is a computational node in a cabinet, the control signaling can timely and accurately dissipate heat of the cabinet.

Based on the first aspect, in an optional implementation manner, the computing load belongs to a cabinet, and the first control device is a management device of the cabinet.

By adopting the implementation mode, the first control device and the computing load are fixed on the same cabinet, so that the first control device is convenient to maintain and configure.

Based on the first aspect, the first control device is configured to manage computing loads in a plurality of racks in the computing system.

By adopting the implementation mode, the first control device can manage the calculation loads of the cabinets so as to obtain the corresponding control signaling, and the heat dissipation efficiency of the cabinets is improved.

Based on the first aspect, the first control device is configured to send the control instruction to a second control device, where the control instruction includes the opening, and the second control device is configured to control the flow rate adjustment valve to the opening according to the control instruction.

By adopting the implementation mode, the second control device controls the opening of the flow regulating valve, so that the timely and accurate heat dissipation of the calculation load is realized.

Based on the first aspect, in an optional implementation manner, the first control device is configured to determine the target flow according to an artificial intelligence AI model and the traffic load information.

By adopting the implementation mode, the target flow is determined through the AI model, so that the efficiency and accuracy for obtaining the target flow of the liquid cooling channel of the calculation load can be improved.

Based on the first aspect, in an optional implementation manner, the first control device is configured to obtain the traffic load information and flow information of the liquid cooling channel of the calculation load reported by the second control device, and is configured to train the AI model according to the traffic load information and the flow information.

By adopting the implementation mode, the first control device determines the control signaling together according to the service load information and the flow information of the liquid cooling channel of the calculation load, so that the accuracy of the second control device for adjusting the flow of the liquid cooling channel of the calculation load according to the control signaling is ensured, the accurate heat dissipation of the calculation load is ensured, and the waste of cold energy generated by the flow of the liquid cooling channel of the calculation load is reduced.

Based on the first aspect, in an optional implementation manner, a relationship between the flow rate of the liquid cooling channel and the opening of the flow regulating valve is stored in the first control device, and the first control device is configured to determine the opening of the flow regulating valve according to the relationship and the target flow rate.

By adopting the implementation mode, the first control device ensures the successful acquisition of the opening according to the relation between the flow of the liquid cooling channel and the opening of the flow regulating valve, so that the flow regulating valve can successfully regulate the flow of the liquid cooling channel for calculating the load.

Based on the first aspect, in an optional implementation manner, the first control device is configured to determine power consumption according to the traffic load information, and determine the target traffic according to the power consumption.

By adopting the implementation mode, the first control device obtains the corresponding power consumption according to the service load information, and obtains the target flow according to the power consumption, so that the cooling capacity generated by calculating the target flow of the liquid cooling channel of the load is ensured, and the heat dissipation requirement in the process of calculating the service of the load can be met.

Based on the first aspect, in an optional implementation manner, the first control device is further configured to send an adjustment instruction to the detection device, where the detection device is configured to control power consumption corresponding to the computing load execution service according to the adjustment instruction.

By adopting the implementation mode, the first control device can adjust the power consumption of the execution service of the computing load according to the adjustment instruction, effectively ensures the cold quantity generated by the liquid cooling channel, can reduce the waste of the cold quantity generated by the liquid cooling channel under the condition of meeting the heat dissipation requirement of the computing load, and realizes the balanced heat dissipation of the computing load.

Based on the first aspect, in an optional implementation manner, after the second control device detects a failure of the liquid cooling channel of the computing load, the flow control valve is controlled to be turned off.

By adopting the implementation mode, if the liquid cooling channel of the calculation load fails, the second control device can switch off the flow regulating valve in time, so that the heat dissipation liquid can not continuously flow into the failed liquid cooling channel, isolation of the failed liquid cooling channel is realized, normal operation of other liquid cooling channels which do not fail can not be influenced, and the reliability of the calculation system is improved.

Based on the first aspect, in an optional implementation manner, the second control device is a baseboard management controller BMC, and the detection device is a scheduler.

Based on the first aspect, in an optional implementation manner, the service load information is used to indicate a computing resource occupied by the computing load to execute the service.

By adopting the implementation mode, the first control device performs flow control according to the computing resources occupied by the computing load execution service, and when the computing resources occupied by the computing load execution service are changed, the first control device can be triggered to control the flow of the liquid cooling channel in time, so that the timeliness of the flow control of the liquid cooling channel is improved.

Based on the first aspect, in an optional implementation manner, the first control device is configured to obtain a target computing resource occupied by a service executed by the computing load; obtaining a first power consumption corresponding to the target computing resource; and obtaining the target flow corresponding to the first power consumption, wherein the first power consumption and the target flow are in positive correlation.

According to the implementation mode, due to different computing load hardware conditions, corresponding target flow is obtained according to the first power consumption corresponding to the target computing resources occupied by the computing load execution service, and the accuracy of flow control is improved.

Based on the first aspect, in an optional implementation manner, the service load information is used to indicate a service type of the computing load executing service.

By adopting the implementation mode, the first control device performs flow control according to the service type of the service executed by the computing load, so that the accuracy and timeliness of heat dissipation of the computing load are improved.

Based on the first aspect, in an optional implementation manner, the first control device is configured to obtain second power consumption corresponding to a service type of the computing load execution service; and obtaining the target flow corresponding to the second power consumption, wherein the second power consumption and the target flow are in positive correlation.

According to the implementation mode, due to different computing load hardware conditions, corresponding target flow is obtained according to the second power consumption corresponding to the target computing resources occupied by the computing load execution service, and the accuracy of flow control is improved.

Based on the first aspect, in an optional implementation manner, the service load information is used to indicate a computing resource occupied by a service that has been executed by the computing load, and/or the service load information is used to indicate at least one of the computing resources occupied by the service to be executed by the computing load.

By adopting the implementation mode, the first control device can adjust the flow according to the service executed by the computing load or the service to be executed by the computing load, effectively ensures the target flow adjusted by the flow adjusting valve, and can meet the requirement of the computing load on the cold when executing the service.

Based on the first aspect, in an optional implementation manner, the adjustment instruction is configured to instruct at least one of the following:

the computing load performs a power saving operation to reduce power consumption corresponding to executing traffic, migration of at least a portion of the traffic executed by the computing load to another computing load, or the computing load executing at least a portion of the traffic from another computing load.

Based on the first aspect, in an optional implementation manner, the first control device is further configured to obtain a heat dissipation status message of the liquid cooling channel, where the heat dissipation status message is related to heat dissipation liquid flowing through the liquid cooling channel; the first control device is configured to obtain the control instruction according to the service load information and the heat dissipation status message.

By adopting the implementation mode, the first control device obtains the corresponding control instruction according to the service load information and the heat dissipation state information, so that the accurate heat dissipation of the calculation load is realized.

A second aspect of the present application provides a control apparatus, comprising: the receiving module is used for receiving the service load information of the calculation load from the detection device, wherein the calculation load is used for executing the service; the processing module is used for determining the target flow of the liquid cooling channel of the calculation load according to the service load information and determining the opening of the flow regulating valve according to the target flow; and the sending module is used for sending a control instruction, and the control instruction is used for adjusting the flow regulating valve to the opening degree so as to regulate the flow of the liquid cooling channel of the calculation load to the target flow.

For an explanation of the beneficial effects of this aspect, please refer to the first aspect, and detailed descriptions thereof are omitted.

Based on the second aspect, in an optional implementation manner, the sending module is further configured to send the control instruction to another control device, where the control instruction includes the opening, and the control instruction is used to control the other control device to adjust the flow rate adjustment valve to the opening.

Based on the second aspect, in an optional implementation manner, the processing module is configured to determine the target traffic according to an artificial intelligence AI model and the traffic load information.

Based on the second aspect, in an optional implementation manner, the processing module is configured to determine power consumption according to the traffic load information, and determine the target traffic according to the power consumption.

A third aspect of the present application provides a heat dissipation method, the method including: the first control device receives service load information of a calculation load from the detection device, wherein the calculation load is used for executing service; the first control device determines the target flow of the liquid cooling channel of the calculation load according to the service load information, and determines the opening of a flow regulating valve according to the target flow; the first control device sends a control instruction which is used for adjusting the flow regulating valve to the opening degree so as to regulate the flow of the liquid cooling channel of the calculated load to the target flow.

Based on the third aspect, in an optional implementation manner, the sending, by the first control device, a control instruction includes: the first control device sends the control instruction to the second control device, wherein the control instruction comprises the opening degree, and the second control device controls the flow regulating valve to the opening degree according to the control instruction.

Based on the third aspect, in an optional implementation manner, the determining, by the first control device, the target flow of the liquid cooling channel of the computing load according to the service load information includes: the first control device determines the target flow according to an artificial intelligence AI model and the service load information.

Based on the third aspect, in an optional implementation manner, the determining, by the first control device, the target flow of the liquid cooling channel of the computing load according to the service load information includes: the first control device determines power consumption according to the service load information and determines the target flow according to the power consumption.

A fourth aspect of the present application provides a computer readable storage medium comprising computer program instructions which, when executed by a computer, perform the method of any of the first aspects above.

Drawings

FIG. 1 is a diagram illustrating an example architecture of a prior computing system;

FIG. 2 is a diagram of a first exemplary architecture of a computing system provided herein;

FIG. 3 is a diagram of a second exemplary architecture of a computing system provided herein;

FIG. 4 is a diagram illustrating an exemplary architecture of a computing system according to one embodiment of the present application;

FIG. 5 is a flowchart illustrating a heat dissipation method according to an embodiment of the present disclosure;

FIG. 6 is a diagram illustrating an exemplary architecture of a computing system according to a second embodiment of the present disclosure;

fig. 7 is a flow chart of steps of a heat dissipation method according to a second embodiment of the present disclosure;

fig. 8 is a step flowchart of a heat dissipation method according to a third embodiment of the present application;

fig. 9 is a step flowchart of a heat dissipation method according to a fourth embodiment of the present application;

fig. 10 is a diagram illustrating a first exemplary structure of an electronic device provided in the present application;

fig. 11 is a diagram illustrating a second exemplary structure of the electronic device provided in the present application.

Detailed Description

The following description of the embodiments of the present application will be made clearly and fully with reference to the accompanying drawings, in which it is evident that the embodiments described are only some, but not all, of the embodiments of the present application. All other embodiments, which can be made by those skilled in the art based on the embodiments herein without making any inventive effort, are intended to be within the scope of the present application.

Fig. 2 is a diagram of a first exemplary architecture of a computing system provided herein. The computing system shown in fig. 2 may be applied to a data center, a supercomputer (super computer), a monitoring center, an industrial control center, etc., without limitation.

The computing system includes a cabinet 210. The cabinet 210 is connected to a CDU220, and the CDU220 is also connected to a cooling tower 230. The cooling tower 230 is connected to the CDU220 through a primary side water supply pipe. Wherein the primary side water supply pipe includes a first pipe 241 and a second pipe 242. Wherein the heat rejection liquid exiting the cooling tower 230 flows into the CDU220 via the flow direction of the first pipe 241. The liquid exiting the CDU220 flows into the cooling tower 210 via the flow direction of the second pipe 242. The CDU220 is connected to the cabinet 230 through a secondary side water supply pipe. The secondary side water supply pipe includes a water inlet pipe 243 and a water outlet pipe 244. The heat sink liquid flowing from the CDU220 flows into the cabinet 210 via the flow direction of the water inlet pipe 243. The heat rejection liquid exiting the cabinet 210 flows into the CDU220 via the flow direction of the outlet pipe 244. Specifically, the heat dissipating liquid flows into the cabinet 210 via the water inlet pipe 243, and takes away heat in the cabinet 210, and the heat in the cabinet 210 flows into the CDU220 via the flow direction of the heat dissipating liquid in the water outlet pipe 244. The cooling tower 230 is capable of heat exchanging within the CDU220.

The cabinet 210 also includes a computing load 211, the present example does not limit the number of computing loads 211 that the cabinet 210 includes. The computing load may be N computing nodes, where N is any positive integer greater than or equal to 1 as shown in this example. For example, N computing nodes may belong to the same enclosure 210. As another example, N computing nodes may also belong to multiple racks. The computing node is used for executing the service, and the computing node includes, but is not limited to, a processor, a memory, a substrate controller (baseboard management controller, BMC), a communication interface, and the like. During the process of executing the service, the processor, the memory and the like in the computing node generate heat, and the higher the service load is, the higher the consumed power is, and the higher the generated heat is. The computing system further includes a liquid cooling channel for dissipating heat generated by each computing node, and a flow regulating valve for regulating a flow of a cooling liquid in the liquid cooling channel, wherein the cooling liquid flowing through the liquid cooling channel is capable of taking away heat generated by the computing node and flowing to the CDU220.

Fig. 3 is a diagram illustrating a second exemplary architecture of a computing system provided herein. The computing system shown in this example is specifically a computer cluster (computer cluster). The computer cluster specifically includes a first control device 301, a second control device 303, a detection device 302, and a flow control valve 421. Wherein the first control device 301 is connected to the second control device 303 and the detection device 302, respectively. The second control device 303 is connected to the flow rate adjustment valve 421. The flow regulating valve 421 is any flow regulating valve fixed to the cabinet 210, and the number of flow regulating valves included in the cabinet 210 is not limited in this example.

The first control device 301 may be fixed to the cabinet, and the first control device 301 is a management device of the cabinet. For example, the top of the cabinet is used to secure the first control device 301. As another example, the side of the cabinet is used to secure the first control device 301. The first control device 301 shown in this example may also be located at the distal end of the cabinet. The detection means 302 provided in the computer cluster are capable of allocating computing resources for executing traffic to the computing nodes such that high priority traffic can be executed preferentially. The detection means 302 may monitor the execution status of a computing node executing a service and change the allocation of resources to the service according to a policy. The detection means 302 may be a scheduler and is adapted to collect traffic load information of the computing node 401, which traffic load information is related to the traffic performed by the computing node 401. For example, the traffic load information is used to indicate the computational resources occupied by the traffic performed by the compute node 401 and/or to indicate the type of traffic performed by the compute node 401. The detection device 302 can report the traffic load information collected from the computing node 401 to the first control device 301. In this example, the first control device 301 and the detection device 302 are two independent devices connected to each other, and in other examples, the first control device 301 may also configure the detection device 302 so that the first control device 301 can also implement the function of the detection device.

The second control device 303 can adjust the opening of the flow rate adjustment valve 421, thereby achieving the purpose of adjusting the flow rate of the heat dissipation liquid in the liquid cooling channel communicated with the flow rate adjustment valve 421. For example, the second control device 303 may be a BMC of a computing node. One interface of the second control device 303 is connected to the flow rate adjustment valve 421, and the other interface of the second control device 303 is connected to the first control device 301. Based on the computing systems shown in fig. 2 and 3, the following describes the implementation procedure of the heat dissipation method provided in the present application in connection with the specific embodiment.

Example 1

Based on the computing system shown in fig. 2 and 3, the structure of the liquid cooling channel and the flow regulating valve is specifically described in this embodiment in conjunction with the structure shown in fig. 4, where fig. 4 is a structural example diagram of the computing system according to the first embodiment of the present application.

The example shown in fig. 4 takes the computing load as one computing node as an example. The liquid cooling channels for cooling the compute node 401 include a split water pipe 413, channels in the liquid cooling plate 402, and a split water pipe 414, which are communicated. Specifically, to implement heat dissipation to the computing node 401, the computing system further includes a liquid cooling plate 402. The heat dissipation liquid flowing through the liquid cooling plate 402 is used for dissipating heat generated during the operation of the computing node 401. The liquid cooling plate 402 in this embodiment may be in direct contact with the computing node 401, so as to implement heat dissipation of the computing node 401 by the liquid cooling plate 402. As another example, a thermally conductive connection may be included between the liquid cooling plate 402 and the compute node 401. One end of the heat conduction connecting piece is in contact with the liquid cooling plate 402, the other end of the heat conduction connecting piece is in contact with the computing node 401, heat generated by the computing node 401 is transferred to the liquid cooling plate 402 through the heat conduction connecting piece, and cooling liquid in the liquid cooling plate 402 cools and dissipates heat. Optionally, the liquid cooling plate 402 is removably mounted within the cabinet 210 for ease of maintenance. The heat conducting connecting piece can be a heat pipe or a heat conducting material such as a heat conducting metal block, so as to ensure that heat generated by the computing node 401 can be timely and efficiently transferred to the liquid cooling plate 402.

The secondary side water supply pipe comprises a water inlet pipe and a water outlet pipe. The water inlet pipe specifically comprises a plurality of main water inlet pipes and branch water inlet pipes communicated with each main water inlet pipe. The water outlet pipe specifically comprises a plurality of main water outlet pipes and a branch water outlet pipe communicated with each main water outlet pipe. The water-dividing pipe 413 and the water-dividing pipe 414 shown in fig. 3 are used for radiating heat to the computing node 401. One port of the water dividing pipe 413 is communicated with the main water inlet pipe 411, and the other port is communicated with the liquid cooling plate 402. One end of the water dividing pipe 414 is communicated with the liquid cooling plate 402, and the other end of the water dividing pipe 414 is communicated with the main water outlet pipe 412. It will be appreciated that the main water inlet pipe 411, the sub water inlet pipe 413, the channels in the liquid cooling plate 402, the sub water outlet pipe 414 and the main water outlet pipe 412 form a closed liquid loop for realizing the flow of the heat dissipating liquid. That is, the heat radiation liquid flowing out of the CDU220 flows into the CDU220 again through the main water inlet pipe 411, the branch water inlet pipe 413, the passage in the liquid cooling plate 402, the branch water outlet pipe 414, and the main water outlet pipe 412 in this order, so that the CDU220 exchanges heat with the heat radiation liquid flowing in from the main water outlet pipe 412.

In this embodiment, a cold plate type liquid cooling system is taken as an example of a computing system, that is, a liquid cooling plate for radiating heat to a computing node is independently inserted into and pulled out from a cabinet, so that the liquid cooling plate is convenient to replace, and the maintenance efficiency of the liquid cooling plate is improved. In other examples, the computing system may also employ an immersion liquid cooling system, where the immersion liquid cooling means that the computing system includes a heat dissipation node, where the heat dissipation node includes an outer shell and a heat dissipation liquid contained in the outer shell, and the computing node is directly immersed in the heat dissipation node, that is, the computing node is immersed in the heat dissipation liquid contained in the heat dissipation node, so that the computing node is directly contacted with the heat dissipation liquid, and heat dissipation efficiency of dissipating heat from the computing node is improved.

The liquid cooling channels used to dissipate heat from the compute node 401 are also in communication with a flow regulating valve 421. The flow control valve 421 in this embodiment is used to control the flow of the heat sink in the liquid cooling channel of the computing node 401. The liquid cooling channels for cooling the computing node 401 include a split water pipe 413, a channel in the liquid cooling plate 402, and a split water pipe 414. The flow regulating valve 421 is communicated with the split water pipe 413 of the liquid cooling channel. In other examples, the flow rate adjusting valve 421 may also be connected to the water outlet pipe 414 of the liquid cooling channel, which is not limited in this embodiment.

The flow rate adjustment valve 421 may also be referred to as a self-operated flow rate control valve, a flow rate balancing valve, a static balancing valve, etc., and is not limited in this embodiment. The type of the flow rate adjustment valve 421 in the present embodiment is not limited as long as the flow rate adjustment valve 421 can adjust the flow rate of the heat radiation liquid flowing through the liquid cooling passage, and for example, the flow rate adjustment valve 421 may be a proportional solenoid valve.

Based on the computing systems shown in fig. 2 to 4, the following description is made with reference to fig. 5 for the execution of the heat dissipation method provided in the present embodiment, where fig. 5 is a flowchart illustrating steps of the heat dissipation method provided in the first embodiment of the present application.

Step 501, the detecting device obtains service load information.

The detection device shown in this embodiment is configured to obtain the traffic load information of the computing node 401. The traffic load information may be at least one of the following message types:

message type 1

The traffic load information detected by the detecting means 302 is used to indicate the computing resources occupied by the computing node 401 having performed the traffic. For example, the computing resources may be utilization of processors of computing node 401, memory occupancy, input/output (I/O) bandwidth, and so on.

Message type 2

The traffic load information detected by the detecting means 302 is used to indicate the computing resources occupied by the service to be executed by the computing node 401, and it can be understood that the detecting means 302 estimates the computing resources occupied by the service to be executed by the computing node 401 at a future time. For example, the computing node 401 may store a service execution list including a plurality of service identities indicating priorities for execution by the computing node 401, e.g., as shown in table 1:

TABLE 1

Service identifier A1
	Service identifier A2
Service identifier A3

The service identifier may be a name of an application executing the service, and as shown in this example, when the computing node 401 executes the service with the service identifier A1, the computing node 401 re-executes the service with the service identifier A2, and so on, when the computing node 401 completes executing the service with the service identifier A2, the computing node 401 re-executes the service with the service identifier A3. It will be appreciated that if the detecting means 302 determines that the service currently executed by the computing node 401 is a service having the service identifier A2, then the detecting means 302 determines that the service to be executed by the computing node 401 is a service having the service identifier A3. For this purpose, the detecting means 302 predicts the computing resources occupied by the computing node 401 executing the service with the service identity A3 and generates the service load information. It will be appreciated that the traffic load information is used to instruct the computing node 401 to execute the computing resources occupied by the traffic with the traffic identification A3.

Message type 3

The traffic load information detected by the detecting means 302 is used to indicate the traffic type of the traffic that the computing node 401 has performed. For example, the service type may be an supercomputer service, an artificial intelligence (artificial intelligence, AI) service, etc., which will not be described in detail in this embodiment.

Message type 4

The service load information detected by the detecting device 302 is used for indicating a service type of the service to be executed by the computing node 401, and for the description of the service to be executed by the computing node 401, please refer to the message type 2, for the description of the service type, please refer to the message type 3, which is not described in detail.

The detecting device shown in this embodiment may report the service load information to the first control device periodically, or may report the service load information to the first control device when the detecting device determines that the service load information has changed quite, which is not limited in this embodiment. For example, if the traffic load information is of message type 1, if the detecting device determines that the difference between the obtained computing resources of two adjacent times is greater than or equal to the threshold value, the current traffic load information is reported to the first control device. As another example, the first control device may send a trigger message to the detection device, and the detection device obtains the traffic load information according to the trigger message.

Step 502, the detecting device sends the service load information to the first control device.

Step 503, the first control device sends a control instruction to the second control device according to the service load information.

The first control device determines the target flow of the liquid cooling channel of the computing node according to the service load information from the detection device. The first control device stores a relationship between the flow rate of the liquid cooling passage and the opening degree of the flow rate adjustment valve, and for this purpose, the first control device obtains the opening degree of the flow rate adjustment valve corresponding to the target flow rate based on the relationship. And the first control device sends a control instruction to the second control device, wherein the control instruction comprises the opening degree. The second control device can adjust the flow regulating valve to the opening according to the control signaling, so that the flow regulating valve can adjust the flow of the liquid cooling channel of the computing node to the target flow.

The first control device shown in this embodiment may store a correspondence list as shown in table 2, which creates a correspondence between the second control device identifier and the computing node identifier.

TABLE 2

Identification of the second control device	Computing node identification
		Second control device B1	Computing node B2
Second control device C1	Computing node C2
		...	...
Second control device N1	Computing node N2

The computing node identified as B2 in table 2 corresponds to the second control device identified as B1, and has a liquid cooling channel controlled by the second control device identified as B1 for dissipating heat from the computing node identified as B2. The computing node with the identifier of C2 corresponds to the second control device with the identifier of C1, and then the liquid cooling channel controlled by the second control device with the identifier of C1 is provided for radiating heat for the computing node with the identifier of C2. And by analogy, the computing node with the identifier N2 corresponds to the second control device with the identifier N1, and then the liquid cooling channel controlled by the second control device with the identifier N1 is used for radiating heat for the computing node with the identifier N2. For this purpose, the traffic load information may also carry an identification of the computing node, e.g. the traffic load information sent by the detection means 302 to the first control means 301 also carries an identification B2 of the computing node 401. The first control device 301 obtains a corresponding control instruction according to the traffic load information from the detection device 302, and the first control device 301 obtains an identification B1 of the second control device corresponding to the identification B2 of the calculation node 402 according to the correspondence relationship as shown in table 2, the control instruction including the opening degree of the flow rate adjustment valve 421. The first control device sends the control instruction to the second control device with the identifier B1, so that the second control device with the identifier B1 can adjust the flow of the heat dissipation liquid in the liquid cooling channel communicated with the flow adjusting valve 421, so as to adjust the flow of the liquid cooling channel of the computing node 401 to the target flow. The following describes alternative ways for the first control device to obtain the control instruction according to the traffic load information:

Mode 1

If the traffic load information is at least one of the message type 1 or the message type 2 shown in step 501, first, the first control device obtains a first computing resource, where the first computing resource is a computing resource occupied by a service executed by the computing node 401. For example, if the traffic load information is message type 1, then the first computing resource is the computing resource occupied by the traffic that has been performed by the computing node 401. If the traffic load information is message type 2, the first computing node is the computing resource occupied by the traffic to be executed by the computing node 401. Next, the first control device obtains a first power consumption corresponding to the first computing resource. Different computing nodes occupy the same computing resources due to the difference of hardware of different computing nodes, but different power consumption is brought, and in order to realize accurate heat dissipation of the computing nodes, the first control device needs to obtain the first power consumption of the computing nodes under the condition that the service executed by the computing nodes occupies the first computing resources. For this purpose, the first control device may create, in advance, a first correspondence according to a hardware condition of the computing node, where the first correspondence has created a correspondence between the first computing resource and the first power consumption. It may be appreciated that, in the case where the first control device obtains that the service executed by the computing node occupies the first computing resource, the corresponding first power consumption is obtained by querying the first correspondence. Again, the first control means obtains a target flow rate corresponding to the first power consumption. The first power consumption and the target flow are in positive correlation. It will be understood that, if the first power consumption obtained by the first control device is larger, the target flow obtained by the first control device is larger, and similarly, if the first power consumption obtained by the first control device is smaller, the target flow obtained by the first control device is smaller. Finally, the first control device obtains the opening degree of the flow regulating valve corresponding to the target flow rate, and sends a control instruction including the opening degree to the second control device.

Optionally, if the traffic load information indicates both the message type 1 and the message type 2, the first control device obtains the power consumption P1 corresponding to the traffic load information indicating the message type 1 and the power consumption P2 corresponding to the traffic load information indicating the message type 2, respectively. The first control device obtains a corresponding control instruction according to the maximum value of the power consumption P1 and the power consumption P2. For example, if the first control device determines that the power consumption P1 is greater than the power consumption P2, the first control device obtains a control instruction corresponding to the power consumption P1, and similarly, if the first control device determines that the power consumption P1 is less than the power consumption P2, the first control device obtains a control instruction corresponding to the power consumption P2.

Mode 2

If the traffic load information is the message type 3 or the message type 4 indicated in step 501, first, the first control device obtains the second power consumption corresponding to the traffic type executed by the computing node. Specifically, the first control device creates a second corresponding relation in advance, and the second corresponding relation has created corresponding relations between different service types and power consumption. It can be understood that, in the case that the first control device obtains the service type executed by the computing node, the corresponding second power consumption is obtained by querying the second correspondence. And the first control device obtains a target flow corresponding to the second power consumption, wherein the second power consumption and the target flow are in positive correlation. It will be appreciated that the greater the second power consumption obtained by the first control device, the greater the target flow obtained by the first control device, and likewise, the smaller the second power consumption obtained by the first control device, the smaller the target flow obtained by the first control device. Finally, the first control device obtains the opening degree of the flow regulating valve corresponding to the target flow rate, and sends a control instruction including the opening degree to the second control device.

Alternatively, if the traffic load information indicates both the message type 3 and the message type 4, the first control device obtains the power consumption P3 corresponding to the traffic load information indicating the message type 3 and the power consumption P4 corresponding to the traffic load information indicating the message type 4, respectively. The first control device obtains a corresponding control instruction according to the maximum value of the power consumption P3 and the power consumption P4. For example, if the first control device determines that the power consumption P3 is greater than the power consumption P4, the first control device obtains a control instruction corresponding to the power consumption P3, and similarly, if the first control device determines that the power consumption P3 is less than the power consumption P4, the first control device obtains a control instruction corresponding to the power consumption P4.

It can be understood that, if the first control device determines, according to the traffic load information, the power consumption P1, the power consumption P2, the power consumption P3, and the power consumption P4, the first control device obtains the target traffic corresponding to the maximum value of the power consumption P1, the power consumption P2, the power consumption P3, and the power consumption P4.

Alternatively, the first control device shown in this embodiment may input the traffic load information into an AI model, and obtain, by using the AI model, the corresponding target flow using the traffic load information. To obtain the AI model, the first control device creates a data set in advance, which is the traffic load information historically received by the first control device, and may perform preprocessing on the data set, for example, performing descriptive statistics, data visualization, data shaping, data segmentation, or the like on the data set. The first control device trains the data set through a machine learning algorithm to obtain the AI model. The target flow is obtained through the AI model, so that the efficiency and accuracy of obtaining the flow of the liquid cooling channel can be improved.

Step 504, the second control device adjusts the flow rate adjustment valve to an opening included in the control signaling.

As further shown in fig. 3 and 4, to implement heat dissipation to the computing node 401, the first control device 301 sends a control instruction to the second control device 303 according to the traffic load information from the detection device 302. The second control device 303 controls the flow rate adjustment valve 421 to the opening included in the control command such that the flow rate adjustment valve 421 will adjust the flow rate of the heat radiation liquid of the liquid cooling passage that radiates heat for the computation node 401 to the target flow rate. The cooling capacity generated by the target flow of the cooling liquid in the liquid cooling channel meets the cooling requirement of the computing node 401. Specifically, the liquid cooling channels for cooling the computing node 401 include a water dividing pipe 413, a channel of the liquid cooling plate 402, and a water dividing pipe 414. The second control device controls the opening of the flow rate adjusting valve 421 according to the control command to adjust the flow rate of the heat dissipation liquid in the liquid cooling channel to the target flow rate.

In this embodiment, steps 501 to 504 may be repeatedly performed until the cooling capacity generated by the flow of the liquid cooling channel regulated by each flow regulating valve can meet the cooling capacity required by cooling of the computing node, so that the cooling capacity generated by the flow of the liquid cooling channel regulated by the flow regulating valve is equal to or approximately equal to the cooling capacity required by cooling of the computing node as much as possible under the condition of cooling of each computing node, and the waste of the cooling capacity is reduced.

The cabinet shown in this embodiment is provided with a plurality of computing nodes, and the heat dissipation requirements of the execution services of different computing nodes are different, so that the first control device can obtain corresponding service load information according to the condition that each computing node executes the services, so as to realize mutual inductance intercommunication between the first control device and the services executed by the computing nodes at the bottom layer. The first control device can obtain a corresponding control instruction according to the service load information, and the second control device can adjust the flow of the heat dissipation liquid of the liquid cooling channel for dissipating heat of the computing nodes according to the control instruction, so that independent heat dissipation of each computing node is realized, and accurate heat dissipation of each computing node is realized. The independent heat dissipation of each computing node can be realized, so that the refined control of the heat dissipation liquid flow in each liquid cooling channel is realized. And the first control device obtains the control instruction according to the service load information related to the service executed by the computing node, so that the second control device adjusts the opening of the flow regulating valve according to the control instruction, and the target flow of the heat dissipation liquid in the liquid cooling channel can be matched with the heat dissipation requirement of the service executed by the computing node.

The first control device obtains a control instruction according to the condition that the computing node executes the service, so that timeliness and reaction speed of adjusting the flow of the heat dissipation liquid in the liquid cooling channel are improved, for example, under the condition that the power consumption of the computing node executes the service is increased or reduced, the first control device can immediately obtain the control instruction, and therefore the flow of the liquid cooling channel is adjusted.

In addition, if the second control device detects that the liquid cooling channel has a leakage condition of heat dissipation liquid, the second control device can directly control the flow control valve communicated with the liquid cooling channel with leakage so that the heat dissipation liquid cannot continuously flow into the liquid cooling channel with leakage, thereby facilitating maintenance or replacement of the liquid cooling channel, avoiding continuous leakage of the heat dissipation liquid, realizing single-point isolation of the liquid cooling channel with leakage, avoiding influence on normal operation of other liquid cooling channels without leakage, and improving reliability of a computing system. It should be clear that, in this example, taking the second control device as an example to determine that the liquid cooling channel of the computing node has liquid leakage and control the flow regulating valve to be turned off, in other examples, the second control device may control the flow regulating valve to be turned off after detecting that the liquid cooling channel of the computing node has any type of fault.

Each flow regulating valve in the computing system shown in the embodiment can independently regulate the flow of one liquid cooling channel, so that computing nodes with different heat dissipation requirements can be deployed in the cabinet, the utilization rate of the computing nodes deployed in the cabinet is improved, and the deployment cost is reduced.

Example two

In a first embodiment, a flow control valve is capable of controlling the flow of cooling fluid in a fluid cooling channel to thereby dissipate heat from a computing node. In the second embodiment, one flow regulating valve can simultaneously regulate the flow of the heat dissipation liquid in the plurality of liquid cooling channels, so as to dissipate heat for the plurality of computing nodes, and the first control device sends a control instruction to the second control device so as to achieve the purpose of simultaneously dissipating heat for the plurality of computing nodes. Fig. 6 is a diagram illustrating an exemplary structure of a computing system according to a second embodiment of the present application.

The computing system shown in this embodiment includes a cabinet, where the cabinet includes a computing load, and the detailed description of the cabinet is referred to in fig. 2, which is not repeated. One computational load shown in this embodiment is M computational nodes, where M is any positive integer greater than 2. In this embodiment, the computing load includes two computing nodes, namely, the computing node 611 and the computing node 612, and for a specific description of each computing node, please refer to the description of fig. 2, details are not repeated. The computing system includes a liquid cooling channel 621 for cooling the computing node 611 and a liquid cooling channel 622 for cooling the computing node 612, which are described in detail in the first embodiment and not described in detail. In this embodiment, the liquid cooling channel 621 and the liquid cooling channel 622 are connected in parallel, and in the case where the liquid cooling channel 621 and the liquid cooling channel 622 are connected in parallel, the computing system further includes a liquid cooling adjusting valve 632 that is respectively connected to the liquid cooling channel 621 and the liquid cooling channel 622, so that the flow adjusting valve 632 can simultaneously adjust the flow rates of the liquid cooling channel 621 and the liquid cooling channel 622, and the flow rate of the heat dissipation liquid flowing through the liquid cooling channel 621 is the same as the flow rate of the heat dissipation liquid flowing through the liquid cooling channel 622. In this embodiment, a plurality of liquid cooling channels communicating with the same flow rate adjusting valve are connected in parallel, and in other examples, a plurality of liquid cooling channels communicating with the same flow rate adjusting valve may also be connected in series, so as to save the length of the pipeline, which is not limited in this embodiment.

Based on the computing system shown in fig. 6, the following description is made with reference to the execution process of the heat dissipation method provided in the present embodiment shown in fig. 7, where fig. 7 is a step flowchart of the heat dissipation method provided in the second embodiment of the present application. In the execution process shown in fig. 5, one flow rate adjusting valve only adjusts the flow rate of the heat dissipation liquid in one liquid cooling channel, whereas in the execution process shown in fig. 7, one flow rate adjusting valve can simultaneously adjust the flow rates of the heat dissipation liquid in a plurality of liquid cooling channels.

Step 701, the detecting device obtains service load information.

In the computing system shown in fig. 6, the computing system further includes detecting means for detecting a situation in which the computing node performs a service to obtain the service load information, and it is understood that the detecting means is used to obtain the service load information from the computing node 611 and the service load information from the computing node 612. The detecting device reports the traffic load information of the computing node 611 and the traffic load information of the computing node 612 to the first control device, and description of the detecting device, the traffic load information and the first control device is shown in the first embodiment, which is not repeated.

Step 702, the detecting device sends service load information to the first control device.

In step 703, the first control device sends a control command to the second control device according to the plurality of service load information.

In this embodiment, the first control device may store a correspondence list as shown in table 3, which creates correspondence between the second control device identifier and the plurality of computing node identifiers.

TABLE 3 Table 3

Wherein, the computing node with the identifier B2 and the computing node with the identifier B3 shown in table 3 correspond to the second control device with the identifier B1, and the flow control valve controlled by the second control device with the identifier B1 can simultaneously control the flow of the liquid cooling channel with the computing node with the identifier B2 and the flow of the liquid cooling channel with the computing node with the identifier B3. Table 3 shows an example of controlling the flow rate of the liquid cooling channels of two computing nodes by one flow rate regulating valve. Wherein the second control device with the identifier B1 may be a BMC with the identifier B2 or a BMC with the identifier B3. By analogy, the computing node with the identifier of N2 and the computing node with the identifier of N3 are corresponding to the second control device with the identifier of N1, and the flow regulating valve controlled by the second control device with the identifier of N1 can simultaneously control the flow of the liquid cooling channel with the computing node with the identifier of N2 and the flow of the liquid cooling channel with the computing node with the identifier of N3. For example, the detecting means transmits traffic load information including an identification B2 of the computing node 611 and traffic load information including an identification B3 of the computing node 612 to the first controlling means, and the first controlling means obtains a corresponding control instruction including the opening degree of the flow rate regulating valve 632 according to the traffic load information of the computing node 611 and the traffic load information of the computing node 612, and the first controlling means obtains an identification B1 of the second controlling means corresponding to the identification B2 of the computing node 611 and the identification B3 of the computing node 612 according to the correspondence as shown in table 3. The first control device sends the control instruction to the second control device having the identifier B1, so that the second control device having the identifier B1 can adjust the flow rate of the heat dissipation liquid in the liquid cooling channel in communication with the flow rate adjusting valve 632, so as to simultaneously adjust the flow rate of the liquid cooling channel of the computing node 611 and the flow rate of the liquid cooling channel of the computing node 612 to target flow rates respectively. The following describes alternative ways in which the first control device obtains the control instruction according to the information from the plurality of traffic loads:

Mode 1

If the traffic load information is at least one of the message types 1 and 2 shown in step 501 corresponding to fig. 5, first, the first control device obtains the traffic load information from the computing node 611 and the traffic load information from the computing node 612 and reports the traffic load information to the detecting device. The first control device obtains, according to the traffic load information, power consumption corresponding to the computing resources occupied by the computing node 611 executing the traffic and power consumption corresponding to the computing resources occupied by the computing node 612 executing the traffic. The first control device obtains the description of the corresponding power consumption according to the service load information, please refer to the mode 1 corresponding to step 503 in fig. 5, which is not described in detail. Next, the first control device obtains a target flow corresponding to first power consumption, which is the maximum value of power consumption corresponding to the computing resources occupied by the computing node 611 executing the service and power consumption corresponding to the computing resources occupied by the computing node 612 executing the service, as shown in this example. The first control device obtains a description of the target flow corresponding to the first power consumption, please refer to mode 1 corresponding to step 503 in fig. 5, which is not described in detail. For example, if the power consumption corresponding to the service executed by the computing node 611 is greater than the power consumption corresponding to the service executed by the computing node 612, the first control device 631 determines that the power consumption corresponding to the service executed by the computing node 611 is the first power consumption, and the first control device determines the corresponding target traffic according to the first power consumption. Finally, the first control device obtains the opening of the flow regulating valve corresponding to the target flow, and sends a control instruction including the opening to the second control device, and the description of obtaining the control instruction according to the target flow is shown in mode 1 corresponding to step 503 of fig. 5, which is not described in detail.

It can be understood that, when the first power consumption is the maximum value of the power consumption corresponding to the service executed by the computing node 611 and the power consumption corresponding to the service executed by the computing node 612, and the flow of the heat dissipation liquid in the liquid cooling channel is the target flow corresponding to the first power consumption, the cooling capacity generated by the heat dissipation liquid in the two liquid cooling channels connected to the flow adjustment valve 632 can not only meet the heat dissipation requirement of the service executed by the computing node 611, but also meet the heat dissipation requirement of the service executed by the computing node 612.

Mode 2

If the traffic load information is at least one of the message types 3 and 4 shown in step 501 corresponding to fig. 5, first, the first control device obtains the traffic load information from the computing node 611 and the traffic load information from the computing node 612 and reports the traffic load information to the detecting device. The first control means obtains the type of traffic performed by the computing node 611 and the type of traffic performed by the computing node 612 based on the traffic load information. The first control means obtains the power consumption corresponding to the execution of the service type by the computing node 611 and the power consumption corresponding to the execution of the service type by the computing node 612. The first control device obtains the description of the corresponding power consumption according to the service type, please refer to the mode 2 corresponding to the step 503 in fig. 5, which is not described in detail. Next, the first control device obtains a target flow rate corresponding to second power consumption, which is the maximum value of power consumption corresponding to the type of service performed by the computing node 611 and power consumption corresponding to the type of service performed by the computing node 612, as shown in this example. The first control device obtains a description of the target flow corresponding to the second power consumption, please refer to a mode 2 corresponding to step 503 in fig. 5, which is not described in detail. Finally, the first control device obtains the opening of the flow regulating valve corresponding to the target flow, and sends a control instruction including the opening to the second control device, and the description of obtaining the control instruction according to the target flow is shown in mode 2 corresponding to step 503 of fig. 5, which is not described in detail.

Optionally, in this embodiment, taking an example that the detecting device reports the traffic load information of each computing node to the first control device, in other examples, the detecting device may obtain the target traffic load information from the plurality of traffic load information obtained by the computing nodes, and report the target traffic load information to the first control device. For example, in combination with the example shown in the above-mentioned mode 1, the detection device obtains the first power consumption, and the description of obtaining the first power consumption is shown in the above-mentioned mode 1, which is not repeated in detail. The detection device sends the target traffic load information including the first power consumption to the first control device, and then the first control device can directly obtain the corresponding target traffic according to the first power consumption included in the target traffic load information. For another example, in combination with the example shown in the above manner 2, the detection device obtains the second power consumption, and the description of obtaining the second power consumption is shown in the above manner 2, which is not repeated in detail. The detection device sends the target traffic load information including the second power consumption to the first control device, and then the first control device can directly obtain the corresponding target traffic according to the second power consumption included in the target traffic load information.

Step 704, the second control device adjusts the flow rate adjustment valve to an opening included in the control signaling.

Taking the liquid cooling channels of the computing nodes 611 and 612 as an example, the opening included in the control signaling sent by the second control device to the second control device connected to the flow rate adjustment valve 632 can adjust both the flow rate of the liquid cooling channel of the computing node 611 and the flow rate of the liquid cooling channel of the computing node 612 to the target flow rate. Under the condition that the flow of the liquid cooling channel of the computing node 611 and the flow of the liquid cooling channel of the computing node 612 are both target flows, the heat dissipation requirement of the computing node 611 for executing the service can be met, and the heat dissipation requirement of the computing node 612 for executing the service can also be met. In the execution of step 704 shown in this embodiment, please refer to step 504 corresponding to fig. 5, which is not described in detail.

By adopting the method shown in the embodiment, a plurality of liquid cooling channels communicated with the same flow regulating valve are in a parallel connection state. The heat dissipation requirements of the execution services of different computing nodes are different, the first control device can obtain corresponding service load information according to the condition that each computing node executes the services, the first control device can obtain corresponding control instructions according to the service load information, and the second control device controls the flow regulating valve to the opening included in the control signaling, so that the flow regulating valve can simultaneously regulate the flow of heat dissipation liquid in a plurality of liquid cooling channels connected in parallel, and the heat dissipation requirements of each computing node are met. And the flow regulating valve is used for simultaneously regulating the flow of the heat dissipation liquid in the plurality of liquid cooling channels which are connected in parallel, so that the efficiency of regulating the flow of the heat dissipation liquid in the liquid cooling channels is improved, and the heat dissipation efficiency of the computing nodes is further effectively improved. And the first control device obtains the control instruction according to the service load information related to the service executed by the computing node, so that the control instruction indicates the target flow regulated by the flow regulating valve and matches the heat dissipation requirement of the service executed by the computing node. The flow rates of the plurality of liquid cooling channels in parallel connection are the same, so that balanced heat dissipation of a plurality of computing nodes is realized, the pressure and power consumption expenditure of a water pump deployed on a primary side water supply are effectively reduced, and the purposes of energy conservation and emission reduction are realized.

Example III

While the above embodiment illustrates how the first control device adjusts the flow of the heat dissipation liquid for dissipating heat to the computing node through the flow adjustment valve, the embodiment shown in fig. 8 also illustrates that the first control device can adjust the power consumption of the computing node for executing the service, so that the heat generated in the process of executing the service by different computing nodes is the same or approximately the same, and the waste of the cold generated by the liquid cooling channel is reduced. Fig. 8 is a flow chart of steps of a heat dissipation method according to a third embodiment of the present application. The computing system to which the method of this embodiment is applied is shown in fig. 6, which is not described in detail.

Step 801, the detecting device obtains service load information.

Step 802, the detecting device sends service load information to the first control device.

Step 803, the first control device sends a control instruction to the second control device according to the service load information.

For the description of the execution process of steps 801 to 803 in this embodiment, please refer to steps 501 to 503 in fig. 5, and also refer to steps 701 to 703 corresponding to fig. 7, and the detailed execution process is not repeated.

Step 804, the first control device sends an adjustment instruction to the detection device.

The execution timing between step 803 and step 804 is not limited in this embodiment.

The detection device shown in this embodiment is configured to control, according to an adjustment instruction, power consumption corresponding to a target computing node executing service. The target computing node is one of the computing loads, such as computing node 611 and computing node 612, as shown in fig. 6, then the target computing node may be one of computing node 611 and computing node 612. And the first control device sends an adjustment instruction to the detection device, and the adjustment instruction is used for instructing the detection device to control the target computing node to execute the power consumption corresponding to the service according to the adjustment instruction. The number of target computing nodes is not limited in this embodiment. The first control device in this embodiment sends the adjustment instruction to the detection device, so that one computing load meets a target condition, where after the target computing node adjusts the power consumption corresponding to the execution service according to the adjustment instruction, the absolute value of the difference between the power consumption corresponding to the execution service of any two computing nodes in the computing load is smaller than or equal to a preset value. For example, as shown in fig. 6, the computing loads are the computing node 611 and the computing node 612, and if the first control device determines that the power consumption P1 corresponding to the execution of the service by the target computing node 611 is greater than the power consumption P2 corresponding to the execution of the service by the computing node 612, the first control device determines that the target computing node 611 is the computing node that needs to perform power consumption adjustment. The executing service of the computing node shown in this embodiment may be an executed service or a service to be executed, and the description of the executed service and the service to be executed of the computing node is referred to in fig. 5, which is not repeated.

In the case where the first control device determines that the target computing node 611 is a computing node that needs to perform power consumption adjustment, the first control device sends an adjustment instruction to the detection device, and the detection device controls the power consumption corresponding to the execution of the service by the target computing node 611 according to the adjustment instruction, so that the absolute value of the difference between the power consumption of the execution of the service by the target computing node 611 and the power consumption of the execution of the service by the computing node 612 is smaller than or equal to a preset value, and the size of the preset value is not limited in this embodiment, and in the case where the absolute value of the difference between the power consumption of the execution of the service by the target computing node 611 and the power consumption of the execution of the service by the computing node 612 is smaller than or equal to the preset value, the power consumption of the execution of the service by the target computing node 611 and the power consumption of the service by the computing node 612 are equal to or nearly equal.

In the case where the difference between the power consumption of the target computing node 611 for executing the service and the power consumption of the computing node 612 for executing the service is relatively large, if the first control device does not adjust the power consumption of the target computing node 611 for executing the service, the control instruction sent by the first control device indicates the flow rate regulated by the flow rate regulating valve, the heat generated by the power consumption of the target computing node 611 for executing the service needs to be satisfied, that is, the control instruction indicates the cold generated by the target flow rate regulated by the flow rate regulating valve to satisfy the heat dissipation requirement of the target computing node 611, and the target computing node 611 and the computing node 612 are in the parallel connection state, but the power consumption of the computing node 612 for executing the service does not need the large target flow rate as indicated by the control instruction, which may cause the waste of the cold generated by the target flow rate indicated by the control instruction in the process of dissipating the heat of the computing node 612.

In the method shown in this embodiment, when the first control device determines that the power consumption of the target computing node 611 for executing the service is greater than the power consumption of the computing node 612 for executing the service, the second control device can control the power consumption of the target computing node 611 for executing the service according to the adjustment instruction from the first control device, so that the power consumption of the target computing node 611 for executing the service is equal to or nearly equal to the power consumption of the computing node 612 for executing the service, and then the heat generated by the target computing node 611 for executing the service and the heat generated by the computing node 612 for executing the service are in an equilibrium state, and the control instruction generated by the first control device can simultaneously meet the heat dissipation requirement of the target computing node 611 and the heat dissipation requirement of the computing node 612, and reduce the waste of the cold energy generated by the heat dissipation liquid in the liquid cooling channel.

The following describes an optional manner in which the second control device controls the target computing node to adjust the power consumption corresponding to the execution service according to the adjustment instruction from the first control device:

mode 1

The adjustment instruction generated by the first control device is used for instructing the target computing node to execute a power saving operation to reduce power consumption corresponding to execution service, wherein the power saving operation can instruct the target computing node to reduce the working frequency of a processor and/or instruct the target computing node to shut down a processor core (core) and the like. The specific operation type of the energy-saving operation is not limited in this embodiment, so long as the power consumption of the target computing node for executing the service can be reduced after the target computing node executes the energy-saving operation. And the second control device controls the target computing node to execute corresponding energy-saving operation according to the adjustment instruction.

Mode 2

The adjustment instruction generated by the first control device is used for indicating that at least part of the services executed by the target computing node are migrated to another computing node. The at least part of the services executed by the target computing node may be the services executed by the target computing node and/or the services to be executed, which are not limited in this embodiment. The liquid cooling channel of the other computing node and the liquid cooling channel of the target computing node can be controlled by the same flow regulating valve. The liquid cooling channel of the other computing node and the liquid cooling channel of the target computing node can be controlled by different flow regulating valves, and the liquid cooling channel is not limited in this embodiment. And the second control device controls at least part of the business executed by the target computing node to migrate to the other computing node according to the adjustment instruction.

As shown in the above-mentioned mode 2, taking an example that the power consumption of the target computing node 611 for executing the service needs to be reduced, in other examples, the power consumption of the target computing node 611 for executing the service may also be increased, so as to ensure that the heat generated by the target computing node 611 for executing the service and the heat generated by the computing node 612 for executing the service are in an equilibrium state. The adjustment instructions generated by the first control means can then also instruct the target computing node 611 to execute at least part of the traffic from another computing node such that the power consumption of the target computing node 611 to execute the traffic is equal or approximately equal to the power consumption of the computing node 612 to execute the traffic.

Step 805, the second control device adjusts the flow rate adjustment valve to an opening degree included in the control signal.

In the execution of step 805 shown in this embodiment, please refer to step 504 corresponding to fig. 5, and detailed description is omitted.

By adopting the method shown in the embodiment, the first control device can perform balanced heat dissipation on the plurality of computing nodes, so that the waste of cold energy generated by heat dissipation liquid of the liquid cooling channel is reduced.

Example IV

In the above embodiment, taking the case that the first control device executes the service according to the computing node as an example, in this embodiment, the first control device may also generate the control command according to the heat dissipation state of the liquid cooling channel, where fig. 9 is a step flowchart of the heat dissipation method provided in the fourth embodiment of the present application. The computing system to which the method of the present embodiment is applied may be a computing system corresponding to the first embodiment or a computing system corresponding to the second embodiment, which will not be described in detail.

Step 901, the detecting device obtains the service load information.

Step 902, the detecting device sends service load information to the first control device.

For the description of the execution process of step 901 to step 902 in this embodiment, please refer to step 501 to step 502 corresponding to fig. 5, and the detailed execution process is not described in detail.

In step 903, the first control device obtains a heat dissipation status message.

The execution timing between step 903 and steps 901 to 902 is not limited in this embodiment. The heat dissipation status message refers to that the heat dissipation status message is related to the heat dissipation liquid flowing through the liquid cooling channel. It can be understood that the first control device can obtain the current cooling condition of the liquid cooling channel according to the heat dissipation status message. For example, the heat dissipation status message may be sent by the flow control valve to the second control device, and the second control device reports the heat dissipation status message to the first control device. The heat radiation state message is used for indicating the current opening degree of the flow regulating valve. For another example, the computing system further includes a flow sensor coupled to the liquid cooling channel for detecting a flow of the cooling liquid through the liquid cooling channel. The heat radiation state message sent by the flow sensor to the second control device is used for indicating the flow of the heat radiation liquid flowing through the liquid cooling channel. For another example, the computing system further includes a pressure sensor in communication with the liquid cooling channel for detecting a pressure of the cooling liquid flowing through the liquid cooling channel, the heat dissipation status message sent by the pressure sensor to the second control device for indicating the pressure of the cooling liquid flowing through the liquid cooling channel. For another example, the computing system further includes a temperature sensor in communication with the liquid cooling channel, the temperature sensor configured to detect a current temperature of the cooling liquid of the flow liquid cooling channel, and the cooling status message sent by the temperature sensor to the second control device is configured to indicate the current temperature of the cooling liquid of the flow liquid cooling channel. As another example, the computing system further includes a temperature sensor for detecting an ambient temperature in an environment in which the liquid cooling channel is located. The heat dissipation status message sent by the temperature sensor to the second control device is used for indicating the environmental temperature in the environment where the liquid cooling channel is located. In the above-mentioned example, each sensor sends a heat dissipation status message to the second control device, and then the second control device reports the heat dissipation status message to the first control device, in other examples, the interface of each sensor may also be directly connected to the first control device, and then the heat dissipation status message of each sensor may be directly reported to the first control device.

Step 904, the first control device sends a control instruction to the second control device according to the service load information and the heat dissipation status message.

The first control device shown in this embodiment can obtain the control instruction according to the service load information and the heat dissipation status message, so that the cooling capacity generated by the heat dissipation liquid of the liquid cooling channel indicated by the control instruction can meet the heat dissipation requirement of the computing node.

Optionally, the first control device shown in this embodiment may input the traffic load information and the heat dissipation status message into an AI model, and obtain the control instruction by using the traffic load information and the heat dissipation status message through the AI model. To obtain the AI model, the first control device creates a data set in advance, where the data set is service load information and heat dissipation status information received by the first control device in a history, and may be preprocessed, for example, descriptive statistics, data visualization, data shaping, or data segmentation, etc. on the data set. The first control device trains the data set through a machine learning algorithm to obtain the AI model. The control instruction is obtained through the AI model, so that the efficiency and accuracy of obtaining the target flow of the liquid cooling channel can be improved.

In step 905, the second control device controls the flow regulating valve to the opening included in the control signaling.

In the execution of step 905 shown in the present embodiment, please refer to step 504 corresponding to fig. 5, and detailed description is omitted.

By adopting the method shown in the embodiment, the first control device obtains the corresponding control instruction according to the service load information and the heat dissipation state information, and the second control device can control the flow regulating valve to the opening included in the control instruction, so as to regulate the flow of the heat dissipation liquid of the liquid cooling channel for dissipating heat of the computing node, thereby being capable of precisely dissipating heat of the computing node according to the relevant condition of the service execution of the computing node and the heat dissipation state information of the liquid cooling channel.

The present application provides an electronic device, referring to fig. 10, where fig. 10 is a diagram illustrating a first structural example of the electronic device provided in the present application. The electronic device comprises a transceiver 1001 and a processor 1002 connected to the transceiver 1001.

The processor 1002 is an arithmetic core and a control core. One or more processor cores may be included in the processor 1002. The processor 1002 may be a very large scale integrated circuit. An operating system and other software programs are installed on the processor 1002. It is appreciated that in this embodiment, the core in the processor 1002 may be, for example, a central processing unit (central processing unit, CPU) and may be other specific integrated circuits (application specific integrated circuit, ASIC). The processor 1002 may also be other general purpose processors, digital signal processors (digital signal processing, DSP), application specific integrated circuits (application specific integrated circuit, ASIC), field programmable gate arrays (field programmable gate array, FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, or the like. In practice, the electronic device may also include a plurality of processors.

For example, the electronic device is the first control device shown in any one of the first to fourth embodiments. The transceiver 1001 is configured to receive traffic load information from a computational load of a detection device, the computational load being configured to perform traffic. The processor 1002 is configured to determine a target flow rate of the liquid cooling channel of the calculation load according to the traffic load information, and determine an opening of the flow regulating valve according to the target flow rate. The transceiver 1001 is further configured to send a control instruction, where the control instruction is configured to adjust the flow rate adjustment valve to the opening degree, so as to adjust the flow rate of the liquid cooling channel of the computation load to the target flow rate.

Specifically, the transceiver 1001 is configured to perform steps related to transmission and reception implemented by the first control device in any of the embodiments shown in the first to fourth embodiments. The processor 1002 is configured to execute the steps relating to processing implemented by the first control device in any of the embodiments shown in the first to fourth embodiments.

As another example, the electronic device may be the second control device shown in any one of the first to fourth embodiments, and the transceiver 1001 is configured to perform the steps related to the transceiving implemented by the second control device in any one of the first to fourth embodiments. The processor 1002 is configured to execute the steps relating to the processing implemented by the second control device in any of the embodiments shown in the first to fourth embodiments.

As another example, the electronic device may be a detection device as shown in any one of the first to fourth embodiments, and the transceiver 1001 is configured to perform the steps related to the transceiving implemented by the detection device in any one of the first to fourth embodiments. The processor 1002 is configured to execute the steps related to processing implemented by the detection device in any of the embodiments shown in the first to fourth embodiments.

The structure of the electronic device will be described with reference to fig. 11 from the perspective of the functional module, where fig. 11 is a diagram illustrating a second structural example of the electronic device provided in the present application. The electronic device 1100 includes a transmitting module 1101, a processing module 1102, and a receiving module 1103, which are sequentially connected.

For example, the electronic device 1100 is the first control device in any of the embodiments shown in the first to fourth embodiments.

The receiving module 1103 is configured to receive traffic load information of a computation load from the detecting device, where the computation load is used to execute a service;

the processing module 1102 is configured to determine a target flow of the liquid cooling channel of the calculation load according to the service load information, and determine an opening of a flow regulating valve according to the target flow;

The sending module 1101 is configured to send a control instruction, where the control instruction is configured to adjust the flow rate adjusting valve to the opening degree, so as to adjust the flow rate of the liquid cooling channel of the calculation load to the target flow rate.

Specifically, the transmitting module 1101 is configured to perform the steps related to transmission implemented by the first control device in any of the embodiments shown in the first to fourth embodiments. The processing module 1102 is configured to execute the steps related to processing implemented by the first control device in any of the embodiments shown in the first to fourth embodiments. The receiving module 1103 is configured to perform the steps related to reception implemented by the first control device in any of the embodiments shown in the first to fourth embodiments.

As another example, the electronic device 1100 may be the second control device in any of the embodiments shown in the first to fourth embodiments. Specifically, the transmitting module 1101 is configured to perform the steps related to transmission implemented by the second control device in any of the embodiments shown in the first to fourth embodiments. The processing module 1102 is configured to execute the steps related to processing implemented by the second control device in any of the embodiments shown in the first to fourth embodiments. The receiving module 1103 is configured to perform the steps related to reception implemented by the second control device in any of the embodiments shown in the first to fourth embodiments.

As another example, the electronic device 1100 may be the detection device in any of the embodiments shown in the first to fourth embodiments. Specifically, the transmitting module 1101 is configured to perform the steps related to transmission implemented by the detecting device in any of the embodiments shown in the first to fourth embodiments. The processing module 1102 is configured to execute the steps related to processing implemented by the detection device in any of the embodiments shown in the first to fourth embodiments. The receiving module 1103 is configured to perform the steps related to reception implemented by the detecting device in any of the embodiments shown in the first to fourth embodiments.

The present application also provides a computer-readable storage medium. The computer readable storage medium may be any available medium that can be stored by a computing device or a data storage device such as a data center containing one or more available media. The usable medium may be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium, or a semiconductor medium (e.g., solid state disk), etc. The computer-readable storage medium includes instructions that instruct a computer to perform the heat dissipation method shown in any one of the first to fourth embodiments.

The above embodiments are merely for illustrating the technical solution of the present application, and not for limiting the same; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the corresponding technical solutions.

Claims

1. A computing system comprising a first control device, a detection device, a flow regulating valve, and a computational load for executing a service;

the detecting device is used for detecting the business load information of the calculation load and sending the business load information to the first control device;

the first control device is used for determining the target flow of the liquid cooling channel of the calculation load according to the service load information and determining the opening of the flow regulating valve according to the target flow;

the first control device is further used for sending a control instruction, and the control instruction is used for adjusting the flow regulating valve to the opening degree so as to adjust the flow of the liquid cooling channel of the calculation load to the target flow.

2. The computing system of claim 1, wherein the computing load is a computing node, a plurality of computing nodes, or a computing node in a rack.

3. The computing system of claim 1, wherein the computing load belongs to a cabinet and the first control device is a management device of the cabinet.

4. The computing system of claim 1, wherein the first control device is to manage computing loads in a plurality of racks in the computing system.

5. The computing system of any one of claims 1 to 4, wherein the first control device is configured to send the control instruction to a second control device, the control instruction including the opening degree, the second control device being configured to control the flow rate adjustment valve to the opening degree according to the control instruction.

6. The computing system of any one of claim 1 to 5,

the first control device is used for determining the target flow according to an artificial intelligence AI model and the business load information.

7. The computing system of claim 6, wherein the first control device is configured to obtain the traffic load information and flow information of the liquid cooling channel of the computing load reported by the second control device, and to train the AI model according to the traffic load information and the flow information.

8. The computing system of any one of claims 1 to 7, wherein the first control means stores therein a relationship between a flow rate of the liquid cooling passage and an opening degree of the flow rate adjustment valve, and the first control means is configured to determine the opening degree of the flow rate adjustment valve based on the relationship and the target flow rate.

9. The computing system of any of claims 1 to 8, wherein the first control means is operable to determine power consumption from the traffic load information and to determine the target traffic from the power consumption.

10. The computing system of any of claims 1 to 9, wherein the first control device is further configured to send an adjustment instruction to the detection device, and the detection device is configured to control power consumption corresponding to the computing load execution service according to the adjustment instruction.

11. The computing system of any one of claims 1 to 10, wherein the flow regulator valve is controlled to shut off after the second control device detects a failure of the liquid cooling channel of the computing load.

12. The computing system of any of claims 1 to 11, wherein the second control device is a baseboard management controller, BMC, and the detection device is a scheduler.

13. A control apparatus, characterized by comprising:

the receiving module is used for receiving the service load information of the calculation load from the detection device, wherein the calculation load is used for executing the service;

the processing module is used for determining the target flow of the liquid cooling channel of the calculation load according to the service load information and determining the opening of the flow regulating valve according to the target flow;

and the sending module is used for sending a control instruction, and the control instruction is used for adjusting the flow regulating valve to the opening degree so as to regulate the flow of the liquid cooling channel of the calculation load to the target flow.

14. The control device according to claim 13, wherein the transmission module is further configured to transmit the control instruction to another control device, the control instruction including the opening degree, the control instruction being configured to control the other control device to adjust the flow rate adjustment valve to the opening degree.

15. The control device according to claim 13 or 14, wherein the processing module is configured to determine the target traffic based on an artificial intelligence AI model and the traffic load information.

16. The control device according to any of claims 13 to 15, wherein the processing module is configured to determine a power consumption from the traffic load information and to determine the target traffic from the power consumption.

17. A method of dissipating heat, the method comprising:

the first control device receives service load information of a calculation load from the detection device, wherein the calculation load is used for executing service;

the first control device determines the target flow of the liquid cooling channel of the calculation load according to the service load information, and determines the opening of a flow regulating valve according to the target flow;

the first control device sends a control instruction which is used for adjusting the flow regulating valve to the opening degree so as to regulate the flow of the liquid cooling channel of the calculated load to the target flow.

18. The method of claim 17, wherein the first control device sending a control instruction comprises:

the first control device sends the control instruction to the second control device, wherein the control instruction comprises the opening degree, and the second control device is used for controlling the flow regulating valve to the opening degree according to the control instruction.

19. The method of claim 17 or 18, wherein the first control means determining the target flow of the liquid cooling channel of the computational load based on the traffic load information comprises:

The first control device determines the target flow according to an artificial intelligence AI model and the service load information.

20. The method of any one of claims 17 to 19, wherein the first control means determining a target flow of the liquid cooling channel of the computational load from the traffic load information comprises:

the first control device determines power consumption according to the service load information and determines the target flow according to the power consumption.

21. A computer readable storage medium comprising computer program instructions which, when executed by a computer, perform the method of any of claims 17 to 20.