Server power supply system and method of operation thereof

Abstract

The application belongs to the technical field of power distribution, and particularly relates to a power supply system and a working method thereof, which comprises a control module and at least two power supply devices electrically connected with the control module; the power supply devices are electrically connected with each other, and a plurality of computing modules are connected to the power supply devices; the control module is adapted to call corresponding computing modules according to computing power requirements, and the control module is adapted to control the corresponding power supply devices to supply power to the called computing modules according to the power supply performance of each power supply device, so that the power supply devices are in the best efficiency interval; the power supply devices remain in the best efficiency interval to ensure the performance of the power supply devices, the stable power supply of the computing modules is ensured, and the performance and service life of the power supply devices are affected by the heat generated by the power supply devices due to the increase of the load.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of power distribution, and particularly relates to a power supply system and a working method thereof. BACKGROUND

[0002] A plurality of computing modules, such as GPUs, are connected to the power supply device of a server. Each power supply device has an optimal efficiency interval when the circuit is designed. When the power output of the power supply device linearly increases in the interval, the power consumption also linearly increases, and the temperature rise is also linear. The power supply device of the server is in the optimal performance interval, and the temperature inside the server is maintained in the appropriate range of the power supply device. If the load of the power supply device increases to cause the power supply device to exceed the optimal efficiency interval, the performance of the power supply device decreases, the temperature increases, and the power supply efficiency of the power supply device is affected, resulting in performance degradation of the computing module.

[0003] Therefore, due to the technical problem of performance degradation caused by the load change of the power supply device exceeding the optimal efficiency interval, a new server power supply system and a working method thereof need to be designed.

[0004] It should be noted that the above information disclosed in the background section of the present application is only used to understand the background of the present application, and therefore, the above description is not considered to constitute prior art information. SUMMARY

[0005] The embodiments of the present disclosure at least provide a server power supply system and a working method thereof.

[0006] In a first aspect, the embodiments of the present disclosure provide a server power supply system, comprising:

[0007] a control module, and at least two power supply devices electrically connected to the control module;

[0008] The power supply devices are electrically connected, and a plurality of computing modules are connected to the power supply devices.

[0009] The control module is adapted to call corresponding computing modules according to the computing power demand, and the control module is adapted to control the corresponding power supply devices to supply power to the called computing modules according to the power supply performance of each power supply device, so that the power supply devices are in the optimal efficiency interval.

[0010] In an optional embodiment, the power supply device comprises a first processor, a second processor, a first power supply module and a second power supply module.

[0011] The first processor is electrically connected to the second processor.

[0012] The first processor is electrically connected to the first power supply module.

[0013] The second processor is electrically connected with the second power module;

[0014] The first power module is electrically connected with the second power module;

[0015] The first power module is connected with a plurality of expansion connectors, and the expansion connectors are adapted to connect the computing modules;

[0016] The second power module is connected with a plurality of expansion connectors, and the expansion connectors are adapted to connect the computing modules;

[0017] The first power module and the second power module are connected with an external power supply to provide required voltages to the connected computing modules.

[0018] In an optional embodiment, the power supply device further comprises a temperature detection module, a power detection module and a power consumption detection module;

[0019] The temperature detection module, the power detection module and the power consumption detection module are electrically connected with the first processor;

[0020] The temperature detection module is adapted to detect the temperature in the power supply device;

[0021] The power detection module is adapted to detect the output power of the first power module and the second power module;

[0022] The power consumption detection module is adapted to detect the power consumption of the first power module and the second power module.

[0023] In an optional embodiment, the power supply devices are connected through cables, i.e. the first power module and the second power module in all power supply devices are connected through cables.

[0024] In an optional embodiment, the first processor is configured to obtain a first performance curve of the first power module and a second performance curve of the second power module in the power supply device according to the temperature in the power supply device and the output power and the power consumption of the first power module and the second power module at each temperature;

[0025] The first processor is configured to obtain the optimal efficiency range of the first power module according to the first performance curve and the optimal efficiency range of the second power module according to the second performance curve in real time.

[0026] In an optional embodiment, the control module is configured to control the power supply of each power supply device according to whether the first power module and the second power module in each power supply device are in the optimal efficiency range.

[0027] In an alternative embodiment, the control module controls each power supply device to supply power, namely

[0028] Each power supply device is sequentially numbered. When the control module controls the corresponding computing module to start working according to the demand, the control module first acquires the first performance curve and the second performance curve corresponding to the first power module and / or the second power module connected to the computing module starting to work, to determine whether the first power module and / or the second power module connected to the computing module starting to work is in the optimal efficiency interval.

[0029] In an alternative embodiment, when the first power module and / or the second power module connected to the computing module starting to work is in the optimal efficiency interval, the control module only controls the corresponding first power module and / or the second power module to supply power.

[0030] In an alternative embodiment, when the first power module connected to the computing module starting to work exceeds the corresponding optimal efficiency interval, if the second power module in the same power supply device works and is in the optimal efficiency interval at this time, the control module controls the second power module to supply power to the first power module, so that the second power module remains in the optimal efficiency interval while the first power module is lowered to the corresponding optimal efficiency interval.

[0031] In an alternative embodiment, when the first power module connected to the computing module starting to work exceeds the corresponding optimal efficiency interval, if the second power module in the same power supply device does not work at this time, the control module controls the second power module to supply power to the first power module, and the first power module is lowered to the corresponding optimal efficiency interval.

[0032] In an alternative embodiment, when the first power module connected to the computing module starting to work exceeds the corresponding optimal efficiency interval, if the second power module in the same power supply device works and is in the optimal efficiency interval at this time, the control module controls the second power module to supply power to the first power module. At this time, if the second power module reaches the upper limit of the optimal efficiency interval but the first power module still exceeds the corresponding optimal efficiency interval, the control module continues to control the first power module and / or the second power module in the optimal efficiency interval or not working in the next numbered power supply device to supply power to the first power module still exceeding the corresponding optimal efficiency interval, so as to lower it to the optimal efficiency interval, and the first power module and the second power module used in each power supply device are in the corresponding optimal efficiency interval.

[0033] In a second aspect, the embodiments of the present disclosure further provide a working method of the server power supply system, comprising:

[0034] The control module calls the corresponding computing module according to the computing power demand, and the control module controls the corresponding power supply device to supply power to the called computing module according to the power supply performance of each power supply device, so that the power supply device is in the best efficiency interval.

[0035] The server power supply system has the advantages that the power supply device is kept in the best efficiency interval, the performance of the power supply device is ensured, the stable power supply of the computing module is ensured, and the performance and service life of the power supply device are affected due to more heat generated by the power supply device caused by load increase.

[0036] Other features and advantages of the present application will be set forth in the following description, and in part will become apparent to those skilled in the art upon examination of the following or can be learned by practice of the application. The objects and other advantages of the application will be realized and attained by the structure particularly pointed out in the written description and claims hereof as well as the appended drawings.

[0037] In order to make the above-mentioned objects, characteristics and advantages of the present application more obvious and easy to understand, the preferred embodiments are described in detail below, and the accompanying drawings are described as follows. BRIEF DESCRIPTION OF DRAWINGS

[0038] In order to more clearly illustrate the specific embodiments of the present application or the technical solutions in the prior art, the following will briefly introduce the drawings needed to be used in the specific embodiments or prior art description. Obviously, the drawings in the following description are some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0039] Figure 1 A server power supply system principle block diagram provided by the embodiment of the present disclosure;

[0040] Figure 2 A power supply device schematic diagram provided by the embodiment of the present disclosure;

[0041] Figure 3 A power supply device principle schematic diagram provided by the embodiment of the present disclosure;

[0042] Figure 4 A power supply judgment schematic diagram provided by the embodiment of the present disclosure. DETAILED DESCRIPTION

[0043] In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions of the present application will be described clearly and completely below with reference to the drawings. Obviously, the described embodiments are some but not all of the embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts should fall into the scope of the present application.

[0044] As used herein, the phrases "in one embodiment", "according to one embodiment", "in some embodiments", and the like, generally mean the fact that a particular feature, structure, or characteristic described after the phrase can be included in at least one embodiment of the present disclosure. Therefore, the particular feature, structure, or characteristic can be included in more than one embodiment of the present disclosure, so that these phrases do not necessarily refer to the same embodiment. As used herein, the terms "example", "exemplary", and the like, are used as an example, instance, or illustration. Any implementation, aspect, or design described herein as "example" or "exemplary" is not necessarily to be construed as preferred or superior to other implementations, aspects, or designs. On the contrary, the use of the terms "example", "exemplary", and the like, is intended to present the concept in a specific manner.

[0045] It is found through research that the prior art has the following disadvantages: a plurality of power supply devices are arranged in the server, and a plurality of computing modules are connected to the power supply devices. As the computing demand increases, the computing modules used for operation increase, and the load of the power supply devices increases. As the load of the power supply devices increases, the output power, power consumption, and heat generated by the power supply devices increase. If the load exceeds the optimal efficiency range of the power supply devices, the performance of the power supply devices decreases, the power supply devices generate a large amount of heat, the internal temperature exceeds the preset range, the performance of the power supply devices is affected, and the service life of the power supply devices is affected. As the power supply devices are used, the power supply devices gradually age, and the heat generated by the power supply devices affects other power supply devices when the power supply devices are stacked together, so that the performance curve of the power supply devices changes. If the performance of the power supply devices is still judged according to the performance curve of the power supply devices when the power supply devices are shipped, errors will occur in the judgment, and the performance of the power supply devices will be affected.

[0046] Some embodiments of the present application will be described in detail below with reference to the accompanying drawings. In the case of no conflict, the embodiments described below and the features in the embodiments can be combined with each other.

[0047] As Figure 1As shown in at least one disclosed embodiment, a server power supply system is provided, comprising: a control module, and at least two power supply devices electrically connected with the control module; the power supply devices are electrically connected, and a plurality of computing modules are connected to the power supply devices; the control module is adapted to call corresponding computing modules according to the computing power demand, and the control module is adapted to control the corresponding power supply device to supply power to the called computing module according to the power supply performance of each power supply device, so that the power supply device is in the best efficiency interval; the power supply device remains in the best efficiency interval to ensure the performance of the power supply device and ensure the stable power supply of the computing module, avoiding the increase of heat generated by the power supply device due to the increase of load, which affects the performance and service life of the power supply device.

[0048] As shown in Figure 2 and Figure 3 In an optional embodiment, the power supply device comprises: a first processor (CPU0), a second processor (CPU1), a first power module and a second power module; the first power module and the second power module can adopt 88096 (PCIe4.0 Switch) chips; the first processor and the second processor can adopt Tengyun S5000C-32 processors; the first processor and the second processor are electrically connected; the first processor and the first power module are electrically connected; the second processor and the second power module are electrically connected; the first power module and the second power module are electrically connected; the first processor and the first power module are connected, the second processor and the second power module are connected, and the first power module and the second power module are connected, which can realize the redundancy of the connection mode; when the power supply device is used, the first processor can control the first power module and the second power module; if the first processor and the first power module are disconnected due to abnormality, the second processor can be controlled at this time to avoid the failure of the power supply device and the failure of the computing module connected to the power supply device; the first power module is connected with a plurality of expansion connectors, and the expansion connectors are adapted to connect the computing modules; the second power module is connected with a plurality of expansion connectors, and the expansion connectors are adapted to connect the computing modules; the first power module and the second power module are connected with an external power supply to provide the required voltage to the connected computing modules; a USB3.0 controller, a gigabit network chip, a SATA3.0 controller chip, a BMC control module and an integrated 8 DDR4 RDIMM slot can be electrically connected to the first processor; an OAM artificial intelligence module of 8 Tianbi100 can also be integrated on the first processor.

[0049] The control module can be electrically connected with the first processor and the second processor; the processor and the power module can be connected through an MCIO port; the first processor and the second processor can be arranged on a CPU board; the power board can provide an external power supply; the power module can provide corresponding voltages for the computing module,

[0050] GPU topology has three mainstream architectures: one Balance, CPU utilization maximization, providing greater uplink bandwidth, limited by the UPI communication bottleneck between CPUs, suitable for HPC, VDI, public cloud, HPC, etc. Scene; two Common, remote GPU communication does not need to cross CPU communication, GPU P2P communication is better, and the throughput is higher, suitable for P2P communication intensive training algorithm models with more CPU tasks, such as Resnet101 / 50; three Cascade, switch cascade provides the strongest GPU peer-to-peer P2P communication, but the CPU to GPU throughput is small, suitable for P2P parameter intensive training algorithm models with less CPU tasks, such as VGG-16.

[0051] The power supply devices in the server are redundantly arranged, and when the server reaches the maximum load, multiple power supply devices are not used, that is, one computing module is not used. The redundant power supply devices can be used to replace the power supply devices that are abnormal, and the power supply devices arranged redundantly can function for other power supply devices, so that the first power module and the second power module in the used power supply devices are in the best efficiency interval.

[0052] In an optional embodiment, the power supply device further comprises a temperature detection module, a power detection module and a power consumption detection module; the temperature detection module, the power detection module and the power consumption detection module are electrically connected with the first processor; the temperature detection module is adapted to detect the temperature in the power supply device; the power detection module is adapted to detect the output power of the first power module and the second power module; the power consumption detection module is adapted to detect the power consumption of the first power module and the second power module; the temperature in the power supply device, the output power and the power consumption of the first power module and the second power module can be detected in real time during the working process of the power supply device, which facilitates real-time updating of the performance curves of the first power module and the second power module, so as to accurately and real-timely determine the best efficiency interval of the first power module and the second power module.

[0053] In an optional embodiment, each power supply device is connected through a cable, that is, the first power module and the second power module in all power supply devices are connected through a cable to meet the functional requirements between the power supply devices.

[0054] In an alternative embodiment, the first processor is configured to obtain a first performance curve of the first power module and a second performance curve of the second power module in the power supply device according to the temperature inside the power supply device and the output power and power consumption of the first power module and the second power module at each temperature; the first processor is configured to obtain the optimal efficiency range of the first power module according to the first performance curve and the optimal efficiency range of the second power module according to the second performance curve in real time; the first performance curve and the second performance curve reflect the relationship between the output power and power consumption of the corresponding power module and the temperature, and will change with the increase of the aging degree of the first power module and the second power module, etc. Therefore, the first performance curve and the second performance curve need to be updated in real time, and the control module needs to update the first performance curve and the second performance curve first when controlling each power supply device to supply power to the computing module, so as to ensure the accuracy of the control of the power supply device.

[0055] In an alternative embodiment, the control module is configured to control each power supply device to supply power according to whether the first power module and the second power module in each power supply device are in the optimal efficiency range.

[0056] In an alternative embodiment, the control module controls each power supply device to supply power, that is, each power supply device is sequentially numbered, when the control module controls the corresponding computing module to start working according to the demand, the control module first obtains the first performance curve and the second performance curve of the first power module and / or the second power module connected to the computing module that starts working, to determine whether the first power module and / or the second power module connected to the computing module that starts working is in the optimal efficiency range; as long as one computing module connected to the first power module is enabled, the first power module will be higher than the minimum value of the corresponding optimal efficiency range, and as long as one computing module connected to the second power module is enabled, the second power module will be higher than the minimum value of the corresponding optimal efficiency range.

[0057] As shown in Figure 4 In an alternative embodiment, when the first power module and / or the second power module connected to the computing module that starts working are both in the optimal efficiency range, the control module only controls the corresponding first power module and / or second power module to supply power; for example, the first power module in the No. 1 power supply device is connected to four computing modules, and the second power module is also connected to four computing modules, at this time the first power module is in the optimal efficiency range and the second power module is in the optimal efficiency range, then it is judged that the first power module and the second power module can both meet the corresponding computing modules at this time, and the control module does not need to call other power supply devices to supply power.

[0058] In an alternative embodiment, when the first power module connected to the computing modules that start to work is out of its corresponding optimal efficiency interval, if the second power module in the same power supply device is working and in its corresponding optimal efficiency interval, the control module controls the second power module to supply power to the first power module, so that the second power module remains in the optimal efficiency interval while the first power module is lowered to the corresponding optimal efficiency interval; for example, the four computing modules connected to the first power module in the No. 1 power supply device are all in use, and two of the four computing modules connected to the second power module have been activated, at this time the first power module is out of its corresponding optimal efficiency interval, the second power module is in its corresponding optimal efficiency interval, the maximum current size corresponding to the optimal efficiency interval of the first power module is 2A, at this time the current size of the first power module is 3A, and the current size of the second power module is 1A, then the second power module provides 1A current to the first power module, so that the first power module provides 2A current while receiving 1A current provided by the second power module, meeting the demand of the first power module at this time, and the first power module is in the optimal efficiency interval at this time, and the second power module is still in the optimal efficiency interval after providing 1A current.

[0059] In an alternative embodiment, when the first power module connected to the computing modules that start to work is out of its corresponding optimal efficiency interval, if the second power module in the same power supply device is not working, the control module controls the second power module to supply power to the first power module, so that the first power module is lowered to the corresponding optimal efficiency interval; for example, the four computing modules connected to the first power module in the No. 1 power supply device are all in use, and the four computing modules connected to the second power module are not in use, at this time the first power module is out of its corresponding optimal efficiency interval, the second power module is not in use, the maximum current size corresponding to the optimal efficiency interval of the first power module is 2A, at this time the current size of the first power module is 3A, then the second power module provides 1A current to the first power module, so that the first power module provides 2A current while receiving 1A current provided by the second power module, meeting the demand of the first power module at this time, and the first power module is in the optimal efficiency interval at this time, and the second power module is in the optimal efficiency interval after providing 1A current.

[0060] In an alternative embodiment, when the first power module connected to the computing module starting to work is out of its corresponding optimal efficiency interval, if the second power module in the same power supply device is working and in the optimal efficiency interval, the control module controls the second power module to supply power to the first power module, and if the second power module reaches the upper limit of the optimal efficiency interval but the first power module is still out of its corresponding optimal efficiency interval, the control module continues to control the first power module and / or the second power module in the optimal efficiency interval or not working in the next numbered power supply device, supplies power to the first power module still out of its corresponding optimal efficiency interval to make it drop into the optimal efficiency interval, and the first power module and the second power module used in each power supply device are in their corresponding optimal efficiency intervals.

[0061] For example, the four computing modules connected to the first power module in the first power supply device are all in use, and the four computing modules connected to the second power module are also all in use, at this time the first power module is out of its corresponding optimal efficiency interval, the second power module is in its corresponding optimal efficiency interval, the maximum current size corresponding to the optimal efficiency interval of the first power module is 2A, at this time the current size of the first power module is 3A, the current size of the second power module is 2A, and the maximum current size corresponding to its optimal efficiency interval is 2A, so providing current to the first power module by the second power module will cause the second power module to be out of its corresponding optimal efficiency interval, therefore the second power module in the first power supply device cannot provide current to the first power module, at this time it is determined whether the first power module in the second power supply device can provide current to the first power module in the first power supply device, if it can, 1A current is provided by the first power module in the second power supply device, if not, it is determined whether the second power module in the second power supply device can provide current, and so on until the first power module and the second power module in each power supply device in use are in the optimal efficiency interval.

[0062] In at least one other disclosed embodiment, a working method of the server power supply system is also provided, comprising: the control module calling corresponding computing modules according to the computing power demand, and the control module controlling the corresponding power supply device to supply power to the called computing modules according to the power supply performance of each power supply device, so that the power supply device is in the optimal efficiency interval.

[0063] In summary, the server power supply system comprises a control module and at least two power supply devices electrically connected with the control module; the power supply devices are electrically connected, and a plurality of computing modules are connected to the power supply devices; the control module is adapted to call corresponding computing modules according to the computing power demand, and the control module is adapted to control the corresponding power supply device to supply power to the called computing module according to the power supply performance of each power supply device, so that the power supply device is in the best efficiency interval; the power supply device keeps the best efficiency interval to ensure the performance of the power supply device, ensure the stable power supply of the computing module, and avoid the increase of the load to cause more heat of the power supply device and affect the performance and service life of the power supply device.

[0064] Based on the above ideal embodiments according to the application, through the above description, relevant personnel can make various changes and modifications without deviating from the technical idea of the application. The technical scope of the application is not limited to the content of the specification, and must be determined according to the scope of the claims.

Claims

1. A server power supply system, characterized by, include: A control module, and at least two power supply devices electrically connected to the control module; The power supply devices are electrically connected to each other, and several computing modules are connected to each power supply device; The control module is adapted to call the corresponding computing module according to the computing power demand, and the control module is adapted to control the corresponding power supply device to supply power to the called computing module according to the power supply performance of each power supply device, so as to keep the power supply device in the optimal efficiency range. The power supply device includes: a first processor, a second processor, a first power module, and a second power module; The first processor and the second processor are electrically connected; The first processor is electrically connected to the first power module; The second processor is electrically connected to the second power module; The first power module is electrically connected to the second power module; The first power module is connected to several expansion connectors, which are adapted to connect to a computing module. The second power module is connected to several expansion connectors, which are adapted to connect to a computing module; The first power module and the second power module are connected to an external power source to provide the required voltage to their respective connected computing modules; The power supply device also includes: a temperature detection module, a power detection module, and a power consumption detection module; The temperature detection module, power detection module, and power consumption detection module are all electrically connected to the first processor; The temperature detection module is suitable for detecting the temperature inside the power supply device; The power detection module is adapted to detect the output power of the first power module and the second power module. The power consumption detection module is suitable for detecting the power consumption of the first power module and the second power module. The first processor is configured to obtain a first performance curve of the first power module and a second performance curve of the second power module in the power supply device based on the internal temperature of the power supply device and the output power and power consumption of the first power module and the second power module at various temperatures. The first processor is configured to obtain the optimal efficiency range of the first power module in real time based on the first performance curve, and to obtain the optimal efficiency range of the second power module based on the second performance curve. The control module is configured to adjust the power supply of each power supply device according to whether the first power module and the second power module in each power supply device are within the optimal efficiency range. The control module regulates the power supply devices to provide power, that is... Each power supply device is sequentially numbered. When the control module controls the corresponding computing module to start working according to the demand, the control module first obtains the first performance curve and the second performance curve corresponding to the first power supply module and / or the second power supply module connected to the computing module that starts working, so as to determine whether the first power supply module and / or the second power supply module connected to the computing module that starts working is in the optimal efficiency range. When the first power module connected to the starting working computing module is out of its corresponding optimal efficiency interval, if the second power module in the same power supply device is working and in the optimal efficiency interval, the control module controls the second power module to supply power to the first power module, so that the second power module remains in the optimal efficiency interval while the first power module is lowered to the corresponding optimal efficiency interval.

2. The server power supply system of claim 1, wherein: The power supply devices are connected through cables, that is, the first power module and the second power module in all power supply devices are connected through cables.

3. The server power supply system of claim 1, wherein: When the first power module and / or the second power module connected to the starting working computing module is in the optimal efficiency interval, the control module only controls the corresponding first power module and / or the second power module to supply power.

4. The server power supply system of claim 1, wherein: When the first power module connected to the starting working computing module is out of its corresponding optimal efficiency interval, if the second power module in the same power supply device is not working, the control module controls the second power module to supply power to the first power module, lowering the first power module to the corresponding optimal efficiency interval.

5. The server power supply system of claim 1, wherein: When the first power module connected to the starting working computing module is out of its corresponding optimal efficiency interval, if the second power module in the same power supply device is working and in the optimal efficiency interval, the control module controls the second power module to supply power to the first power module, and if the second power module reaches the upper limit of the optimal efficiency interval but the first power module is still out of its corresponding optimal efficiency interval, the control module continues to control the first power module and / or the second power module in the optimal efficiency interval or not working in the next numbered power supply device to supply power to the first power module that is still out of its corresponding optimal efficiency interval, so that it is lowered to the optimal efficiency interval, and the first power module and the second power module used in each power supply device are in their corresponding optimal efficiency intervals.

6. A method of operating a server power supply system as claimed in claim 1, characterized in that including: The control module calls the corresponding computing module according to the computing power demand, and the control module controls the corresponding power supply device to supply power to the called computing module according to the power supply performance of each power supply device, so that the power supply device is in the optimal efficiency interval.

Images