Consumer graphics card multiplexing scheduling method, system and equipment and medium

Abstract

The invention provides a consumption-level graphics card multiplexing scheduling method, system and device and a medium, and the method comprises the steps: obtaining a resource application submitted by a specified service platform, determining an access point of a network used by the specified service platform according to the resource application, and positioning a target edge cloud to which the access point belongs according to the access point; under the condition that the target computing power resource meeting the business demand condition exists in the target edge cloud, locking the target computing power resource and distributing the target computing power resource to a specified business platform for use; wherein the condition of meeting the business requirement comprises that idle bare metal resources exist in a resource pool of the target edge cloud and specified diskless services required by a specified business platform exist in the resource pool. Management and maintenance of computing equipment are simplified through a diskless technology, and meanwhile, through an intelligent computing scheduling engine, the computing power scheduling of the graphics card is realized, dynamic combination of computing power and diskless services is realized, capability is quickly provided to the outside, human input is reduced, rapid reuse of graphics card resources is realized, and resource waste is avoided, so that the hardware cost is remarkably reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of resource scheduling, and in particular to a method, system, device and medium for scheduling the reuse of consumer-grade graphics cards based on diskless technology. Background Art

[0002] With the rapid development of artificial intelligence, big data, virtual reality and high-performance computing, the demand for graphics card (GPU) computing power in the consumer and professional markets is increasing day by day. The traditional way of using graphics cards is usually one computer corresponds to one or more graphics cards, which has low resource utilization efficiency, especially in a multi-user and multi-tasking environment. This method cannot give full play to the computing power of graphics cards. Summary of the invention

[0003] The embodiment of the present invention provides a method, system, device and medium for scheduling the reuse of consumer-grade graphics cards to solve the problems existing in related technologies. The technical solution is as follows:

[0004] In a first aspect, an embodiment of the present invention provides a method for scheduling the reuse of a consumer-grade graphics card, comprising:

[0005] Obtain resource applications submitted by designated service platforms, determine access points of the network used by designated service platforms based on the resource applications, and locate the target edge cloud to which it belongs based on the access points;

[0006] When there are target computing resources that meet the business requirements in the target edge cloud, the target computing resources are locked and allocated to the designated business platform for use; wherein, meeting the business requirements includes that there are idle bare metal resources in the resource pool of the target edge cloud and that there are designated diskless services required by the designated business platform in the resource pool.

[0007] In one implementation, locating a target edge cloud to which an access point belongs according to the access point includes:

[0008] Determine the unique cloud access identifier of the network used by the designated service platform based on the access point;

[0009] According to the unique cloud access identifier, determine whether the city area where the network used by the specified business platform belongs has an edge cloud. If so, it will be identified as the access point to locate the target edge cloud to which it belongs.

[0010] In one embodiment, it further includes:

[0011] If the city or prefecture where the network used by the designated business platform belongs does not have an edge cloud, or if the target computing resources that meet the business requirements do not exist in the target edge cloud, analyze the network link quality between the access point and other city or prefecture edge clouds to obtain the network distance;

[0012] Other city edge clouds with the shortest network distance will be re-identified as target edge clouds.

[0013] In one implementation, the idle bare metal resources in the resource pool include:

[0014] Find the bare metal information summarized in the resource pool, which includes the machine macId, machine graphics card type, machine power on / off status, and machine occupancy status;

[0015] The load value of the resource pool is calculated according to the bare metal information, and the load value is compared with a preset first threshold. When the load value is lower than or equal to the first threshold, it is determined that there are idle bare metal resources in the resource pool.

[0016] In one embodiment, it further includes:

[0017] When the load value is higher than the preset second threshold, a test request is sent to the resource pool of other city edge clouds, and the delay time from the service switch where the resource pool of other city edge clouds is located from receiving the test request to responding to the test request is counted, and the network link quality is determined based on the delay time to obtain the network distance, and the other city edge clouds with the shortest network distance are re-determined as the target edge cloud.

[0018] In one implementation, locking the target computing resources and allocating them to the designated service platform for use includes:

[0019] Lock the target computing power resources and modify the switch port connected to the computing power network card corresponding to the target computing power resources to the VLAN ID pre-allocated by the specified diskless service, so that the bare metal resources and the specified diskless service are in the same LAN at the network level.

[0020] In one implementation, locking the target computing resources and allocating them to the designated service platform for use further includes:

[0021] When the bare metal resource is started, the DHCP server and TFTP service information of the specified diskless service are broadcast in the local area network. Through the DHCP server and TFTP service information, the bare metal resource runs the corresponding operating system and provides the required computing power service on the specified business platform.

[0022] In a second aspect, an embodiment of the present invention provides a consumer-grade graphics card reuse scheduling system that executes the consumer-grade graphics card reuse scheduling method as described above.

[0023] In a third aspect, an embodiment of the present invention provides an electronic device, the device comprising: a memory and a processor. The memory and the processor communicate with each other through an internal connection path, the memory is used to store instructions, the processor is used to execute the instructions stored in the memory, and when the processor executes the instructions stored in the memory, the processor executes the method in any one of the above-mentioned embodiments.

[0024] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, which stores a computer program. When the computer program runs on a computer, the method in any one of the above-mentioned embodiments is executed.

[0025] The advantages or beneficial effects of the above technical solution include at least:

[0026] The present invention combines diskless technology and an intelligent computing scheduling engine, determines the target edge cloud to which the access point belongs by specifying the access point of the network used by the business platform, and preferentially searches for a consumer-grade graphics card resource pool in the target edge cloud with the shortest distance on the user's physical link to provide computing power services to customers. If the resources in the nearest resource pool are not sufficient, the computing power resources with the minimum latency are searched in the consumer-grade graphics card resource pool in the edge cloud of other cities and allocated to the user.

[0027] The present invention simplifies the management and maintenance of computing devices through diskless technology, and realizes the separation of computing power resources and operating system images through diskless technology, so that the two are no longer dependent on the binding of physical lines, but are flexibly connected and bound through the network. The interaction between computing power resources and operating system images is transmitted through the network, thereby avoiding the limitations of traditional hardware and reducing the problems caused by hard disk failures.

[0028] At the same time, through the intelligent computing scheduling engine to schedule graphics card computing power, a dynamic combination of computing power and diskless services can be achieved, capabilities can be quickly provided to the outside world, manpower investment can be reduced, graphics card resources can be quickly reused, resource waste can be avoided, and thus hardware costs can be significantly reduced.

[0029] The above summary is for illustrative purposes only and is not intended to be limiting in any way. In addition to the illustrative aspects, embodiments and features described above, further aspects, embodiments and features of the present invention will be readily apparent by reference to the accompanying drawings and the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] In the accompanying drawings, unless otherwise specified, the same reference numerals throughout the multiple drawings represent the same or similar parts or elements. These drawings are not necessarily drawn to scale. It should be understood that these drawings only depict some embodiments disclosed in the present invention and should not be regarded as limiting the scope of the present invention.

[0031] Figure 1 It is a flow chart of the method for multiplexing and scheduling consumer-grade graphics cards of the present invention;

[0032] Figure 2 A schematic diagram of the connection between the intelligent scheduling engine of the present invention and each designated business platform;

[0033] Figure 3 A schematic diagram of the resource scheduling process of the intelligent scheduling engine of the present invention;

[0034] Figure 4 This is a schematic diagram of the resource pool load analysis process of the present invention;

[0035] Figure 5 This is a schematic diagram of target computing power resource allocation of the present invention;

[0036] Figure 6 A schematic diagram of the intelligent computing scheduling engine of the present invention performing computing power scheduling and network scheduling;

[0037] Figure 7 A schematic diagram of the intelligent scheduling engine of the present invention dynamically adjusting the service priorities of the service platform in different time periods;

[0038] Figure 8 The structure block diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0039] In the following, only some exemplary embodiments are briefly described. As those skilled in the art will appreciate, the described embodiments may be modified in various ways without departing from the spirit or scope of the present invention. Therefore, the drawings and descriptions are considered to be exemplary and non-restrictive in nature.

[0040] With the rapid development of artificial intelligence, big data, virtual reality and high-performance computing, the demand for graphics card (GPU) computing power in the consumer and professional markets is increasing day by day. The traditional way of using graphics cards is usually one computer corresponds to one or more graphics cards, which has low resource utilization efficiency, especially in a multi-user and multi-tasking environment. This method cannot give full play to the computing power of graphics cards.

[0041] In order to solve the above problems, an embodiment of the present invention proposes a consumer-grade graphics card reuse scheduling method, which can realize the free combination of computing power graphics card resources and diskless services according to business scenarios, and realize efficient scheduling and reuse of computing resources.

[0042] like Figure 1 As shown, the consumer-grade graphics card reuse scheduling method of this embodiment specifically includes the following steps:

[0043] Step S1: Obtain a resource application submitted by a designated service platform, determine the access point of the network used by the designated service platform according to the resource application, and locate the target edge cloud to which it belongs according to the access point.

[0044] It should be noted that a corresponding edge cloud is built in advance for each city in the province, and the consumer-grade graphics card resource pool information of each province and city is synchronized to the intelligent computing scheduling engine through the underlying cloud management interface for consumer-grade graphics card resource scheduling.

[0045] It should be noted that if Figure 2 As shown, the consumer-grade graphics card resource pool information mainly includes resource pool information, bare metal information, and diskless service information; the resource pool information includes the resource pool name, resource pool uuid, the local network where the resource pool is located, and the resource pool type (consumer-grade graphics card resource pool); the bare metal information in the resource pool includes the machine MAC Id, the machine graphics card type, the machine power on and off status, the machine occupancy status and other information; the diskless service information in the resource pool includes the diskless ecosystem manufacturer, the diskless service name, the capabilities provided by the diskless service (such as e-sports diskless can provide users with games, rendering diskless can provide users with rendering capabilities, etc.), the VID of the diskless service on the switch, the business switch code connected to the diskless service and other information.

[0046] In addition to determining the relationship between consumer-grade graphics cards and in-cloud service switches in the intelligent computing dispatching engine, the port relationship between the network card of each bare metal server and the in-cloud service switch also needs to be identified in the intelligent computing dispatching engine.

[0047] After the intelligent computing scheduling engine completes the above resource integration, the resource system uniformly plans the user access points and the user's home broadband optical modem to enter the edge cloud network channel to achieve interconnection between home broadband and edge cloud.

[0048] Users apply for resource allocation through designated business platforms, where designated business platforms include but are not limited to cloud e-sports platform, cloud rendering platform, cloud design platform, and cloud VR platform.

[0049] like Figure 3 As shown, after receiving the resource application, the intelligent scheduling engine determines the source of the request, preferentially determines the access point of the broadband network based on the user's home width, determines the city area where the access point is located based on the access point's unique cloud entry identifier, and determines whether there is a corresponding edge cloud in the city area where the access point is located. If an edge cloud exists in the city area where the access point is located, the edge cloud to which the access point belongs is determined as the target edge cloud.

[0050] If the city or prefecture where the network used by the designated business platform belongs does not have an edge cloud pre-built in place, the network link quality between the access point and other city or prefecture edge clouds is analyzed to obtain the network distance; the other city or prefecture edge clouds with the shortest network distance are re-determined as the target edge cloud.

[0051] Specifically, the network link quality between the access point and other municipal edge clouds is analyzed, specifically, the time required for data to be transmitted from the access point to the target edge cloud is calculated. The shorter this time is, the closer the network distance is and the higher the data transmission efficiency is.

[0052] In the resource allocation process of the intelligent computing scheduling engine, priority is given to the target edge cloud to which the access point belongs for resource allocation. When the target edge cloud does not meet business needs, the resource pool with the smallest delay is found from the consumer-grade graphics card resource pools of other edge clouds in other cities across nodes and allocated to users to ensure that users can get the best service experience. Assuming that an edge cloud resource pool has a short network distance, low network latency, sufficient bandwidth, and a stable network, it is more likely to be selected as the resource pool serving users. The resource scheduling strategy of this embodiment helps to reduce the transmission time of data in the network and improve the response speed and reliability of services.

[0053] Step S2: When there are target computing resources that meet the business requirements in the target edge cloud, the target computing resources are locked and allocated to the designated business platform for use.

[0054] In this embodiment, the business demand conditions must be met at the same time: there is a resource pool in the target edge cloud, there are idle bare metal resources in the resource pool, and there are specified diskless services required by the specified business platform in the resource pool. If any one of the conditions cannot be met, it is determined that the business demand conditions are not met.

[0055] It should be noted that bare metal resources usually refer to services that use physical servers as resources and provide them to users. Consumer-grade graphics card resource pools are all bare metal machines allocated to users of the business platform, and are immediately released and returned to the resource pool after use. The intelligent computing capability scheduling engine controls the status of these bare metal resources. After a bare metal machine is allocated to a platform user, the machine is directly marked as unschedulable and will be marked as schedulable after the user releases it.

[0056] The method for determining whether there are idle bare metal resources in the resource pool of the target edge cloud includes:

[0057] Step S21: Search for all bare metal information summarized in the resource pool of the target edge cloud, where the bare metal information includes machine macId, machine graphics card type, machine power on / off status, and machine occupancy status;

[0058] Step S22: Calculate the load value of the resource pool of the target edge cloud based on the bare metal information, compare the load value with a preset first threshold, and determine that there are idle bare metal resources in the resource pool when the load value is lower than or equal to the first threshold.

[0059] The calculation formula of the resource pool load value is:

[0060] resPoolLoad = N / M;

[0061] N is the number of allocated bare metals, and M is the total number of bare metals in the entire resource pool.

[0062] In this embodiment, the first threshold is set to 50%. When the load value is lower than or equal to 50%, it means that the computing power resources of the resource pool are idle. At the same time, a second threshold is set, and the second threshold can be set to 80%. If the load value is higher than 50% and lower than or equal to 80%, it means that the computing power resources of the resource pool are relatively tight; if the load value is greater than 80%, it means that the computing power resources of the resource pool are relatively scarce.

[0063] At the same time, based on the resource request, it searches for the target edge cloud to see whether the user's required consumer-grade graphics card type and diskless capability manufacturer exist. If the three conditions are met, that there is a resource pool in the target edge cloud, and there are idle bare metal resources in the resource pool, and there are specified diskless services required by the specified business platform in the resource pool, then the idle bare metal resources in the resource pool in the target edge cloud and the specified diskless services required by the specified business platform in the resource pool are locked as target computing resources and allocated to the user.

[0064] If the target edge cloud does not have the consumer-grade graphics card type required by the user or the resource pool does not have a diskless capability vendor, it means that the target edge cloud does not meet the business requirements. In this case, the network link quality between the access point and other city edge clouds is analyzed to obtain the network distance, and the other city edge clouds with the shortest network distance are re-determined as the target edge cloud.

[0065] This means that, in the process of calculating the load value of the target edge cloud, when the load value is higher than the preset second threshold, or when the load value is higher than the first threshold and lower than or equal to the second threshold, it means that the bare metal resources of the target edge cloud are not idle. In order to support cross-edge cloud scheduling, the intelligent computing scheduling engine Figure 4 As shown, the trigger rule sends a test request to the resource pool of the edge cloud in other cities, counts the delay time from the service switch where the resource pool of the edge cloud in other cities receives the test request to the time it responds to the test request, determines the network link quality based on the delay time, and obtains the network distance; and stores these data to provide a basis for the subsequent cross-node scheduling delay judgment.

[0066] When the intelligent computing scheduling engine fails to find idle resources to allocate to users through distance priority scheduling, it will automatically trigger cross-node scheduling to allocate consumer-grade graphics card computing power resource pools to users. The intelligent computing scheduling engine traverses all consumer-grade graphics card resource pools in the province to find consumer-grade graphics card resource pools that have idle bare metal resources and diskless services required by the business platform, and allocates bare metal resources with the lowest latency to users' access points.

[0067] It should be explained that the above-mentioned network distance analysis method is the same as the network distance analysis method in step S1. When business needs are not met, cross-edge cloud scheduling continues to look for idle resources to allocate to users, which helps to reduce data transmission time in the network and improve service response speed and reliability.

[0068] When a user applies for consumer-grade graphics card resources from the intelligent computing scheduling engine through a designated business platform, the above analysis determines the target computing resources such as the consumer-grade graphics card resource pool, bare metal MAC address, and required diskless services. At this time, the bare metal cannot directly find the required diskless service when it is turned on (because bare metal can be reused in different diskless services to support different business scenarios. At this time, the bare metal and the diskless service required by the user may not be in the same LAN).

[0069] Therefore, if Figure 5 As shown, in this embodiment, based on the three factors of consumer-grade graphics card resource pool, bare metal MAC address, and diskless service required by users, combined with the relationship between the consumer-grade graphics card MAC address and the service switch port, and the diskless service information in the resource pool, the intelligent computing scheduling engine can directly obtain the VID of the service switch where the currently requested diskless service in the resource pool is located, and at the same time, it can connect to the service switch through the SDN controller, and modify the service switch port of the bare metal network card connected to the other end to be the same as the diskless one, that is, modify the switch port connected to the computing power network card corresponding to the target computing power resource to the VLAN ID of the diskless service. The purpose is to ensure that the network communication between the target computing power resource and the diskless service can be carried out in the same VLAN, thereby realizing dynamic connection of the network connection between the target computing power resource and the diskless service.

[0070] It should be noted that, during the construction of the diskless service, corresponding VLAN IDs (VIDs for short) have been allocated in different resource pools.

[0071] During the boot process, the bare metal server broadcasts in the local area network to find the DHCP server and TFTP server of the diskless service, so that the target computing power resources can allocate system images and boot loaders through the diskless service to provide computing power services to customers.

[0072] Specifically, the DHCP server will assign an IP address to the target computing resource so that the target computing resource can be identified in the network and can receive other information from the network.

[0073] The TFTP (Trivial File Transfer Protocol) server stores the IPXE boot file. It responds to the target computing resource's request, announces its location and TFTP service information, and guides the target computing resource where to obtain the boot file. The target computing resource uses the IPXE network boot program to download the boot file from the server through the TFTP protocol. Among them, IPXE allows the operating system to be started from the network without the need for local storage devices.

[0074] Once IPXE boots, it requests a differential copy of the system master disk via the HTTP protocol, which means that only system files that have changed since the last copy will be transferred, saving bandwidth and speeding up the boot process. The downloaded system files are dynamically mounted to the target computing resource over the network, so that the target computing resource can start and run the operating system without the need for local storage system images.

[0075] The whole process realizes diskless booting of computing resources, that is, computing resources do not need to rely on local hard disks or other storage devices to start the operating system. This can simplify hardware configuration, reduce costs, and improve system flexibility and maintainability. In addition, since the operating system and data are stored on the network server, they can be updated and maintained more easily, while also improving data security.

[0076] like Figure 6 As shown, the intelligent computing scheduling engine provides a unified consumer-grade graphics card application service to the business platforms (such as cloud e-sports platform, cloud rendering platform, cloud design, cloud VR, etc.) facing the north. Customers apply for consumer-grade graphics card computing resources through these platforms; the intelligent computing scheduling engine dispatches to the idle consumer-grade graphics card resource pool through the previous consumer-grade graphics card resource identifier, and through network scheduling, the network between the computing resources and the corresponding diskless services is connected (for example, if the user applies for bare metal computing resources through the cloud e-sports platform, the scheduling engine dispatches the platform request to the consumer-grade graphics card resource pool 1, and allocates bare metal GPU server 1 to the customer. At this time, the allocated bare metal computing resources have not yet been allocated an operating system. The diskless service information in the resource pool can be used to know the VLAN ID of the cloud e-sports diskless service in the current resource pool, the switch code, switch IP and switch port connected to the computing mac port, and directly modify the vid of the bare metal computing resource connection port and the diskless service to be consistent (both located in vid1 in the switch), so as to achieve network connection between computing resources and diskless services).

[0077] In addition, according to statistics, different business platforms have different demands for consumer-grade graphics cards at different times of the day, such as Figure 7 As shown in the figure, the intelligent computing scheduling engine dynamically adjusts the business priorities of the business platform in different time periods, realizes the dynamic combination of the same consumer-grade graphics card computing power and diskless services in different time periods, and provides customers with diversified computing power services. The intelligent computing scheduling engine realizes the dynamic combination of computing power resources and diskless services, and the computing power machines are combined with different diskless services at different time periods, so that the same computing power can provide diversified computing power capabilities at different time periods.

[0078] Furthermore, to achieve computing power reuse driven by diskless technology, it is necessary to coordinate multiple key operations at the same time, including computing power scheduling, diskless service scheduling, network scheduling, computing power startup and shutdown, etc. The close coordination of these links can ensure the efficient reuse of computing power resources, realize dynamic allocation and management of resources through reasonable scheduling strategies, and improve the flexibility and utilization of the overall system.

[0079] Specifically, computing power scheduling and diskless service scheduling are dynamically combined through the intelligent computing scheduling engine; consumer-grade computing power resources have no local storage, but obtain operating systems, applications and data from central servers through the network; diskless services are distributed in different resource pools, and through the information aggregation of consumer-grade graphics card resource pools, it can be perceived that the current resource pool already has a list of diskless systems that support allocation, and the required diskless services are determined according to the business type association requested by the business system.

[0080] The principle of computing power scheduling and diskless service scheduling is to first find the edge cloud associated with the user's metropolitan area network based on the unique identifier of the user's home broadband cloud access, and find out whether there is a consumer-grade graphics card resource pool in the edge cloud, and whether there are idle bare metal resources and diskless services required by the business platform. If so, the intelligent computing scheduling engine determines the consumer-grade graphics card resource pool and the bare metal MAC address, locks the resources and allocates them to the user. If not, cross-node scheduling continues to find idle resources to allocate to the user.

[0081] It should be noted that since the network between computing resources and diskless services is isolated, in the process of implementing resource scheduling, it is necessary to open up the network between computing resources and corresponding diskless services to bind computing resources and diskless services and provide computing services to customers.

[0082] Computing power startup refers to the process in which the client initiates a startup request to the intelligent computing scheduling platform, and the platform issues instructions to control the startup process of computing power resources. When the computing power resources are powered on, they will broadcast a request to the DHCP server for diskless services in the local area network to obtain the assigned IP address, and at the same time announce the information of the TFTP service (the service stores the IPXE boot file). Next, the computing power resources enter the IPXE boot process. IPXE triggers the differential replication of the system mother disk through the HTTP protocol, and dynamically mounts the operating system to the computing power resources through the network, so that the diskless computing power resources appear to have disk resources.

[0083] Computing power shutdown refers to the process in which the client initiates a shutdown request to the intelligent computing scheduling platform, and the platform issues a shutdown command to control the shutdown of computing power; at the same time, the VID of the computing power resource mac connection switch port is reset to unbind the computing power resources and business association relationship (computing power resources change from stateful to stateless) to achieve diskless technology-driven computing power reuse.

[0084] This embodiment uses diskless technology to flexibly realize the free combination of computing resources and services, and meet the computing power reuse requirements of cloud e-sports, cloud design, cloud rendering and other scenarios at the business level. The system can dynamically schedule computing resources according to the peaks and troughs of business requests, thereby maximizing the use of limited computing power and achieving efficient allocation and optimization of resources.

[0085] Another embodiment of the present invention provides a consumer-grade graphics card reuse scheduling system, which executes the above-mentioned consumer-grade graphics card reuse scheduling method. The system simplifies the management and maintenance of computing devices through diskless technology, reduces the problems caused by hard disk failures, and at the same time, through the intelligent computing scheduling engine graphics card computing power scheduling, realizes the dynamic combination of computing power and diskless services, quickly provides capabilities to the outside world, reduces manpower investment, realizes rapid reuse of graphics card resources, avoids resource waste, and thus significantly reduces hardware costs.

[0086] It should be noted that the functions of the various modules of the system of the embodiment of the present invention can be found in the corresponding description of the above method, which will not be repeated here.

[0087] Another embodiment of the present invention provides an electronic device, Figure 8 FIG. 2 shows a structural block diagram of an electronic device according to an embodiment of the present invention. Figure 8 As shown, the electronic device includes: a memory 100 and a processor 200, wherein the memory 100 stores a computer program that can be run on the processor 200. When the processor 200 executes the computer program, the consumer-grade graphics card reuse scheduling method in the above embodiment is implemented. The number of the memory 100 and the processor 200 can be one or more.

[0088] The electronic device also includes:

[0089] The communication interface 300 is used to communicate with external devices and perform data exchange transmission.

[0090] If the memory 100, the processor 200 and the communication interface 300 are implemented independently, the memory 100, the processor 200 and the communication interface 300 can be connected to each other through a bus and communicate with each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc.

[0091] Optionally, in a specific implementation, if the memory 100, the processor 200 and the communication interface 300 are integrated on a chip, the memory 100, the processor 200 and the communication interface 300 can communicate with each other through an internal interface.

[0092] An embodiment of the present invention provides a computer-readable storage medium storing a computer program, which implements the method provided in the embodiment of the present invention when executed by a processor.

[0093] An embodiment of the present invention further provides a chip, which includes a processor for calling and executing instructions stored in the memory from the memory, so that a communication device equipped with the chip executes the method provided by the embodiment of the present invention.

[0094] An embodiment of the present invention also provides a chip, including: an input interface, an output interface, a processor and a memory, wherein the input interface, the output interface, the processor and the memory are connected via an internal connection path, and the processor is used to execute the code in the memory. When the code is executed, the processor is used to execute the method provided by the embodiment of the invention.

[0095] It should be understood that the processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. It is worth noting that the processor may be a processor supporting the advanced RISC machines (ARM) architecture.

[0096] Further, optionally, the above-mentioned memory may include a read-only memory and a random access memory, and may also include a non-volatile random access memory. The memory may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may include a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may include a random access memory (RAM), which is used as an external cache. By way of example but not limitation, many forms of RAM are available. For example, static RAM (SRAM), dynamic random access memory (DRAM), synchronous DRAM (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM) and direct rambus RAM (DR RAM).

[0097] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function according to the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium.

[0098] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine different embodiments or examples described in this specification and the features of different embodiments or examples, unless they are contradictory.

[0099] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0100] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of various changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for scheduling the reuse of consumer-grade graphics cards, characterized in that: include: Obtain a resource application submitted by a designated service platform, determine an access point of a network used by the designated service platform according to the resource application, and locate a target edge cloud to which the access point belongs according to the access point; When there are target computing resources that meet the business requirements in the target edge cloud, the target computing resources are locked and allocated to the designated business platform for use; wherein, meeting the business requirements includes that there are idle bare metal resources in the resource pool of the target edge cloud and that there are designated diskless services required by the designated business platform in the resource pool.

2. The method for scheduling the reuse of consumer-grade graphics cards according to claim 1, characterized in that: The locating the target edge cloud to which the access point belongs according to the access point includes: Determine the cloud entry unique identifier of the network used by the designated service platform according to the access office point; According to the cloud entry unique identifier, it is determined whether the city area where the network used by the designated service platform belongs has an edge cloud. If so, it is determined as the access point to locate the target edge cloud to which it belongs.

3. The method for scheduling the reuse of consumer-grade graphics cards according to claim 2, characterized in that: Also includes: If the city area to which the network used by the designated service platform belongs does not have an edge cloud, or if the target computing resources that meet the service requirements do not exist in the target edge cloud, the network link quality between the access point and other city edge clouds is analyzed to obtain the network distance; The other city edge cloud with the shortest network distance is re-determined as the target edge cloud.

4. The method for scheduling the reuse of consumer-grade graphics cards according to claim 1, characterized in that: The idle bare metal resources in the resource pool include: Searching for bare metal information summarized in the resource pool, the bare metal information including machine macId, machine graphics card type, machine power on / off status, and machine occupancy status; The load value of the resource pool is calculated according to the bare metal information, the load value is compared with a preset first threshold, and when the load value is lower than or equal to the first threshold, it is determined that there are idle bare metal resources in the resource pool.

5. The method for scheduling the reuse of consumer-grade graphics cards according to claim 4, characterized in that: Also includes: When the load value is higher than a preset second threshold, a test request is sent to the resource pool of other city edge clouds, and the delay time between the service switch where the resource pool of other city edge clouds is located receiving the test request and responding to the test request is counted, and the network link quality is determined according to the delay time to obtain the network distance; the other city edge cloud with the shortest network distance is re-determined as the target edge cloud.

6. The method for scheduling the reuse of consumer-grade graphics cards according to claim 1, characterized in that: The locking of the target computing resources and allocating them to the designated business platform for use includes: The target computing power resource is locked and the switch port connected to the computing power network card corresponding to the target computing power resource is modified to the VLAN ID pre-allocated by the designated diskless service, so that the bare metal resource and the designated diskless service are in the same local area network at the network level.

7. The method for scheduling the reuse of consumer-grade graphics cards according to claim 6, characterized in that: The locking of the target computing resources and allocating them to the designated service platform for use further includes: When the bare metal resource is started, the DHCP server and TFTP service information of the designated diskless service are broadcast in the local area network, and the bare metal resource is enabled to run the corresponding operating system through the DHCP server and the TFTP service information, so as to provide the required computing power service on the designated business platform.

8. A consumer-grade graphics card reuse scheduling system, characterized in that: Execute the consumer-grade graphics card reuse scheduling method as described in any one of claims 1 to 7.

9. An electronic device, characterized in that: include: A processor and a memory, wherein the memory stores instructions, and the instructions are loaded and executed by the processor to implement the consumer-grade graphics card reuse scheduling method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method for multiplexing and scheduling consumer-grade graphics cards according to any one of claims 1 to 7 is implemented.

Images