Array server configuration method suitable for AIPC and server

Abstract

The invention discloses an array server configuration method suitable for AIPC and a server, and relates to the field of server configuration.The method comprises the steps that calculation demand characteristics of a target application scene are obtained, and the calculation demand characteristics comprise the single-machine calculation power demand, the network interconnection demand and the storage capacity demand; calculating the number of target computing power modules according to the single-machine computing power demand, and calculating the number of target storage modules according to the storage capacity demand; constructing a server topology model according to the calculation demand characteristics, and calculating an estimated flow of data transmission between modules in the server topology model according to the server topology model; according to the number of the target computing power modules, the number of the target storage modules and the estimated flow, a server layout mode comprising the computing power modules, the storage modules and the network switching module is determined in a server topology model, and by implementing the method, the resource allocation balance of the array server can be improved.

Description

Technical Field

[0001] The present application relates to the field of server configuration, and in particular to an array server configuration method and server suitable for AIPC. Background Art

[0002] With the booming development of the artificial intelligence industry, the demand for computing infrastructure is growing. As a server specifically designed for artificial intelligence computing, the server configuration of AIPC is particularly important. Different artificial intelligence application scenarios have different requirements for computing resources, storage resources, and network resources, which puts higher requirements on the configuration of AIPC.

[0003] In the related art, a standardized configuration scheme can be adopted, that is, the computing power module, storage module and network switching module are configured according to the preset server specifications. However, in the actual operation process, the standardized configuration scheme leads to uneven resource allocation and reduces the overall performance of the server. Summary of the invention

[0004] The present application provides an array server configuration method suitable for AIPC, which is used to improve the resource allocation balance of the array server.

[0005] In the first aspect, the present application provides an array server configuration method suitable for AIPC, which is applied to a server, and the method includes: obtaining the computing demand characteristics of the target application scenario, the computing demand characteristics including stand-alone computing power requirements, network interconnection requirements and storage capacity requirements; calculating the target number of computing power modules according to the stand-alone computing power requirements, and calculating the target number of storage modules according to the storage capacity requirements; constructing a server topology model according to the computing demand characteristics, the server topology model including a computing power module, a storage module and a network switching module, and the server topology model is a preset specification server model; calculating the estimated traffic flow of data transmission between each module in the server topology model according to the server topology model; determining the server layout method including the computing power module, the storage module and the network switching module in the server topology model according to the target number of computing power modules, the target number of storage modules and the estimated traffic flow.

[0006] By adopting the above technical solution, we first obtain the computing demand characteristics of the target application scenario, including stand-alone computing power, network interconnection and storage capacity requirements. According to the stand-alone computing power requirements, we determine the target number of computing power modules, and according to the storage capacity requirements, we calculate the target number of storage modules. Then, we build a server topology model with preset specifications based on the computing demand characteristics, and determine the server layout after estimating the traffic between computing modules. This can accurately match the requirements of different application scenarios, make server resource allocation more reasonable, avoid resource waste or shortage, and improve the overall performance of the server.

[0007] In combination with some embodiments of the first aspect, in some embodiments, the server topology model is constructed according to the computing demand characteristics, and the server topology model includes a computing power module, a storage module and a network switching module. The server topology model is a preset specification server model, and the steps specifically include: obtaining the specification parameters of the preset server, and the specification parameters include height, width and depth; establishing a basic topology model according to the specification parameters and preset functional modules; placing the computing power module, the storage module and the network switching module in the basic topology model based on preset arrangement rules to obtain a server topology model, and the server topology model is a preset specification server model.

[0008] By adopting the above technical solution, after obtaining the preset server specification parameters, the basic topology model is established in combination with the preset functional modules, and then the computing power, storage and network switching modules are placed according to the preset arrangement rules to obtain the server topology model. The specification parameters provide a framework for module placement, the functional modules affect the layout planning, and the preset arrangement rules take into account many factors, so that the internal structure of the server is orderly, the layout of each module is reasonable, and the heat dissipation and wiring are optimized, which improves the maintainability and scalability of the server.

[0009] In combination with some embodiments of the first aspect, in some embodiments, the step of determining the server layout method including the computing power module, the storage module and the network switching module in the server topology model according to the target number of computing power modules, the target number of storage modules and the estimated traffic specifically includes: selecting corresponding module arrangement rules in a preset data table according to the target number of computing power modules, the target number of storage modules and the estimated traffic, and the module arrangement rules include computing power module arrangement rules, storage module arrangement rules, target network switching module and preset function arrangement rules; arranging the target computing power module according to the computing power module arrangement rules in the module arrangement rules, and arranging the target storage module according to the storage module arrangement rules; determining the target network switching module according to the estimated traffic, and arranging the target network switching module according to the network switching module arrangement rules; arranging the preset function modules according to the preset function arrangement rules, and finally obtaining the server layout method.

[0010] By adopting the above technical solution, the module arrangement rules are selected in the preset data table according to the target number of computing power modules, the target number of storage modules and the estimated traffic, and each module is arranged according to the rules. The target number of computing power modules determines the calculation scale and affects the arrangement of computing power modules, the target number of storage modules determines the storage layout, and the estimated traffic dominates the location of network switching modules. The reasonable layout of each module reduces data transmission delay and improves data exchange efficiency.

[0011] In combination with some embodiments of the first aspect, in some embodiments, after the step of determining a server layout method including the computing power module, the storage module and the network switching module in a server topology model according to the target number of computing power modules, the target number of storage modules and the estimated traffic, the method further includes: determining a target heat dissipation module, a type of the target heat dissipation module and specifications of the target heat dissipation module according to the target application scenario and the target number of computing power modules; adding the target heat dissipation module to the server layout method based on the type and the specifications, and a distance between the target heat dissipation module and the target computing power module is less than a preset distance threshold.

[0012] By adopting the above technical solution, after determining the server layout, the target heat dissipation module and its type and specifications are determined according to the target application scenario and the target number of computing power modules. The target application scenario determines the workload characteristics of the server. The heating conditions of the computing power module are different in different scenarios, and the target number of computing power modules affects the overall scale of heat generation. Based on this, selecting a suitable heat dissipation module can accurately match the server's heat dissipation requirements. Adding the target heat dissipation module to the layout and ensuring that the distance from the target computing power module is less than the preset distance threshold achieves efficient heat dissipation.

[0013] In combination with some embodiments of the first aspect, in some embodiments, after the step of determining the server layout including the computing power module, the storage module and the network switching module in the server topology model according to the target number of computing power modules, the target number of storage modules and the estimated traffic, the method also includes: obtaining each real-time business demand dynamics under the target application scenario, the real-time business demand dynamics including the number of tasks, task types and resource occupancy of each real-time business; calculating the computing power allocation ratio of each real-time business corresponding to the number of computing power modules according to the real-time business demand dynamics; and allocating computing power resources according to the computing power allocation ratio.

[0014] By adopting the above technical solutions, we can obtain the real-time business demand dynamics in the target application scenario, including the number of tasks, task types and resource occupancy. The number and type of tasks reflect the complexity of the business, and the resource occupancy reflects the degree of demand for server resources by the business. Based on these calculations, the number of computing power modules corresponds to the computing power allocation ratio for each real-time business, so that the computing power allocation is more in line with the actual business needs.

[0015] In combination with some embodiments of the first aspect, in some embodiments, after the step of allocating computing power resources according to the computing power allocation ratio, the method also includes: obtaining the storage requirement of each of the real-time services; allocating storage resources to each of the real-time services according to the storage requirement to obtain the storage resource allocation ratio; allocating storage resources according to the storage resource allocation ratio.

[0016] By adopting the above technical solution, after computing power resources are allocated, the storage demand of each real-time business is obtained, and the storage resource allocation ratio is calculated according to the storage demand, so that the server storage resources can be reasonably divided. Storage resources are allocated according to the ratio, and storage areas are divided for each real-time business through technical means such as logical address mapping and partitioning.

[0017] In combination with some embodiments of the first aspect, in some embodiments, after the step of determining the server layout including the computing power module, the storage module and the network switching module in the server topology model according to the target number of computing power modules, the target number of storage modules and the estimated traffic, the method also includes: obtaining the memory data, storage data and load data of all target hosts; ranking the memory data, the storage data and the load data to obtain a host memory occupancy list, a host storage occupancy list and a host load occupancy list; and displaying the host memory occupancy list, the host storage occupancy list and the host load occupancy list on a preset front-end page.

[0018] By adopting the above technical solution, the host memory occupancy list, host storage occupancy list and host load occupancy list are displayed on the preset front-end page, so that the administrator can intuitively understand the overall resource allocation status of the server.

[0019] In a second aspect, an embodiment of the present application provides a server, comprising: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors call the computer instructions to enable the server to execute the method described in the first aspect and any possible implementation method of the first aspect.

[0020] In a third aspect, an embodiment of the present application provides a computer program product comprising instructions, which, when executed on a server, enables the server to execute the method described in the first aspect and any possible implementation of the first aspect.

[0021] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, comprising instructions, which, when executed on a server, enable the server to execute the method described in the first aspect and any possible implementation of the first aspect.

[0022] It is understandable that the server provided in the second aspect, the computer program product provided in the third aspect, and the computer storage medium provided in the fourth aspect are all used to execute the method provided in the embodiment of the present application. Therefore, the beneficial effects that can be achieved can refer to the beneficial effects in the corresponding method, which will not be repeated here.

[0023] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: 1. This application first obtains the computing demand characteristics of the target application scenario, including stand-alone computing power, network interconnection and storage capacity requirements. Determine the target number of computing power modules based on the stand-alone computing power requirements, calculate the target number of storage modules based on the storage capacity requirements, and then build a preset specification server topology model based on the computing demand characteristics. After estimating the traffic between computing modules, determine the server layout method, which can accurately match the requirements of different application scenarios, make server resource allocation more reasonable, avoid resource waste or shortage, and improve the overall performance of the server.

[0024] 2. After obtaining the preset server specification parameters, this application establishes a basic topology model in combination with the preset functional modules, and then places the computing power, storage and network switching modules according to the preset arrangement rules to obtain the server topology model. The specification parameters provide a framework for module placement, the functional modules affect the layout planning, and the preset arrangement rules take into account many factors, so that the internal structure of the server is orderly, the layout of each module is reasonable, and the heat dissipation and wiring are optimized, which improves the maintainability and scalability of the server.

[0025] 3. This application selects module arrangement rules in the preset data table according to the target number of computing power modules, the target number of storage modules and the estimated traffic, and arranges each module according to the rules. The target number of computing power modules determines the calculation scale and affects the arrangement of computing power modules, the target number of storage modules affects the storage layout, and the estimated traffic dominates the location of network switching modules. The reasonable layout of each module reduces data transmission delay and improves data exchange efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 It is a flow chart of an array server configuration method suitable for AIPC in an embodiment of the present application; Figure 2 is another flow chart of the array server configuration method suitable for AIPC in the embodiment of the present application; Figure 3 It is a schematic diagram of a server model suitable for the array server configuration method of AIPC in the embodiment of the present application.

[0027] Figure 4 It is another flowchart diagram of the array server configuration method suitable for AIPC in the embodiment of the present application; Figure 5 It is a schematic diagram of a physical device structure of a server in an embodiment of the present application. DETAILED DESCRIPTION

[0028] The terms used in the following embodiments of the present application are only for the purpose of describing specific embodiments, and are not intended to be used as limitations to the present application. As used in the specification of the present application, the singular expressions "one", "a kind of", "above", "the" and "this" are intended to also include plural expressions, unless there is a clear indication to the contrary in the context. It should also be understood that the term "and / or" used in the present application refers to any or all possible combinations comprising one or more of the listed items.

[0029] In the following, the terms "first" and "second" are used for descriptive purposes only and are not to be understood as suggesting or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features, and in the description of the embodiments of the present application, unless otherwise specified, "plurality" means two or more.

[0030] It should be noted that AIPC is a device or server specifically designed for AI computing. Compared with ordinary computers, its hardware configuration is more focused on meeting the needs of AI application scenarios. For example, in terms of computing power, it may be equipped with more powerful CPU, GPU, NPU and other computing cores to cope with complex AI algorithm computing tasks.

[0031] For ease of understanding, the following is a description of the process of the method provided by this implementation. Figure 1 , which is a flow chart of an array server configuration method suitable for AIPC in an embodiment of the present application.

[0032] S101. Obtain computing demand characteristics of a target application scenario, where the computing demand characteristics include single-machine computing power requirements, network interconnection requirements, and storage capacity requirements.

[0033] Among them, the target application scenario refers to the specific environment or business field in which the AIPC array server is applied, such as cloud desktop application scenario, cloud game application scenario, small-scale cloud design application scenario, and end-side AI text and image application scenario. Computing demand characteristics refer to the requirements of the target application scenario for server computing power, network connection, and data storage. Single-machine computing power requirements refer to the computing power required by a single computing device (such as a single AIPC) in the target application scenario, such as the size of the CPU, GPU, NPU, etc. computing power required to execute specific algorithms, process images or data, etc., measured in floating-point operations per second (FLOPS), etc. For example, in complex 3D game scenarios, a single device may require a higher GPU computing power to ensure the rendering effect. Network interconnection requirements refer to the requirements for network bandwidth, latency, stability, etc. when data is transmitted and communicated between modules within the server and with external devices in the target application scenario. For example, in large-scale data parallel processing scenarios, a high-bandwidth and low-latency network is required to ensure rapid data exchange. Storage capacity requirements refer to the amount of storage space required to store data, programs, models and other information in the target application scenario. For example, in a cloud storage service scenario, a large amount of storage capacity is required to save files uploaded by users.

[0034] Specifically, when obtaining the computing demand characteristics of the target application scenario, it is first necessary to conduct a detailed analysis of the target application scenario. The computing demand characteristics vary greatly for different application scenarios. For example, in a cloud gaming scenario, it is necessary to consider factors such as the number of simultaneous online game players and the game screen quality requirements to determine the single-machine computing power requirements; determine the network interconnection requirements based on the transmission frequency and real-time requirements of game data; and determine the storage capacity requirements based on the size of the game installation package and the player's game data storage requirements. By comprehensively considering these factors, the computing demand characteristics of the scenario can be accurately obtained, including the single-machine computing power requirements, network interconnection requirements, and storage capacity requirements, so that the server can be reasonably configured later.

[0035] In some embodiments, the computing demand characteristics may be obtained in a variety of ways: Optionally, for known standard application scenarios, their computing demand characteristics can be determined by consulting relevant industry standards, technical documents, or previous similar project experience. For example, for common cloud desktop office scenarios, the computing power, network bandwidth, and storage capacity required for running office software as specified by industry standards, combined with the expected number of concurrent online users, can be used to determine the computing demand characteristics for this scenario.

[0036] Optionally, actual testing and simulation can be performed for new or special application scenarios. Build a small-scale test environment, run typical tasks in the target application scenario, use performance monitoring tools to collect data, including CPU, GPU usage, network traffic, storage read and write speeds, etc., and analyze these data to determine the computing demand characteristics. For example, when developing new AI image recognition application scenarios, by running test image data on a small-scale server cluster, monitoring the computing power consumption and network transmission of each device, accurate computing demand characteristics can be obtained. It is understandable that computing demand characteristics can also be obtained by investigating the needs of the target user group, analyzing the configuration of competitors' products, etc., which are not limited here.

[0037] S102. Calculate the target number of computing power modules according to the single-machine computing power requirement, and calculate the target number of storage modules according to the storage capacity requirement.

[0038] The single-machine computing power requirement is the computing power required by a single computing device in the target application scenario obtained in step S101. The target computing power module quantity represents the number of computing power modules (such as APU) required to be configured in the server to meet the single-machine computing power requirement of the target application scenario.

[0039] It should be noted that APU (Accelerated Processing Unit) is a processor that integrates multiple computing cores such as CPU, GPU, NPU, etc., and can provide comprehensive computing capabilities. The storage capacity requirement is also the storage space required for the target application scenario determined in step S101. The target number of storage modules refers to the number of storage modules (such as HDD hard disk modules) that should be configured in the server to meet the storage capacity requirement.

[0040] Specifically, when calculating the target number of computing power modules based on the computing power requirements of a single machine, the performance parameters of each computing power module (such as APU) must first be clarified, including its CPU core frequency, GPU core number and performance, NPU computing power, etc. Then compare and calculate the single-machine computing power requirements of the target application scenario with the performance of a single computing power module. For example, if the single-machine computing power requirement of the target application scenario is 10,000 FLOPS, and the computing power of a single APU is 2,000 FLOPS, then the target number of computing power modules is 5 (10,000 ÷ 2,000 = 5). When calculating the target number of storage modules based on storage capacity requirements, it is necessary to understand the capacity of each storage module (such as HDD). If the storage capacity requirement of the target application scenario is 10 TB, and the capacity of a single HDD storage module is 2 TB, then the target number of storage modules is 5 (10 ÷ 2 = 5). Through such calculations, the number of computing power modules and storage modules required to meet the target application scenario can be preliminarily determined, providing a basis for the reasonable configuration of the server.

[0041] In some embodiments, the target number of computing power modules and storage modules can be calculated in a variety of ways: Optionally, when calculating the target number of computing power modules, a redundant design can be used. Considering the possible performance fluctuations or failures of computing power modules, a certain proportion (such as 20%) of computing power modules can be added. For example, according to the above calculation, 5 computing power modules are originally required. After adding 20% ​​redundancy, 6 computing power modules are configured (5×1.2=6). When calculating the target number of storage modules, the data growth trend can also be considered and a certain amount of storage space can be reserved. If the data volume is expected to increase by 30% in the future, the number of target storage modules will increase accordingly. If the original calculation requires 5 storage modules, 7 storage modules will be configured after the increase (5×1.3 rounded up=7).

[0042] Optionally, the calculation is based on the dynamic changes in the load of the application scenario. For application scenarios with large load fluctuations, such as server load during e-commerce promotions, the average load demand is calculated by real-time monitoring of the peak and trough values ​​of the load in historical data, and the number of additional computing power and storage modules required is determined based on the standard deviation of the load change. For example, the average single-machine computing power requirement is 8000FLOPS and the standard deviation is 2000FLOPS. In order to cope with the load peak, the number of computing power modules that meet the computing power requirement of 10000FLOPS (8000+2000) can be configured. For storage modules, the appropriate target number of storage modules is determined based on the growth rate and fluctuation of the data volume during the promotion period. It is understandable that the target number can also be calculated based on cost-benefit analysis, taking into account factors such as the procurement cost and operating cost of computing power and storage modules, which is not limited here.

[0043] S103. Construct a server topology model according to the computing demand characteristics. The server topology model includes a computing power module, a storage module, and a network switching module. The server topology model is a preset specification server model.

[0044] Among them, computing demand characteristics refer to the requirements of the target application scenario for the server in terms of computing, network, and storage. The server topology model is an abstract structure that describes the connection relationship and layout of each module in the server. The computing power module provides computing power, such as APU (including CPU, GPU, NPU, etc.) can perform various computing tasks. The storage module stores data, such as HDD for saving information. The network switching module is responsible for data exchange communication. The preset specification server model has pre-set physical dimensions and other standards, such as the 5U height chassis has its internal space and other specifications.

[0045] Specifically, when building a server topology model based on computing demand characteristics, determine the type and quantity of computing power modules based on the computing power requirements of a single machine, such as selecting a suitable APU based on demand and configuring it according to the number of calculations. Determine the number and type of storage modules based on storage capacity requirements. Select network switching modules based on network interconnection requirements. Then build a basic topology model based on preset server specification parameters (such as height, etc.) and preset functional modules (such as fans, power supplies), and place each module according to the preset arrangement rules (considering heat dissipation, wiring, etc.) to obtain a server topology model that meets the preset specifications.

[0046] In some embodiments, the server topology model can be constructed in a variety of ways: Optional, based on modular design. First divide the server into functional areas (computing power, storage, and network areas), place corresponding modules in each area according to needs (such as placing a suitable APU in the computing power area), and connect the modules in each area with standard interfaces and cables, such as using a high-speed bus to connect the computing power and network switching modules.

[0047] Optionally, a simulation optimization algorithm is used to build the model. Use simulation software to build a virtual model according to the requirements and preset specification parameters, try different module combination layouts, simulate data transmission and other performance evaluations (such as inter-module delay, computing power utilization), select the optimal layout to determine the model, such as using genetic algorithm optimization. It is understandable that other methods can also be used, which are not limited here.

[0048] S104: Calculate the estimated flow of data transmission between each module in the server topology model according to the server topology model.

[0049] The server topology model is a model constructed in step S103 that includes each module and meets the preset specifications. The estimated flow of data transmission between modules refers to the expected data transmission volume (unit such as B / s, etc.) between modules when the server runs the target scenario, such as the data volume transmitted between modules in cloud games.

[0050] Specifically, when calculating the estimated traffic, first analyze the data flow and business logic of the target scenario, such as the data flow of user operations in the cloud desktop. Combined with factors such as task frequency and data size, for example, calculate the estimated traffic of HD video data transmitted from storage to the computing power module (considering video size and reading frequency). Analyze and calculate each transmission path in this way to obtain the estimated traffic between modules.

[0051] In some embodiments, the estimated flow of data transmission between modules can be calculated in multiple ways: Optional, based on historical data statistical analysis and calculation. Collect historical data of similar scenarios or tests, and perform statistical analysis (mean, maximum, trend, etc.) to predict the estimated traffic of the target scenario, such as predicting the future based on the data transmission volume records of a certain period in the past.

[0052] Optionally, a performance testing tool is used to simulate the calculation. Use professional tools to simulate the target scenario test, simulate business operations and monitor data transmission, and determine the estimated traffic based on the results, such as using a network performance testing tool to simulate multiple clients accessing the cloud desktop service to monitor traffic. It is understandable that the calculation can also be based on theoretical calculation models combined with hardware parameters, which is not limited here.

[0053] S105. Determine a server layout including the computing power module, the storage module, and the network switching module in a server topology model according to the target number of computing power modules, the target number of storage modules, and the estimated traffic.

[0054] Among them, the target number of computing power modules is the number of computing power modules (such as APU) that should be configured based on the computing power requirements of the target application scenario, which determines the computing scale of the server. The target number of storage modules is determined by the storage capacity requirements to meet data storage. The estimated traffic is the expected transmission volume between modules, reflecting the degree of interaction. The server topology model describes the connection and layout of modules. The server layout method, that is, the physical location and connection arrangement of each module in the chassis, affects the performance of the server. For example, in the big data processing scenario, there are many computing power modules, and the layout needs to take into account both heat dissipation and data transmission.

[0055] Specifically, when determining the layout based on these factors, consider the computing power module first. If the number is large and the estimated traffic is large, the layout can be dispersed to shorten the connection lines between adjacent modules for heat dissipation and transmission efficiency. The storage module is selected according to the number and transmission requirements with other modules, and the modules that interact frequently are placed nearby. The network switching module should be easy to connect with other modules efficiently and is often placed in the middle or back. Comprehensive considerations determine the optimal layout. For example, in a cloud gaming server, ensure that the computing power module works efficiently with the storage and network modules.

[0056] In some embodiments, the server layout can be determined in a variety of ways: Optional, based on heat dissipation priority. Analyze the heat distribution of computing modules under different layouts, and use thermal simulation software or tests to obtain temperature data. According to the target number of computing modules, place them in low temperature or well-ventilated areas, place storage modules in a place with stable temperature and not affected by the heat of computing modules, and place network switching modules in a place that does not affect heat dissipation and is easy to connect, and then optimize the connection lines.

[0057] Optional, according to the principle of the shortest data transmission path. Determine the pair of modules with frequent transmission according to the estimated traffic and place them in the nearest position. Multi-computing modules can be layered or grouped. The transmission requirements of modules in the same layer or group are similar and the connection path with related modules is short, while taking into account the physical structure and wiring convenience. Cost and maintainability can also be comprehensively considered, such as using universal interfaces and standardized layouts to facilitate later maintenance, which is not limited here.

[0058] The following is a more detailed description of the process of the method provided by this implementation. Figure 2 , is another flow chart of an array server configuration method suitable for AIPC in an embodiment of the present application.

[0059] S201. Obtain computing demand characteristics of a target application scenario, where the computing demand characteristics include single-machine computing power requirements, network interconnection requirements, and storage capacity requirements.

[0060] It can be understood that this step is similar to step S101 and will not be described in detail here.

[0061] S202. Calculate the target number of computing power modules according to the single-machine computing power requirement, and calculate the target number of storage modules according to the storage capacity requirement.

[0062] It can be understood that this step is similar to step S102 and will not be described in detail here.

[0063] S203: Obtain specification parameters of a preset server, where the specification parameters include height, width, and depth.

[0064] This step is to determine the physical space limitations of the server. The preset server specifications are like the blueprint frame size for building the server, such as 5U height (but not limited to 5U, it can also be 4U, 6U and higher).

[0065] S204: Establish a basic topology model according to the specification parameters and preset functional modules.

[0066] Based on the obtained server specification parameters, the basic topology model is constructed in combination with the preset functional modules. The preset functional modules include fans, power supplies, etc. Considering the 5U height of the chassis space, the positions of functional modules such as fan modules and power supplies should be reasonably arranged to meet the heat dissipation and power supply requirements. According to the height, width and depth of the chassis, the approximate layout area of ​​the functional modules is determined. For example, the placement of the fan module should be conducive to air circulation, and the location of the power supply should be convenient for wiring and power distribution, thereby laying the foundation for the placement of computing power modules, storage modules and network switching modules.

[0067] S205. The computing power module, the storage module and the network switching module are placed in the basic topology model based on preset arrangement rules to obtain a server topology model, which is a preset specification server model.

[0068] According to the preset arrangement rules, computing modules (such as APU), storage modules (such as HDD) and network switching modules are placed in the basic topology model. For example, the computing modules are arranged in two rows in the front, the storage modules and the rear computing modules are mutually exclusive, and the central fan module maximizes the shared heat dissipation solution. The switch module, out-of-band management module, storage module (HDD) and power supply are post-maintenance. Each APU module is connected to the switching module or the middle board (MIB) through a cable to achieve network conversion and interconnection. This arrangement method fully considers factors such as heat dissipation, data transmission and maintenance, and finally obtains a server topology model that meets the preset specifications (such as 5U height, etc.). This model can meet the computing power, storage and network requirements in different application scenarios and realize efficient and stable operation of the server.

[0069] like Figure 3 As shown, Figure 3 It is a schematic diagram of a server model suitable for the array server configuration method of AIPC in the embodiment of the present application.

[0070] exist Figure 3 In the example, there are three server model solutions, A, B, and C. In part (A), the overall layout builds a server topology model based on a 5U (but not limited to 5U) chassis specification. From the front view, rear view, side view, and top view, it can be seen that the chassis space is reasonably divided. The computing power module (APU) is arranged in two rows in a tight and regular manner in the front, which is its notable feature. Compared with other examples, this arrangement achieves a higher computing power module density in a limited space, and the upper and lower layout is conducive to forming a stable cold air channel, ensuring that each APU can obtain good heat dissipation conditions and provide guarantees for computing performance. For example, in a cloud gaming scenario, more APUs can process computing tasks such as game screen rendering in parallel, and good heat dissipation can avoid performance degradation due to overheating. The storage module (HDD) and the back computing power module adopt a mutually exclusive combination. From the rear view, it can be seen that their layout is compact and orderly. This layout optimizes the data interaction path with the back computing power module while ensuring the reasonable distribution of storage capacity. The switch module, out-of-band management module, storage module (HDD) and power supply adopt a rear maintenance method and are located at the rear of the chassis, which is convenient for centralized management and maintenance and reduces interference with the operation of the front-end computing power module. Each APU module is connected to the switch module or the middle board (MIB) through a cable to realize network conversion and interconnection functions. The cable direction shown in the top view ensures the efficiency of data transmission. This connection method provides stable support for data exchange between computing power modules and with the outside in this model. Differences in connection layout from other examples may affect the efficiency and stability of data transmission.

[0071] In part (B), the computing power module (APU) is also arranged in two rows, the storage module (HDD) is adjusted in position, and the switch module, out-of-band management module and other functional modules are still located at the rear of the chassis, using the rear maintenance method. From the rear view, you can see the connection relationship between these modules and other components. Each APU module is connected to the switch module or the middle board (MIB) through a cable to ensure network interconnection. During the operation of the server, the network switching module can efficiently handle the data exchange requirements between modules, for example, in the cloud desktop application scenario, it can achieve rapid transmission and response of user operation data.

[0072] In the (C) part, the arrangement of the computing power module (APU) still maintains a certain regularity, and the layout of the upper and lower rows is conducive to heat dissipation and space utilization. At the same time, the combination of the storage module (HDD) and the computing power module in this solution is relatively obvious. As shown in the rear view, multiple HDD storage modules are concentrated in a specific area, forming a close cooperation relationship with the computing power module. Network-related modules such as switch modules are located at the rear of the chassis and adopt a rear maintenance method. Each APU module is connected to the switch module through a cable. The layout of preset functional modules such as fan modules also cooperates with other modules. The position of the fan module can be seen from the side view, which can provide a good heat dissipation environment for the entire server and ensure that each module works within the normal temperature range. The power module provides a stable power supply for the server, and its layout is also convenient for wiring and management.

[0073] S206: Calculate the estimated flow of data transmission between each module in the server topology model according to the server topology model.

[0074] It can be understood that this step is similar to step S104 and will not be described in detail here.

[0075] S207. Select corresponding module arrangement rules in the preset data table according to the target number of computing power modules, the target number of storage modules and the estimated traffic. The module arrangement rules include computing power module arrangement rules, storage module arrangement rules, target network switching modules and preset function arrangement rules.

[0076] Based on the three key factors of target number of computing power modules, target number of storage modules and estimated traffic, find the appropriate module arrangement rules in the preset data table. The target number of computing power modules determines the scale of computing power. For example, if the number is large, a more dispersed arrangement is adopted to facilitate heat dissipation and efficient data interaction. The target number of storage modules affects the layout of storage capacity. The estimated traffic reflects the frequency of data transmission between modules. If the estimated traffic in the cloud gaming scenario is large, it is necessary to select an arrangement rule that can ensure that the network switching module efficiently connects the modules. This rule covers the specific arrangement of computing power modules, storage modules, network switching modules and preset functional modules (such as fans, power supplies, etc.).

[0077] S208. Arrange the target computing power modules according to the computing power module arrangement rule in the module arrangement rule, and arrange the target storage modules according to the storage module arrangement rule.

[0078] The target computing power module is placed according to the computing power module arrangement rules in the selected module arrangement rules. For example, if the rules require layout according to the number of computing power modules and heat dissipation requirements, a double-row single-card heat dissipation method will be used for more computing power modules to ensure that all computing modules can have direct cold air access to ensure the best heat dissipation conditions. At the same time, the target storage module is placed according to the storage module arrangement rules, considering its data interaction requirements with the computing power module. If the interaction is frequent, it will be placed nearby to reduce data transmission delays and improve overall performance.

[0079] S209: Determine a target network switching module according to the estimated traffic, and arrange the target network switching module according to the network switching module arrangement rule.

[0080] Determine the appropriate target network switching module based on the estimated traffic. When the estimated traffic is large, you may need to select a high-performance network switching chip or module. The switching module can use a single-chip or dual-chip solution to meet different network configuration requirements, and then place it according to the network switching module arrangement rules. It is usually placed in a position that is convenient for efficient connection with other modules, such as the middle or back, to ensure that data is transmitted quickly and stably between modules to meet the server's requirements for data exchange.

[0081] S210: Arrange the preset functional modules according to the preset functional arrangement rules to finally obtain a server layout.

[0082] Finally, the preset functional modules (such as fans, power supplies, etc.) are laid out according to the preset functional arrangement rules. The layout of the fan module must ensure effective heat dissipation, such as the center fan module to maximize the shared heat dissipation solution. The power module must consider power supply stability and redundancy. The power module can realize full system power supply and redundancy (3+1 or 2+2 redundancy) functions, and its location must be convenient for wiring and management.

[0083] S211. Determine a target heat dissipation module, a type of the target heat dissipation module, and specifications of the target heat dissipation module according to the target application scenario and the target number of computing power modules.

[0084] Specifically, considering the characteristics of the target application scenario, different application scenarios have different load conditions on the server. For example, in a cloud gaming scenario, a large number of players online at the same time will cause the computing power module to run at high load for a long time, generating more heat; while in a small-scale cloud design scenario, the computing tasks may be relatively concentrated but last for a long time. According to the characteristics of these scenarios, combined with the target number of computing power modules, determine the required heat dissipation module. If the number of computing power modules is large and the application scenario generates a lot of heat, a type with high heat dissipation efficiency may be required, such as a high-performance fan or liquid cooling module. Its specifications are determined based on the power and heat density of the computing power module. For example, for a double-row single-card heat dissipation method, select a fan of appropriate size and heat dissipation capacity to ensure that the heat generated by the computing power module can be effectively removed to ensure stable operation of the server.

[0085] S212: Add the target heat dissipation module to the server layout based on the type and the specification, and the distance between the target heat dissipation module and the target computing power module is less than a preset distance threshold.

[0086] After determining the type and specification of the cooling module, add it to the server layout. Make sure that the distance between the target cooling module and the target computing power module is less than the preset distance threshold to achieve the best cooling effect. As shown in the attached figure of the technical briefing, the layout of the central fan module is to better dissipate heat for the front and rear rows of computing power modules. Place the cooling module reasonably so that its wind direction can blow directly to the computing power module, or through reasonable air duct design, let the cold air pass through the cooling module before flowing to the computing power module to improve the cooling efficiency. At the same time, the compatibility of the cooling module with other components (such as power supply and storage module) must also be considered to avoid mutual interference, ensure that the overall layout of the server is reasonable, and that all components work together to maintain the server running within the normal temperature range, thereby improving the reliability and performance of the server.

[0087] The following is a more detailed description of the process of the method provided by this implementation. Figure 4 , is another flow chart of an array server configuration method suitable for AIPC in an embodiment of the present application.

[0088] S401. Determine a server layout including the computing power module, the storage module and the network switching module in a server topology model according to the target number of computing power modules, the target number of storage modules and the estimated traffic.

[0089] It can be understood that this step is similar to step S105 and will not be described in detail here.

[0090] S402: Acquire each real-time business demand dynamics in the target application scenario, where the real-time business demand dynamics include the task quantity, task type, and resource occupancy of each real-time business.

[0091] During the operation of the target application scenario, it is necessary to obtain the demand dynamics of each real-time business in real time. First, all currently running tasks are counted through the task monitoring system built into the server to determine the number of tasks for each real-time business. For example, in a cloud gaming application scenario, there may be multiple people playing different games at the same time. At this time, the monitoring system will record the number of ongoing game tasks. For example, 10 users are playing 5 different games, and the number of tasks for each game is 2.

[0092] Next, analyze the business type of each task. You can judge based on the program or service features called by the task. For example, if some tasks mainly call graphics rendering related programs, they can be judged as 3D game business types; if the task mainly involves data storage and reading operations, it may belong to the data storage business type, etc.

[0093] Then, the resource usage of each real-time business is accurately calculated. This requires the use of server performance monitoring tools to collect and analyze multiple resource usage data such as CPU usage, GPU usage, memory usage, and network bandwidth usage. For example, in a cloud gaming scenario, when running a large 3D game task, it is monitored that the CPU usage reaches 60%, the GPU usage is 80%, the memory usage is 2GB, and the network bandwidth usage is 5Mbps. These data combined accurately reflect the resource usage of the game business. Through such comprehensive monitoring and analysis, the detailed task quantity, task type, and resource usage of each real-time business in the target application scenario can be obtained.

[0094] S403: Dynamically calculate the computing power allocation ratio of the number of computing power modules corresponding to each real-time business according to the real-time business demand.

[0095] Based on the real-time business demand dynamics obtained in the previous step, start calculating the computing power allocation ratio. First, determine the total number of computing power modules in the server, assuming that the server has a total of 100 computing power modules (APUs).

[0096] For each real-time business, computing power is allocated based on the number of tasks, task type, and resource usage. Taking the cloud gaming scenario as an example, if there are two game businesses, Game A has 3 tasks, with an average CPU usage of 30% and a GPU usage of 40% for each task; Game B has 5 tasks, with an average CPU usage of 20% and a GPU usage of 30% for each task.

[0097] First, calculate the total resource usage weight of each game business. Game A's total CPU usage weight is 3×30%=90%, and its total GPU usage weight is 3×40%=120%; Game B's total CPU usage weight is 5×20%=100%, and its total GPU usage weight is 5×30%=150%.

[0098] Then, the computing power allocation ratio is calculated based on the total resource occupancy weight. The total CPU occupancy weight is 90%+100%=190%, the CPU computing power allocation ratio of game A is 90%÷190%≈47.4%, and the CPU computing power allocation ratio of game B is 100%÷190%≈52.6%. Similarly, the total GPU occupancy weight is 120%+150%=270%, the GPU computing power allocation ratio of game A is 120%÷270%≈44.4%, and the GPU computing power allocation ratio of game B is 150%÷270%≈55.6%.

[0099] S404: Allocate computing power resources according to the computing power allocation ratio.

[0100] After obtaining the computing power allocation ratio for each real-time business, the actual allocation of computing power resources begins. The computing power resource allocation of the server is mainly achieved through the scheduling system at the software level.

[0101] Assume that the server has a total of 100 computing power modules (APUs), based on the previously calculated computing power allocation ratios of Game A and Game B (the CPU computing power allocation ratio of Game A is approximately 47.4%, and the GPU computing power allocation ratio is approximately 44.4%; the CPU computing power allocation ratio of Game B is approximately 52.6%, and the GPU computing power allocation ratio is approximately 55.6%).

[0102] For CPU computing power allocation, the number of computing power modules that can be allocated to Game A is approximately 100×47.4%≈47) and the number of computing power modules that can be allocated to Game B is approximately 100×52.6%≈53 (rounded).

[0103] The GPU computing power allocation is also allocated in proportion. The scheduling system will allocate the GPU resources in the corresponding number of computing power modules (APU) to the corresponding game business. For example, the GPU resources in 47 computing power modules are allocated to game A, and the GPU resources in 53 computing power modules are allocated to game B. In the allocation process, the performance characteristics of each computing power module will be optimized and combined to ensure that the allocated computing power can meet the needs of the game business.

[0104] S405: Obtain storage requirements for each real-time service.

[0105] To obtain the storage requirements of each real-time service, we first need to comprehensively monitor the data storage during the service operation. For services involving file storage, the storage management system will track the creation, modification and deletion of files in real time, accurately record the storage space occupied by each file, and cumulatively calculate the storage requirements of the service. By accumulating the storage space of the data involved in each link in the data processing process, we can get the accurate storage requirements. For multimedia services, such as video or audio stream processing services, the storage requirements are determined by using a specific calculation formula based on the encoding format, bit rate, duration and other parameters of the media file. At the same time, the storage management system will also take into account the system's own overhead, such as the storage space occupied by index files, configuration files, etc., and add these additional parts to the service data storage requirements, so as to accurately obtain the storage requirements of each real-time service during operation.

[0106] S406: Allocate storage resources to each real-time service according to the storage demand to obtain a storage resource allocation ratio.

[0107] After determining the storage requirement of each real-time service, start calculating the storage resource allocation ratio. First, determine the total available storage resources of the server, assuming that the total storage capacity of the server is a known value. Compare and calculate the storage requirement of each real-time service with the total storage requirement of all services. For a real-time service, its storage resource allocation ratio is equal to the storage requirement of the service divided by the total storage requirement of all services. By performing such calculations for each real-time service, the corresponding storage resource allocation ratio is obtained.

[0108] S407: Allocate storage resources according to the storage resource allocation ratio.

[0109] After obtaining the storage resource allocation ratio, the server's storage management system will actually allocate storage resources based on these ratios. The storage management system calculates the specific storage capacity that should be allocated to each real-time business based on the ratio. Then, on the physical storage device, the corresponding storage area is divided for each real-time business through technical means such as logical address mapping and partitioning. During the allocation process, the storage management system will optimize the storage layout to ensure the continuity and efficiency of data storage and improve the data reading and writing speed. At the same time, the storage management system will establish a strict access control mechanism to ensure that each real-time business can only access the storage area to which it belongs to prevent data leakage and illegal access. In addition, the storage management system will continuously monitor the use of storage resources. When the business storage requirements change, the allocation ratio and storage area will be adjusted in a timely manner to ensure that the storage resources can always meet the needs of each real-time business and maintain the stable operation of the server storage system.

[0110] S408. Obtain memory data, storage data, and load data of all target hosts.

[0111] By reading key data structures such as memory allocation tables and page tables, the number and size of memory pages allocated to each process can be accurately obtained, and the amount of used memory can be calculated. For storage data, the data acquisition component works closely with the host's storage device driver. It sends specific instructions to obtain the sector allocation information of the storage device, calculates the number of sectors occupied by data, and thus obtains the size of the used storage space. In terms of load data collection, it involves the monitoring of multiple subsystems. It interacts with the host's CPU scheduling module to obtain the length of time each process occupies the CPU in each time slice, and then calculates the CPU usage rate. At the same time, it collaborates with the network protocol stack module to count the number and size of data packets received and sent by the network interface to determine the network bandwidth occupancy.

[0112] S409: rank the memory data, the storage data, and the load data to obtain a host memory occupancy list, a host storage occupancy list, and a host load occupancy list.

[0113] For memory data, the data processing unit first extracts the used memory data of each target host. Then, an efficient sorting algorithm, such as a quick sorting algorithm, is used to sort these data in order from large to small. During the sorting process, the used memory value of each host is strictly compared to ensure the accuracy of the sorting. Finally, a host memory occupancy list is formed, in which the host with the largest memory usage is ranked first in the list, and then descending. For storage data, the used storage space is also used as the key indicator for sorting. The data processing unit traverses the storage data of all hosts, compares their used storage space one by one, and uses a stable sorting algorithm, such as merge sort, to ensure that the order of hosts with the same storage usage remains consistent, and finally obtains a host storage occupancy list. For load data, the data processing unit calculates the load value of each host based on a pre-set load evaluation model, comprehensive CPU usage, network bandwidth occupancy and other factors. After that, a suitable sorting algorithm, such as heap sort, is used to sort these load values ​​from large to small to generate a host load occupancy list.

[0114] S410: Display the host memory occupancy list, the host storage occupancy list, and the host load occupancy list on a preset front-end page.

[0115] A special front-end display service program is deployed on the server side, and its main function is to present data on the preset front-end page. The program first reads the sorted host memory usage list, host storage usage list and host load usage list from the data storage area. For the host memory usage list, the display service program fills its data into the special memory usage display area on the front-end page, and presents it in a clear table format. Each row corresponds to a host, and the columns respectively display key information such as the host name and the amount of memory used. In addition, different display formats are set according to different ranges of memory usage. For example, for hosts with memory usage exceeding 80%, it is displayed in bold red font. For the host storage usage list, the data is accurately displayed in the storage usage display area of ​​the front-end page, and the host-related information is also displayed in a table. In addition, a click event is set for each host entry. When the administrator clicks it, a detailed storage usage pop-up window can be popped up, including detailed information such as the partition structure of the storage device and the proportion of space used in each partition.

[0116] In some embodiments, data about the hyper-converged cluster can also be collected in the resource pool data collector, including data such as CPU, memory, storage, and load. The host data collector collects data about the host, including data such as CPU, memory, storage, and load. The data collector stores this data and time series data in the database. At the same time, the data collector sends the data to the front end for display, which is divided into overall resource occupancy, host resource occupancy ranking, and operating status. The overall resource occupancy shows the overall occupancy of the hyper-converged cluster, including CPU data display, memory data display, storage data display, and load data display. The host resource occupancy ranking shows the host resource occupancy ranking, including CPU ranking display, memory ranking display, storage ranking display, and load ranking display. The operating status shows the health of the hyper-converged cluster, including host status, container engine status, container status, and virtual machine status. The status categories are divided into normal, abnormal, maintenance, and shutdown.

[0117] In some embodiments, management signals from different computing power modules or storage modules can be aggregated. The management information includes alarm information such as the operating status, health monitoring, temperature, voltage, etc. of each module. The data stream signal of each module is used for local or remote online upgrades of each module to meet operation and maintenance or changes in customer business applications. The video signal is used to support the remote IPKVM function, realize remote desktop transmission and desktop switching between different modules. (Fan) control signals can realize functions such as on-demand fan speed regulation and fan health monitoring. For scenarios equipped with storage modules, the hard disk control signal is used for hard disk status check, basic information query, hard disk ambient temperature check and other functions.

[0118] In some embodiments, the server configuration includes the layout and settings of the computing power module, storage module, and network switching module, etc. The workflow of the unattended installation system can be realized by using PXE to automatically complete the installation of the server operating system. After the server hardware layout is completed, a special step is added to explain the activation of the PXE unattended installation system. For example, after determining the server layout mode (such as after step S210), add the step of starting the PXE unattended installation system and performing the automatic installation of the server operating system. When the server obtains the IP address from the DHCP server, explain how the allocation of the IP address matches the network environment where the server is located and its own network switching module configuration to ensure the correctness of the network connection. When obtaining the network boot program and related configuration files through the TFTP server, explain the relationship between the acquisition path of these files and the server storage module, for example, the boot program and configuration files may be stored in a specific storage area, which is related to the storage module layout of the server. When installing the Linux system according to the automatic response file, mention how the initial configuration after the system is installed cooperates with the resource configuration of the server's computing power module, memory, etc.

[0119] The following describes the server in the embodiment of the present invention from the perspective of hardware processing. Figure 5 , is a schematic diagram of a physical device structure of a server in an embodiment of the present application.

[0120] It should be noted that Figure 5 The structure of the server shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0121] like Figure 5 As shown, the server includes a central processing unit (CPU) 501, which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 502 or the program loaded from the storage part 508 to the random access memory (RAM) 503, such as executing the method described in the above embodiment. Various programs and data required for system operation are also stored in RAM503. CPU501, ROM502 and RAM503 are connected to each other through bus 504. Input / output (I / O) interface 505 is also connected to bus 504.

[0122] The following components are connected to the I / O interface 505: an input section 506 including an audio input device, a button switch, etc.; an output section 507 including a liquid crystal display (LCD) and an audio output device, an indicator light, etc.; a storage section 508 including a hard disk, etc.; and a communication section 509 including a network interface card such as a LAN (Local Area Network) card, a modem, etc. The communication section 509 performs communication processing via a network such as the Internet. A drive 510 is also connected to the I / O interface 505 as needed. A removable medium 511, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 510 as needed so that a computer program read therefrom is installed into the storage section 508 as needed.

[0123] In particular, according to an embodiment of the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present invention includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a computer program for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication part 509, and / or installed from a removable medium 511. When the computer program is executed by a central processing unit (CPU) 501, various functions defined in the present invention are performed.

[0124] It should be noted that specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a flash memory, an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in combination with an instruction execution system, apparatus, or device.

[0125] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present invention. Each box in the flowchart or block diagram may represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions marked in the box may also occur in an order different from that marked in the accompanying drawings.

[0126] Specifically, the server of this embodiment includes a processor and a memory. The memory stores a computer program. When the computer program is executed by the processor, the array server configuration method suitable for AIPC provided in the above embodiment is implemented.

[0127] As another aspect, the present invention further provides a computer-readable storage medium, which may be included in the server described in the above embodiment; or may exist independently without being assembled into the server. The above storage medium carries one or more computer programs, and when the above one or more computer programs are executed by a processor of the server, the server implements the array server configuration method suitable for AIPC provided in the above embodiment.

[0128] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.

[0129] As used in the above embodiments, the term "when..." may be interpreted to mean "if..." or "after..." or "in response to determining..." or "in response to detecting...", depending on the context. Similarly, the phrases "upon determining..." or "if (the stated condition or event) is detected" may be interpreted to mean "if determining..." or "in response to determining..." or "upon detecting (the stated condition or event)" or "in response to detecting (the stated condition or event)", depending on the context.

[0130] Those skilled in the art can understand that to implement all or part of the processes in the above-mentioned embodiments, the processes can be completed by computer programs to instruct related hardware, and the programs can be stored in computer-readable storage media. When the programs are executed, they can include the processes of the above-mentioned method embodiments. The aforementioned storage media include: ROM or random access memory RAM, magnetic disk or optical disk and other media that can store program codes.

Claims

1. A method for configuring an array server suitable for AIPC, characterized in that: Applied to a server, the method comprises: Obtaining computing demand characteristics of the target application scenario, wherein the computing demand characteristics include single-machine computing power requirements, network interconnection requirements, and storage capacity requirements; Calculate the target number of computing power modules according to the single-machine computing power requirement, and calculate the target number of storage modules according to the storage capacity requirement; Building a server topology model according to the computing demand characteristics, wherein the server topology model includes a computing power module, a storage module, and a network switching module, and the server topology model is a preset specification server model; Calculate the estimated flow of data transmission between each module in the server topology model according to the server topology model; A server layout including the computing power module, the storage module and the network switching module is determined in a server topology model according to the target number of computing power modules, the target number of storage modules and the estimated traffic.

2. The method according to claim 1, characterized in that The step of building a server topology model according to the computing demand characteristics, wherein the server topology model includes a computing power module, a storage module, and a network switching module, and the server topology model is a server model with preset specifications, specifically includes: Obtaining specification parameters of a preset server, wherein the specification parameters include height, width and depth; Establishing a basic topology model according to the specification parameters and preset functional modules; The computing power module, the storage module and the network switching module are placed in the basic topology model based on preset arrangement rules to obtain a server topology model, which is a preset specification server model.

3. The method according to claim 2, characterized in that The step of determining a server layout mode including the computing power module, the storage module and the network switching module in a server topology model according to the target number of computing power modules, the target number of storage modules and the estimated traffic specifically includes: Selecting a corresponding module arrangement rule in a preset data table according to the target number of computing power modules, the target number of storage modules and the estimated traffic, wherein the module arrangement rule includes a computing power module arrangement rule, a storage module arrangement rule, a target network switching module and a preset function arrangement rule; Arrange the target computing power modules according to the computing power module arrangement rule in the module arrangement rule, and arrange the target storage modules according to the storage module arrangement rule; Determine a target network switching module according to the estimated traffic, and arrange the target network switching module according to the network switching module arrangement rule; The preset functional modules are arranged according to the preset functional arrangement rules, and finally a server layout is obtained.

4. The method according to claim 2, characterized in that: After the step of determining a server layout including the computing power module, the storage module and the network switching module in a server topology model according to the target number of computing power modules, the target number of storage modules and the estimated traffic, the method further includes: Determine a target heat dissipation module, a type of the target heat dissipation module, and a specification of the target heat dissipation module according to the target application scenario and the target number of computing power modules; The target heat dissipation module is added to the server layout based on the type and the specification, and the distance between the target heat dissipation module and the target computing power module is less than a preset distance threshold.

5. The method according to claim 2, characterized in that: After the step of determining a server layout including the computing power module, the storage module and the network switching module in a server topology model according to the target number of computing power modules, the target number of storage modules and the estimated traffic, the method further includes: Obtaining each real-time business demand dynamics under the target application scenario, wherein the real-time business demand dynamics include the number of tasks, task type, and resource occupancy of each real-time business; Dynamically calculate the computing power allocation ratio of the number of computing power modules corresponding to each real-time business according to the real-time business needs; The computing power resources are allocated according to the computing power allocation ratio.

6. The method according to claim 5, characterized in that After the step of allocating computing power resources according to the computing power allocation ratio, the method further includes: Obtaining storage requirements for each of the real-time services; Allocate storage resources to each of the real-time services according to the storage demand to obtain a storage resource allocation ratio; The storage resources are allocated according to the storage resource allocation ratio.

7. The method according to claim 1, characterized in that After the step of determining a server layout including the computing power module, the storage module and the network switching module in a server topology model according to the target number of computing power modules, the target number of storage modules and the estimated traffic, the method further includes: Get the memory data, storage data and load data of all target hosts; Ranking the memory data, the storage data and the load data to obtain a host memory occupancy list, a host storage occupancy list and a host load occupancy list; The host memory occupancy list, the host storage occupancy list and the host load occupancy list are displayed on a preset front-end page.

8. A server, characterized in that: The server includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors call the computer instructions to enable the server to execute the method described in any one of claims 1-7.

9. A computer-readable storage medium comprising instructions, characterized in that: When the instructions are executed on a server, the server is caused to execute the method according to any one of claims 1 to 7.

10. A computer program product, characterized in that When the computer program product is run on a server, the server is caused to execute the method according to any one of claims 1 to 7.

Images