Resource scheduling method and device
Through dynamic resource scheduling method, dynamic deployment and expansion of resource pools are carried out according to algorithm encoding and resource level, solving the problems of low resource utilization and resource waste in the existing technology, and achieving efficient unified scheduling and maximum resource allocation for algorithms from different manufacturers Use.
Patent Information
- Application Number
- CN202410140077.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-01-31
- Publication Date
- 2024-11-19
- Estimated Expiration
- 2044-01-31
AI Technical Summary
In the video surveillance localization project, the existing technology resource scheduling method leads to low resource utilization and is unable to effectively respond to business changes, especially when the business volume is large, resulting in resource waste, and the algorithms of different manufacturers cannot be uniformly dispatched.
By receiving algorithm analysis tasks, determine whether the algorithm exists in the resource pool based on algorithm encoding and resource level, and dynamically deploy it according to the task load and capacity of the resource pool, give priority to the use of idle resources, and schedule resources in real time to maximize resource utilization, and Expand capacity as needed when resources are insufficient.
The resource utilization rate is improved, and the unified scheduling of different versions of algorithms from different manufacturers is realized, ensuring the maximum resource utilization of the resource pool and the stable operation of the algorithm.
Smart Images

Figure CN118152116B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of cloud computing technology, and in particular to a resource scheduling method and device. Background Art
[0002] In actual localized video surveillance projects, manual deployment of algorithm images is generally adopted. This static deployment method has poor flexibility, and the resource usage of the algorithm is fixed. In related technologies, the scheduling of algorithms is generally based on the sorting of resource load conditions, and the deployment of algorithms and the issuance of algorithm analysis tasks are performed on servers with less resource consumption through polling. This method assumes that the algorithms use the same resources by default, does not differentiate based on the algorithm capabilities, and only considers the current usage of algorithm resources. It does not respond promptly to actual business changes. The scheduling model of related technologies causes a large waste of resources when the business volume is large, and it is impossible to uniformly schedule different algorithms from multiple manufacturers. Summary of the invention
[0003] The embodiments of the present application provide a resource scheduling method and device to solve the problem in the related art that scheduling causes a large amount of resource waste and low resource utilization when the business volume is large.
[0004] In a first aspect, a resource scheduling method is provided, comprising:
[0005] receiving an algorithm analysis task, wherein the algorithm analysis task carries an algorithm code and an algorithm resource level of an algorithm to be analyzed;
[0006] According to the algorithm code, determining whether the algorithm to be analyzed exists in a resource pool, wherein multiple algorithms are deployed in the resource pool;
[0007] If the algorithm to be analyzed exists in the resource pool and the task load of the algorithm in the resource pool meets the expansion condition, or if the algorithm does not exist in the resource pool, deploy the algorithm to be analyzed according to the algorithm resource level of the algorithm to be analyzed and the resource capacity in the resource pool;
[0008] The algorithm analysis task is sent to the algorithm to be analyzed deployed in the resource pool, and the algorithm analysis task is analyzed by the algorithm to be analyzed deployed in the resource pool.
[0009] In a second aspect, a resource scheduling device is provided, including:
[0010] A receiving module, used for receiving an algorithm analysis task, wherein the algorithm analysis task carries an algorithm code and an algorithm resource level of an algorithm to be analyzed;
[0011] A first determination module is used to determine whether the algorithm to be analyzed exists in a resource pool according to the algorithm code, and multiple algorithms are deployed in the resource pool;
[0012] A deployment module, configured to deploy the algorithm to be analyzed according to the algorithm resource level of the algorithm to be analyzed and the resource capacity in the resource pool, when the algorithm to be analyzed exists in the resource pool and the task load of the algorithm in the resource pool meets the expansion condition, or when the algorithm does not exist in the resource pool;
[0013] The first analysis module is used to send the algorithm analysis task to the algorithm to be analyzed deployed in the resource pool, and analyze the algorithm analysis task through the algorithm to be analyzed deployed in the resource pool.
[0014] In a third aspect, an electronic device is provided, which includes a processor, a memory, and a program or instruction stored in the memory and executable on the processor, wherein the program or instruction, when executed by the processor, implements the steps of the method described in the first aspect.
[0015] In a fourth aspect, a readable storage medium is provided, on which a program or instruction is stored. When the program or instruction is executed by a processor, the steps of the method described in the first aspect are implemented.
[0016] In an embodiment of the present application, an algorithm analysis task is first received, and the algorithm analysis task carries the algorithm code and algorithm resource level of the algorithm to be analyzed. Then, based on the algorithm code, it is determined whether the algorithm to be analyzed exists in the resource pool. There are multiple algorithms deployed in the resource pool. When the algorithm to be analyzed exists in the resource pool and the task load of the algorithm in the resource pool meets the expansion conditions, or the algorithm does not exist in the resource pool, the algorithm to be analyzed is deployed according to the algorithm resource level of the algorithm to be analyzed and the resource capacity in the resource pool. Finally, the algorithm analysis task is sent to the algorithm to be analyzed deployed in the resource pool, and the algorithm analysis task is analyzed by the algorithm to be analyzed deployed in the resource pool. The embodiment of the present application preferentially schedules idle resources of the same level in the resource pool for algorithm deployment, and when the actual number of algorithm tasks meets the expansion requirements, the resources are scheduled in real time to complete the algorithm deployment, so as to maximize the utilization of the computing power resources of the existing resource pool and improve the resource utilization rate. When the resource pool is insufficient, the resource pool is expanded on demand to complete the algorithm deployment. BRIEF DESCRIPTION OF THE DRAWINGS
[0017] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:
[0018] Figure 1is a flow chart of a resource scheduling method provided in an embodiment of the present application;
[0019] Figure 2 is a detailed flow chart of the resource scheduling method provided in the embodiment of the present application;
[0020] Figure 3 is a schematic diagram of a resource scheduling device provided in an embodiment of the present application;
[0021] Figure 4 It is a structural block diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION
[0022] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.
[0023] The terms "first", "second", etc. in the specification and claims of the present application are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable under appropriate circumstances, so that the embodiments of the present application can be implemented in an order other than those illustrated or described here. In addition, the "and / or" in the specification and claims represents at least one of the connected objects, and the character " / " generally represents that the objects associated with each other are in an "or" relationship.
[0024] In the following, in conjunction with the accompanying drawings, a resource scheduling method and device provided by an embodiment of the present application are described in detail through specific embodiments and their application scenarios.
[0025] like Figure 1 As shown, it is a flow chart of a resource scheduling method provided by an embodiment of the present application. Figure 1 As shown, the resource scheduling method may include: contents shown in S101 to S104.
[0026] In S101, an algorithm analysis task is received, where the algorithm analysis task carries the algorithm code and algorithm resource level of the algorithm to be analyzed.
[0027] The algorithm analysis task comes from the algorithm configured by the customer in the upper application system. The algorithm analysis task also includes information such as the task unique code (Identity document, ID), algorithm name, algorithm analysis type, and algorithm task load threshold.
[0028] In S102, according to the algorithm code, it is determined whether the algorithm to be analyzed exists in the resource pool, and multiple algorithms are deployed in the resource pool.
[0029] That is, the algorithm code of the algorithm to be analyzed may be searched in the algorithms deployed in the resource pool to determine whether the algorithm to be analyzed exists in the resource pool.
[0030] In S103, if the algorithm to be analyzed exists in the resource pool and the task load of the algorithm in the resource pool meets the expansion condition, or the algorithm does not exist in the resource pool, the algorithm to be analyzed is deployed according to the algorithm resource level of the algorithm to be analyzed and the resource capacity in the resource pool.
[0031] That is to say, if the algorithm to be analyzed has been deployed in the resource pool, it is further determined whether the current number of tasks of the algorithm in the resource pool has reached the expansion condition. If the expansion condition is reached, the algorithm to be analyzed is deployed according to the algorithm resource level of the algorithm to be analyzed and the resource capacity in the resource pool. Or if the algorithm to be analyzed is not deployed in the resource pool, it is also necessary to deploy the algorithm to be analyzed according to the algorithm resource level of the algorithm to be analyzed and the resource capacity in the resource pool. The specific process is described in detail in the subsequent implementation method.
[0032] In S104, the algorithm analysis task is sent to the algorithm to be analyzed deployed in the resource pool, and the algorithm analysis task is analyzed by the algorithm to be analyzed deployed in the resource pool.
[0033] In an embodiment of the present application, an algorithm analysis task is first received, and the algorithm analysis task carries the algorithm code and algorithm resource level of the algorithm to be analyzed. Then, based on the algorithm code, it is determined whether the algorithm to be analyzed exists in the resource pool. There are multiple algorithms deployed in the resource pool. When the algorithm to be analyzed exists in the resource pool and the task load of the algorithm in the resource pool meets the expansion conditions, or the algorithm does not exist in the resource pool, the algorithm to be analyzed is deployed according to the algorithm resource level of the algorithm to be analyzed and the resource capacity in the resource pool. Finally, the algorithm analysis task is sent to the algorithm to be analyzed deployed in the resource pool, and the algorithm analysis task is analyzed by the algorithm to be analyzed deployed in the resource pool. The embodiment of the present application preferentially schedules idle resources of the same level in the resource pool for algorithm deployment, and when the actual number of algorithm tasks meets the expansion requirements, the resources are scheduled in real time to complete the algorithm deployment, so as to maximize the utilization of the computing power resources of the existing resource pool and improve the resource utilization rate. When the resource pool is insufficient, the resource pool is expanded on demand to complete the algorithm deployment.
[0034] In a possible implementation of the present application, after determining whether the algorithm to be analyzed exists in the resource pool based on the algorithm code, and multiple algorithms are deployed in the resource pool, the resource scheduling method may also include: in the case that the algorithm to be analyzed exists in the resource pool, determining the number of analysis tasks of the first algorithm according to the algorithm code of the first algorithm corresponding to the algorithm to be analyzed in the resource pool; determining the task load after the first algorithm receives the algorithm to be analyzed according to the algorithm resource level of the algorithm to be analyzed and the algorithm code of the algorithm to be analyzed; in the case that the task load is less than the task load threshold, sending the algorithm analysis task to the first algorithm, and analyzing the algorithm analysis task by the first algorithm.
[0035] That is to say, when the algorithm to be analyzed exists in the resource pool, the current number of analysis tasks of the first algorithm is determined according to the algorithm code of the first algorithm corresponding to the algorithm to be analyzed in the resource pool, and it is judged whether the task load of the first algorithm after receiving the algorithm to be analyzed meets the capacity expansion conditions; if the task load is less than the task load threshold, there is no need to expand the capacity, and the algorithm analysis task is directly sent to the first algorithm, and the first algorithm can analyze the algorithm analysis task.
[0036] Since different manufacturers and different versions of algorithms use different GPU, video memory, and memory of graphics processing unit (GPU) servers of the same specifications, the exclusive whole-card deployment algorithm used in the related art will cause a waste of computing resources. This embodiment classifies the resource specifications of GPU cards. For example, they can be divided into three levels (1C4G 2G), two levels (4C 16G 8G), and one level (8C32G 16G). The algorithm is tested to determine the level supported by each algorithm and the maximum task threshold at the supported resource specification level. This embodiment uses the lowest resource level supported by the algorithm for deployment when the algorithm is first deployed. Taking 4 GPU cards as an example, a server has 4 GPU cards. The exclusive whole-card deployment of the related art can only deploy 4 algorithms. The hierarchical deployment provided by this embodiment can deploy 12 three-level algorithms or 8 two-level algorithms or 4 first-level algorithms. The use of mixed deployment according to the different capabilities of the algorithms can improve resource utilization.
[0037] Deploying the algorithm on the whole card causes a waste of resources, and different levels of algorithms deployed on the same card interfere with each other and easily cause algorithm instability. Therefore, this embodiment comprehensively considers the stable deployment of algorithms and the maximization of resource utilization, and classifies resources for algorithms of different manufacturers and versions. To ensure the stable operation of the algorithm, it is stipulated that only algorithms of the same resource level are deployed on a single GPU card until all are occupied before occupying other GPU cards. Specifically, the maximum task threshold of algorithm i under different resource specifications k, k = 1, 2, 3 is a ik , the actual analysis task load of the algorithm is Ai , i = 1, 2, 3, ..., the number of currently deployed algorithm containers is n ik , n ik ≥0. When A i <n ik* a ik When A i ≥n ik* a ik , it means that the algorithm meets the expansion conditions and needs to be expanded before deploying the algorithm to be analyzed.
[0038] In a possible implementation of the present application, deploying the algorithm to be analyzed according to the algorithm resource level of the algorithm to be analyzed and the resource capacity in the resource pool may include: determining whether there are remaining resources in the resource pool; if there are remaining resources in the resource pool, judging whether there are free resources on the first GPU card with the same algorithm resource level as the algorithm to be analyzed; if there are free resources on the first GPU card, deploying the algorithm to be analyzed on the first GPU card; if there are no free resources on the first GPU card, creating a new container on the free GPU card in the resource pool and deploying the algorithm to be analyzed; if there are no remaining resources in the resource pool, expanding the resource pool; creating a new container on the second GPU in the expanded resource pool and deploying the algorithm to be analyzed, the algorithm resource level of the second GPU being the same as the algorithm resource level of the algorithm to be analyzed.
[0039] Among them, the total amount of resources in the resource pool can be obtained by multiplying the number of GPU cards in the resource pool by the capacity of each card, and the current actual amount of resources used is the sum of the GPU card resource usage of each deployed algorithm. In order to enable real-time deployment of algorithms and real-time analysis of algorithm tasks, this embodiment performs minimum pre-expansion on the resource pool. When the total amount of resources minus the current actual amount of resources used is less than 1.5 times the resources of a single server, it means that there are insufficient resources in the current resource pool and the current resource pool needs to be expanded. Create a new container on the newly expanded GPU server according to the resource level of the algorithm to be analyzed, deploy the algorithm to be analyzed, and mark the GPU as the resource level corresponding to the algorithm to be analyzed. If there are remaining resources in the current resource pool, find the GPU card of the resource level corresponding to the current algorithm to be analyzed, and determine whether there are free resources on the GPU card of the resource level. When there are free GPU cards of the same level of resources, the algorithm to be analyzed is deployed on the free GPU cards of the same resource level as the algorithm to be analyzed according to the principle of optimal resource utilization; when all resources of the same level are occupied, a new container is created on the free GPU cards in the resource pool and the algorithm to be analyzed is deployed, and the GPU is marked as the resource level corresponding to the algorithm.
[0040] It is worth noting that when expanding the resource pool, you can only expand one server at a time.
[0041] like Figure 2 The figure is a detailed flow chart of a resource scheduling method provided in an embodiment of the present application. Specifically, it is as follows:
[0042] Receive an algorithm analysis task; determine whether the algorithm to be analyzed already exists in the resource pool; if the algorithm to be analyzed has been deployed, calculate the number of analysis tasks of the first algorithm corresponding to the algorithm to be analyzed in the resource pool; determine whether the first algorithm meets the expansion conditions; if the expansion conditions are met, or the algorithm to be analyzed is not deployed in the resource pool, determine whether there are remaining resources in the existing resource pool; if there are remaining resources, determine whether there are idle resources on the GPU card corresponding to the resource level of the algorithm to be analyzed; if there are idle resources, deploy the algorithm to be analyzed on the GPU card corresponding to the resource level of the algorithm to be analyzed, and send the algorithm analysis task to the newly deployed algorithm to be analyzed, and the algorithm to be analyzed analyzes the algorithm analysis task; if the GPU card corresponding to the resource level of the algorithm to be analyzed If there are no idle resources on the U card, a new container is created on the spare GPU card and the algorithm to be analyzed is deployed according to the principle of optimal resource utilization and smooth algorithm deployment, and the algorithm analysis task is sent to the newly deployed algorithm to be analyzed, which analyzes the algorithm analysis task; if there are no remaining resources in the existing resource pool, it is prompted that the resources are insufficient, and the resource pool is expanded, a new container is created on the newly expanded GPU card and the algorithm to be analyzed is deployed, and the algorithm analysis task is sent to the newly deployed algorithm to be analyzed, which analyzes the algorithm analysis task; if the first algorithm does not meet the expansion conditions, the algorithm deployed in the resource pool is searched according to the algorithm code of the algorithm to be expanded, and the algorithm analysis task is sent to the algorithm, which analyzes the algorithm analysis task.
[0043] The embodiments of the present application comprehensively consider the optimal utilization of resources and the smooth deployment of algorithms, and utilize the resource scheduling method provided in the above embodiments to achieve unified scheduling of algorithms of different versions from different manufacturers, maximize the utilization of existing resources, and ensure the stable operation of the algorithm.
[0044] like Figure 3 As shown, it is a schematic diagram of a resource scheduling device provided in an embodiment of the present application. Figure 3 As shown, the resource scheduling device may include: a receiving module 301 , a first determining module 302 , a deployment module 303 and a first analyzing module 304 .
[0045] Among them, the receiving module 301 is used to receive an algorithm analysis task, which carries the algorithm code and algorithm resource level of the algorithm to be analyzed; the first determination module 302 is used to determine whether the algorithm to be analyzed exists in the resource pool according to the algorithm code, and multiple algorithms are deployed in the resource pool; the deployment module 303 is used to deploy the algorithm to be analyzed according to the algorithm resource level of the algorithm to be analyzed and the resource capacity in the resource pool when the algorithm to be analyzed exists in the resource pool and the task load of the algorithm in the resource pool meets the expansion condition, or the algorithm does not exist in the resource pool; the first analysis module 304 is used to send the algorithm analysis task to the algorithm to be analyzed deployed in the resource pool, and analyze the algorithm analysis task through the algorithm to be analyzed deployed in the resource pool.
[0046] In an embodiment of the present application, an algorithm analysis task is first received, and the algorithm analysis task carries the algorithm code and algorithm resource level of the algorithm to be analyzed. Then, based on the algorithm code, it is determined whether the algorithm to be analyzed exists in the resource pool. There are multiple algorithms deployed in the resource pool. When the algorithm to be analyzed exists in the resource pool and the task load of the algorithm in the resource pool meets the expansion conditions, or the algorithm does not exist in the resource pool, the algorithm to be analyzed is deployed according to the algorithm resource level of the algorithm to be analyzed and the resource capacity in the resource pool. Finally, the algorithm analysis task is sent to the algorithm to be analyzed deployed in the resource pool, and the algorithm analysis task is analyzed by the algorithm to be analyzed deployed in the resource pool. The embodiment of the present application preferentially schedules idle resources of the same level in the resource pool for algorithm deployment, and when the actual number of algorithm tasks meets the expansion requirements, the resources are scheduled in real time to complete the algorithm deployment, so as to maximize the utilization of the computing power resources of the existing resource pool and improve the resource utilization rate. When the resource pool is insufficient, the resource pool is expanded on demand to complete the algorithm deployment.
[0047] In a possible implementation manner of the present application, the resource scheduling device may further include: a second determination module, a third determination module and a second analysis module.
[0048] Among them, the second determination module is used to determine the number of analysis tasks of the first algorithm according to the algorithm code of the first algorithm corresponding to the algorithm to be analyzed in the resource pool when the algorithm to be analyzed exists in the resource pool; the third determination module is used to determine the task load after the first algorithm receives the algorithm to be analyzed according to the algorithm resource level of the algorithm to be analyzed and the algorithm code of the algorithm to be analyzed; the second analysis module is used to send the algorithm analysis task to the first algorithm when the task load is less than the task load threshold, and analyze the algorithm analysis task through the first algorithm.
[0049] In a possible implementation of the present application, the deployment module 303 is used to: determine whether there are remaining resources in the resource pool; if there are remaining resources in the resource pool, determine whether there are free resources on the first GPU card with the same algorithm resource level as the algorithm to be analyzed; if there are free resources on the first GPU card, deploy the algorithm to be analyzed on the first GPU card.
[0050] In a possible implementation of the present application, the deployment module 303 is used to: when there are no free resources on the first GPU card, create a new container on a free GPU card in the resource pool and deploy the algorithm to be analyzed.
[0051] In a possible implementation of the present application, the deployment module 303 is used to: expand the resource pool when there are no remaining resources in the resource pool; create a new container on the second GPU in the expanded resource pool, and deploy the algorithm to be analyzed, and the algorithm resource level of the second GPU is the same as the algorithm resource level of the algorithm to be analyzed.
[0052] The function of the resource scheduling device described in this application has been Figure 1-2 The method embodiment shown is described in detail, so for any details not provided in the description of this embodiment, please refer to the relevant descriptions in the aforementioned embodiments, which will not be repeated here.
[0053] like Figure 4 As shown, an embodiment of the present application also provides an electronic device 400, including a processor 401, a memory 402, and a program or instruction stored in the memory 402 and executable on the processor 401. When the program or instruction is executed by the processor 401, each process of the above-mentioned resource scheduling method embodiment is implemented, and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.
[0054] The embodiment of the present application also provides a storage medium, on which a program or instruction is stored, and when the program or instruction is executed by a processor, each process of the resource scheduling method embodiment provided in any of the above embodiments is implemented. And the same technical effect can be achieved, and to avoid repetition, it will not be repeated here.
[0055] The processor is the processor in the electronic device described in the above embodiment. The storage medium includes computer storage media, such as computer read-only memory (ROM), random access memory (RAM), magnetic disk or optical disk, etc.
[0056] It should be noted that, in this article, the term "comprises", "includes" or any other variant thereof is intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "including one..." do not exclude the presence of other identical elements in the process, method, article or device including the element. In addition, it should be pointed out that the scope of the method and device in the embodiment of the present application is not limited to performing functions in the order shown or discussed, and may also include performing functions in a substantially simultaneous manner or in reverse order according to the functions involved, for example, the described method may be performed in an order different from that described, and various steps may also be added, omitted, or combined. In addition, the features described with reference to certain examples may be combined in other examples.
[0057] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a computer software product, which is stored in a storage medium (such as ROM / RAM, a disk, or an optical disk), and includes a number of instructions for a terminal (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present application.
[0058] The embodiments of the present application are described above in conjunction with the accompanying drawings, but the present application is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present application, ordinary technicians in this field can also make many forms without departing from the purpose of the present application and the scope of protection of the claims, all of which are within the protection of the present application.
Claims
1. A resource scheduling method, characterized in that: include: receiving an algorithm analysis task, wherein the algorithm analysis task carries an algorithm code and an algorithm resource level of an algorithm to be analyzed; According to the algorithm code, determining whether the algorithm to be analyzed exists in a resource pool, wherein multiple algorithms are deployed in the resource pool; If the algorithm to be analyzed exists in the resource pool and the task load of the algorithm in the resource pool meets the expansion condition, or if the algorithm does not exist in the resource pool, deploy the algorithm to be analyzed according to the algorithm resource level of the algorithm to be analyzed and the resource capacity in the resource pool; Sending the algorithm analysis task to the algorithm to be analyzed deployed in the resource pool, and analyzing the algorithm analysis task by the algorithm to be analyzed deployed in the resource pool; After determining, according to the algorithm code, whether the algorithm to be analyzed exists in a resource pool, and multiple algorithms are deployed in the resource pool, the method further includes: In the case that the algorithm to be analyzed exists in the resource pool, determining the number of analysis tasks of the first algorithm according to the algorithm code of the first algorithm corresponding to the algorithm to be analyzed in the resource pool; Determining, according to the algorithm resource level of the algorithm to be analyzed and the algorithm code of the algorithm to be analyzed, a task load after the first algorithm receives the algorithm to be analyzed; When the task load is less than a task load threshold, the algorithm analysis task is sent to the first algorithm, and the algorithm analysis task is analyzed by the first algorithm.
2. The resource scheduling method according to claim 1, characterized in that: The deploying the algorithm to be analyzed according to the algorithm resource level of the algorithm to be analyzed and the resource capacity in the resource pool includes: Determining whether there are remaining resources in the resource pool; If there are remaining resources in the resource pool, determining whether there are idle resources on a first GPU card having the same algorithm resource level as the algorithm to be analyzed; When there are idle resources on the first GPU card, the algorithm to be analyzed is deployed on the first GPU card.
3. The resource scheduling method according to claim 2, characterized in that: The method further comprises: In the case that there are no idle resources on the first GPU card, a new container is created on an idle GPU card in the resource pool, and the algorithm to be analyzed is deployed.
4. The resource scheduling method according to claim 2, characterized in that: The method further comprises: When there are no remaining resources in the resource pool, expanding the resource pool; A new container is created on the second GPU in the expanded resource pool, and the algorithm to be analyzed is deployed, where the algorithm resource level of the second GPU is the same as the algorithm resource level of the algorithm to be analyzed.
5. A resource scheduling device, characterized in that: include: A receiving module, used for receiving an algorithm analysis task, wherein the algorithm analysis task carries an algorithm code and an algorithm resource level of an algorithm to be analyzed; A first determination module is used to determine whether the algorithm to be analyzed exists in a resource pool according to the algorithm code, and multiple algorithms are deployed in the resource pool; A deployment module, configured to deploy the algorithm to be analyzed according to the algorithm resource level of the algorithm to be analyzed and the resource capacity in the resource pool, when the algorithm to be analyzed exists in the resource pool and the task load of the algorithm in the resource pool meets the expansion condition, or when the algorithm does not exist in the resource pool; A first analysis module, used for sending the algorithm analysis task to the algorithm to be analyzed deployed in the resource pool, and analyzing the algorithm analysis task through the algorithm to be analyzed deployed in the resource pool; The device also includes: A second determination module is used to determine the number of analysis tasks of the first algorithm according to the algorithm code of the first algorithm corresponding to the algorithm to be analyzed in the resource pool when the algorithm to be analyzed exists in the resource pool; A third determination module, configured to determine a task load after the first algorithm receives the algorithm to be analyzed according to the algorithm resource level of the algorithm to be analyzed and the algorithm code of the algorithm to be analyzed; A second analysis module is configured to send the algorithm analysis task to the first algorithm and analyze the algorithm analysis task through the first algorithm when the task load is less than a task load threshold.
6. The resource scheduling device according to claim 5, characterized in that: The deployment module is used to: Determining whether there are remaining resources in the resource pool; If there are remaining resources in the resource pool, determining whether there are idle resources on a first GPU card having the same algorithm resource level as the algorithm to be analyzed; When there are idle resources on the first GPU card, the algorithm to be analyzed is deployed on the first GPU card.
7. The resource scheduling device according to claim 6, characterized in that: The deployment module is used to: When there are no remaining resources in the resource pool, expanding the resource pool; A new container is created on the second GPU in the expanded resource pool, and the algorithm to be analyzed is deployed, where the algorithm resource level of the second GPU is the same as the algorithm resource level of the algorithm to be analyzed.
8. An electronic device, characterized in that: It comprises a processor, a memory and a program or instruction stored in the memory and executable on the processor, wherein the program or instruction, when executed by the processor, implements the steps of the resource scheduling method as described in any one of claims 1 to 4.
