Collaborative resource scheduling method and system based on programmable network

Abstract

The invention discloses a collaborative resource scheduling method and system based on a programmable network. The method comprises the following steps: performing performance modeling on heterogeneous computing resource data in all computing resource data to obtain a uniform resource performance index; defining target computing resource data required by multi-dimensional resource demand services, and determining matching relationship information between each multi-dimensional resource demand service and the current computer cluster; in the deployment process, according to all the matching relation information, the correspondingly selected multi-dimensional resource demand service is deployed to each computer cluster; determining a shortest deployment path between the routing nodes in each computer cluster based on a Floyd algorithm; and in the actual scheduling process, executing the computing power arrangement scheme generated in the deployment process. According to the collaborative resource scheduling method and system based on the programmable network provided by the embodiment of the invention, the utilization rate of computing power resources in a wide area network can be improved, and a computing power scheduling strategy is dynamically generated.

Description

Technical Field

[0001] The present invention relates to the technical field of computer computing power resources, and in particular to a collaborative resource scheduling method and system based on a programmable network. Background Art

[0002] With the rapid development of cloud computing, edge computing, and network communication technologies, traditional computing resource scheduling methods still have obvious limitations when facing multi-cloud, cross-regional large-scale computing scenarios.

[0003] Existing computing resource scheduling systems usually rely on pre-set scheduling strategies when facing dynamically changing business needs, lacking flexibility and real-time response capabilities. When business needs change rapidly or sudden computing loads occur, these systems often find it difficult to adjust computing power allocation in a timely manner, resulting in reduced resource utilization.

[0004] It can be seen that how to improve the utilization of computing resources in a wide area network has become a technical problem that technical personnel in this field need to solve urgently. Summary of the invention

[0005] The present invention provides a collaborative resource scheduling method and system based on a programmable network to solve the technical problem of low utilization of computing resources, so as to achieve the effect of improving the utilization of computing resources and dynamically generating computing scheduling strategies.

[0006] In order to solve the above technical problems, an embodiment of the present invention provides a collaborative resource scheduling method based on a programmable network, comprising:

[0007] Acquire all computing resource data of the computer cluster, and perform performance modeling on heterogeneous computing resource data in all the computing resource data to obtain a unified resource performance index;

[0008] Defining each target computing resource data required by the multi-dimensional resource demand service, and determining matching relationship information between each of the multi-dimensional resource demand services and the current computer cluster based on a linear relationship between different target computing resource data in the current computer cluster;

[0009] During the deployment process, according to all the matching relationship information, the corresponding selected multi-dimensional resource demand services are deployed to each of the computer clusters respectively; and during the deployment process, the shortest deployment path between each routing node in each of the computer clusters is determined based on the Floyd algorithm;

[0010] During the actual scheduling process, the computing power scheduling plan generated by the deployment process is executed.

[0011] As one preferred solution, the linear relationship between different target computing resource data in the computer cluster is expressed as:

[0012] Score(cluster i )=α cpu X CPU +α gpu X GPU +α mem X Mem +α io X IO

[0013] Among them, Score(cluster i ) represents the matching relationship information between the multi-dimensional resource demand business and the current computer cluster, and its value range is [0,4], α x is the weight of different target computing resource data to all computing resource data. The x type includes CPU resources, GPU resources, memory resources and IO resources. Memory resources are expressed in Mem. x Indicates the matching relationship between different target computing resource data types and the current computer cluster of the multi-dimensional resource demand business;

[0014] The matching relationship between the different target computing resource data types and the current computer cluster of the multi-dimensional resource demand business is expressed as:

[0015]

[0016] Among them, cluster i (x) represents the weight coefficient of the x-type computing resource data of the ith computer cluster, x max Indicates the highest weight coefficient of type x resource in all current computer clusters.

[0017] As one of the preferred solutions, in the actual scheduling process, executing the computing power scheduling plan generated by the deployment process includes:

[0018] Construct an SDN architecture, establish a network topology structure based on the multi-dimensional resources on the SDN architecture, and adopt an SRV6-based computing power orchestration solution generated by the deployment process.

[0019] As one of the preferred solutions, the SDN architecture includes an SDN controller, a southbound interface, a soft routing device and an SDN control plane;

[0020] The SDN control plane is located on the SDN controller, and the SDN control plane implements routing management and path orchestration based on network programming and path management technology;

[0021] The soft routing device is connected to the SDN controller via a southbound interface, and is used to execute instructions issued by the SDN controller.

[0022] As one of the preferred solutions, the computing power orchestration solution generated by the deployment process based on SRV6 includes:

[0023] Using SID as a network node identifier, generating a SID list, and dynamically allocating and adjusting the SID according to network status and traffic;

[0024] The Floyd algorithm is used to calculate the shortest path to obtain a routing path;

[0025] When the data is sent from the source node, the SID list is combined with SRV6. When the data reaches the intermediate node from the routing path, the content of SRV6 is parsed, the adjusted SID that needs to be processed is identified, and the processed SID is sent to its corresponding next routing node to execute and complete the computing power orchestration plan.

[0026] Another embodiment of the present invention provides a collaborative resource scheduling system based on a programmable network, comprising:

[0027] A resource perception module is used to obtain all computing resource data of a computer cluster, and to perform performance modeling on heterogeneous computing resource data in all the computing resource data to obtain a unified resource performance index;

[0028] A cluster allocation module, used to define each target computing resource data required by the multi-dimensional resource demand business, and determine the matching relationship information between each multi-dimensional resource demand business and the current computer cluster based on the linear relationship between different target computing resource data in the current computer cluster;

[0029] A cross-cloud routing control module is used to deploy the corresponding selected multi-dimensional resource demand services to each of the computer clusters according to all the matching relationship information during the deployment process; and to determine the shortest deployment path between each routing node in each of the computer clusters based on the Floyd algorithm during the deployment process;

[0030] The cloud computing power orchestration module is used to execute the computing power orchestration plan generated by the deployment process during the actual scheduling process.

[0031] As one preferred solution, the linear relationship between different target computing resource data in the computer cluster is expressed as:

[0032] Score(clusteri )=α cpu X CPU +α gpu X GPU +α mem X Mem +α io X IO

[0033] Among them, Score(cluster i ) represents the matching relationship information between the multi-dimensional resource demand business and the current computer cluster, and its value range is [0,4], α x is the weight of different target computing resource data to all computing resource data. The x type includes CPU resources, GPU resources, memory resources and IO resources. Memory resources are expressed in Mem. x Indicates the matching relationship between different target computing resource data types and the current computer cluster of the multi-dimensional resource demand business;

[0034] The matching relationship between the different target computing resource data types and the current computer cluster of the multi-dimensional resource demand business is expressed as:

[0035]

[0036] Among them, cluster i (x) represents the weight coefficient of the x-type computing resource data of the ith computer cluster, x max Indicates the highest weight coefficient of type x resource in all current computer clusters.

[0037] As one of the preferred solutions, in the actual scheduling process, the computing power scheduling plan generated by the deployment process is executed, including

[0038] Construct an SDN architecture, establish a network topology structure based on the multi-dimensional resources on the SDN architecture, and adopt an SRV6-based computing power orchestration solution generated by the deployment process.

[0039] As one of the preferred solutions, the SDN architecture includes an SDN controller, a southbound interface, a soft routing device and an SDN control plane;

[0040] The SDN control plane is located on the SDN controller, and the SDN control plane implements routing management and path orchestration based on network programming and path management technology;

[0041] The soft routing device is connected to the SDN controller via a southbound interface, and is used to execute instructions issued by the SDN controller.

[0042] As one of the preferred solutions, the computing power orchestration solution generated by the deployment process based on SRV6 includes:

[0043] Using SID as a network node identifier, generating a SID list, and dynamically allocating and adjusting the SID according to network status and traffic;

[0044] The Floyd algorithm is used to calculate the shortest path to obtain a routing path;

[0045] When the data is sent from the source node, the SID list is combined with SRV6. When the data reaches the intermediate node from the routing path, the content of SRV6 is parsed, the adjusted SID that needs to be processed is identified, and the processed SID is sent to its corresponding next routing node to execute and complete the computing power orchestration plan.

[0046] Compared with the prior art, the embodiments of the present invention have the following advantages:

[0047] (1) The present invention provides a unified abstract description of different types of computing resources, and through refined modeling of computing performance, establishes a globally unique computing resource identification system, thereby solving the scheduling problem of diverse and highly heterogeneous computing resources in a complex environment;

[0048] (2) The present invention proposes a weight-based multi-dimensional resource scheduling method for multiple dimensions such as CPU, GPU, memory and IO resources. The system dynamically adjusts the weight of resources in each dimension according to the actual needs of the business, optimizes resource utilization through precise resource matching and allocation strategies, and ensures efficient computing power scheduling and system performance improvement;

[0049] (3) In a cross-cloud and cloud-edge environment, the present invention combines network topology information and computing node status, and uses the Floyd algorithm to calculate the shortest path between any pair of nodes in the network, thereby achieving efficient coordination and scheduling of cross-domain resources. Through SID expansion and dynamic path adjustment, it ensures that computing power requests can be transmitted to the best computing power node via the optimal path, thereby maximizing resource utilization and improving business processing performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 It is a flowchart of a collaborative resource scheduling method based on a programmable network in one embodiment of the present invention;

[0051] Figure 2 It is a schematic diagram of a structure block diagram of a collaborative resource scheduling system based on a programmable network in one embodiment of the present invention;

[0052] Reference numerals:

[0053] Among them, 11. Resource perception module; 12. Cluster allocation module; 13. Cross-cloud routing control module; 14. In-cloud computing power orchestration module. DETAILED DESCRIPTION

[0054] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. The purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0055] In the description of this application, the terms "first", "second", "third", etc. are used for descriptive purposes only and are not to be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first", "second", "third", etc. may explicitly or implicitly include one or more of the feature. In the description of this application, unless otherwise specified, "plurality" means two or more.

[0056] In the description of the present application, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be a connection between the two elements. The terms "vertical", "horizontal", "left", "right", "upper", "lower" and similar expressions used herein are only for illustrative purposes, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention. The term "and / or" used herein includes any and all combinations of one or more related listed items. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood according to specific circumstances.

[0057] In the description of this application, it should be noted that, unless otherwise defined, all technical and scientific terms used in the present invention have the same meaning as those commonly understood by those skilled in the art. The terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood by specific circumstances.

[0058] An embodiment of the present invention provides a collaborative resource scheduling method based on a programmable network. For details, see Figure 1, Figure 1 It is a schematic flow chart of a collaborative resource scheduling method based on a programmable network in one embodiment of the present invention.

[0059] The flowchart of the collaborative resource scheduling method based on programmable network includes steps S1 to S4:

[0060] S1. Obtain all computing resource data of a computer cluster, and perform performance modeling on heterogeneous computing resource data in all the computing resource data to obtain a unified resource performance index;

[0061] S2, defining each target computing resource data required by the multi-dimensional resource demand service, and determining matching relationship information between each multi-dimensional resource demand service and the current computer cluster based on the linear relationship between different target computing resource data in the current computer cluster;

[0062] S3. During the deployment process, according to all the matching relationship information, the corresponding selected multi-dimensional resource demand services are deployed to each of the computer clusters respectively; and during the deployment process, the shortest deployment path between each routing node in each of the computer clusters is determined based on the Floyd algorithm;

[0063] S4. During the actual scheduling process, the computing power scheduling plan generated by the deployment process is executed.

[0064] Obtain all computing resource data of the computer cluster, and measure the resource situation by using Prometheus and node Exporter. Computing resource data includes computing power resources, memory resources, etc. Memory resources are calculated and measured in a unified unit. Computing power resources form a computing power capacity template by making a unified abstract description of different computing types, which can provide standard computing power measurement rules for computing power routing, computing power equipment management, etc.

[0065] Existing computing modules usually adopt various different architectures and have different capabilities. By modeling heterogeneous computing resources in different computing devices, the heterogeneity of the underlying physical resources can be shielded from the service. In step S1, the performance modeling of heterogeneous computing resources can be performed to build a unified and measurable resource performance indicator, so as to uniformly identify the performance of different computing devices in various aspects.

[0066] In S2, each target computing resource data required by the multi-dimensional resource demand business is classified. The business basic resource demand types include the following four types: CPU resources, GPU resources, memory resources and IO resources. Based on the linear relationship between the different target computing resource data in the current computer cluster, this relationship is expressed as:

[0067] Score(cluster i )=α cpu X CPU +α gpu X GPU +α mem X Mem +α io X IO

[0068] Among them, Score(cluster i ) represents the matching relationship information between the multi-dimensional resource demand business and the current computer cluster, and its value range is [0,4], α x is the weight of different target computing resource data to all computing resource data. The x type includes CPU resources, GPU resources, memory resources and IO resources. Memory resources are expressed in Mem. x Indicates the matching relationship between different target computing resource data types and the current computer cluster of the multi-dimensional resource demand business;

[0069] The matching relationship between the different target computing resource data types and the current computer cluster of the multi-dimensional resource demand business is expressed as:

[0070]

[0071] Among them, cluster i (x) represents the weight coefficient of the x-type computing resource data of the ith computer cluster, x max Indicates the highest weight coefficient of type x resource in all current computer clusters.

[0072] Specifically, the multi-dimensional resource demand business is mainly used in scenarios to prevent power lines from being damaged, including one or more of equipment detection, abnormal alarm, damage analysis, and aging warning. According to the business application resources and cluster computing resources, the cluster that best meets the business resource requirements is automatically allocated. When allocating clusters, based on the type of computing resources required by the business, unified identification technologies such as performance modeling are combined to use allocation algorithms from the data collection module to obtain the cluster that best meets the current business needs in the current cross-cloud environment. By connecting the computing resource types required by the business and the existing computing resource conditions of the cluster, dynamic on-demand allocation can be achieved, thereby improving the overall performance of the system and improving system operation efficiency.

[0073] In the actual scheduling process, an SDN architecture is constructed, a network topology structure based on the multi-dimensional resources is established on the SDN architecture, and a computing power orchestration solution generated by the deployment process is executed based on SRV6.

[0074] The SDN architecture includes an SDN controller, a southbound interface, a soft routing device and an SDN control plane. The SDN control plane is located on the SDN controller. The SDN control plane implements routing management and path orchestration based on network programming and path management technology. The soft routing device is connected to the SDN controller via a southbound interface and is used to execute instructions issued by the SDN controller.

[0075] Specifically, in this embodiment, the controller communicates with the soft routing device through the REST API to instruct the device how to handle traffic. The REST API is a commonly used southbound interface protocol that can implement communication between the controller and network devices through standard HTTP requests.

[0076] The steps of executing the computing power orchestration scheme generated by the deployment process based on SRV6 include: (1) SID management, using SID as a network node identifier, generating a SID list, and dynamically allocating and adjusting the SID based on network status and traffic; (2) path calculation, using the Floyd algorithm to perform shortest path calculation to obtain a routing path; (3) routing control, when data is sent from a source node, combining the SID list with SRV6, and when the data reaches an intermediate node, parsing the content of the SRV6, identifying the SID that currently needs to be processed, and sending the processed SID to the next routing node corresponding to the SID to execute and complete the computing power orchestration scheme.

[0077] The SDN control plane is located on the SDN controller, which is responsible for the management and decision-making of the entire network. The control plane can monitor the network status from the perspective of the entire network and dynamically adjust the network configuration when necessary to achieve the purpose of global optimization.

[0078] The SRV6 controller combines network programming and path management technologies, and implements routing management and path orchestration based on the pyroute2 library. The controller uses pyroute2 to create, update, delete, and query the status of SRV6 paths, and communicates with the router through the REST API to implement dynamic path management. Path management is organized using a data dictionary to facilitate quick search and modification of path information. The controller monitors the network status in real time, dynamically adjusts the path to optimize traffic transmission, and can expand the application of SIDs to achieve flexible service programming and path optimization, such as combining Floyd, Dijkstra and other algorithms for optimal path selection. In this embodiment, the advantage of the Floyd algorithm is that it can simultaneously calculate the shortest path between all node pairs in a multi-source situation, which is suitable for the optimal path calculation scenario of the entire network.

[0079] An embodiment of the present invention provides a collaborative resource scheduling system based on a programmable network. For details, see Figure 2 , Figure 2 The figure is a schematic diagram of a structure block diagram of a collaborative resource scheduling system based on a programmable network in one embodiment of the present invention, comprising:

[0080] The resource perception module 11 is used to obtain all computing resource data of the computer cluster, and perform performance modeling on the heterogeneous computing resource data in all the computing resource data to obtain a unified resource performance index;

[0081] The cluster allocation module 12 is used to define each target computing resource data required by the multi-dimensional resource demand business, and determine the matching relationship information between each multi-dimensional resource demand business and the current computer cluster based on the linear relationship between different target computing resource data in the current computer cluster;

[0082] The cross-cloud routing control module 13 is used to deploy the corresponding selected multi-dimensional resource demand services to each of the computer clusters according to all the matching relationship information during the deployment process; and to determine the shortest deployment path between each routing node in each of the computer clusters based on the Floyd algorithm during the deployment process;

[0083] The cloud computing power orchestration module 14 is used to execute the computing power orchestration plan generated by the deployment process during the actual scheduling process.

[0084] Furthermore, in the above embodiment, the linear relationship between different target computing resource data in the computer cluster is expressed as:

[0085] Score(cluster i )=α cpu X CPU +α gpu X GPU +α mem X Mem +α io X IO

[0086] Among them, Score(cluster i ) represents the matching relationship information between the multi-dimensional resource demand business and the current computer cluster, and its value range is [0,4], α x is the weight of different target computing resource data to all computing resource data. The x type includes CPU resources, GPU resources, memory resources and IO resources. Memory resources are expressed in Mem. x Indicates the matching relationship between different target computing resource data types and the current computer cluster of the multi-dimensional resource demand business;

[0087] The matching relationship between the different target computing resource data types and the current computer cluster of the multi-dimensional resource demand business is expressed as:

[0088]

[0089] Among them, cluster i (x) represents the weight coefficient of the x-type computing resource data of the ith computer cluster, x max Indicates the highest weight coefficient of type x resource in all current computer clusters.

[0090] Furthermore, in the above embodiment, in the actual scheduling process, the computing power orchestration plan generated by the deployment process is executed, including building an SDN architecture, establishing a network topology structure based on the multi-dimensional resources on the SDN architecture, and using SRV6 to execute the computing power orchestration plan generated by the deployment process.

[0091] Furthermore, in the above embodiment, the SDN architecture includes an SDN controller, a southbound interface, a soft routing device and an SDN control plane; the SDN control plane is located on the SDN controller, and the SDN control plane implements routing management and path orchestration based on network programming and path management technology; the soft routing device is connected to the SDN controller through the southbound interface, and is used to execute instructions issued by the SDN controller.

[0092] Furthermore, in the above embodiment, the computing power orchestration scheme generated by the deployment process based on SRV6 includes: (1) SID management, using SID as a network node identifier to generate a SID list, and dynamically allocating and adjusting the SID based on network status and traffic; (2) path calculation, using the Floyd algorithm to perform shortest path calculation to obtain a routing path; (3) routing control, when data is sent from a source node, combining the SID list with SRV6, and when the data reaches an intermediate node, parsing the content of the SRV6, identifying the SID that currently needs to be processed, and sending the processed SID to the next routing node corresponding to the SID to execute and complete the computing power orchestration scheme.

[0093] Compared with the prior art, the embodiments of the present invention have the following advantages:

[0094] (1) The present invention provides a unified abstract description of different types of computing resources, and through refined modeling of computing performance, establishes a globally unique computing resource identification system, thereby solving the scheduling problem of diverse and highly heterogeneous computing resources in a complex environment;

[0095] (2) The present invention proposes a weight-based multi-dimensional resource scheduling method for multiple dimensions such as CPU, GPU, memory and IO resources. The system dynamically adjusts the weight of resources in each dimension according to the actual needs of the business, optimizes resource utilization through precise resource matching and allocation strategies, and ensures efficient computing power scheduling and system performance improvement;

[0096] (3) In a cross-cloud and cloud-edge environment, the present invention combines network topology information and computing node status, and uses the Floyd algorithm to calculate the shortest path between any pair of nodes in the network, thereby achieving efficient coordination and scheduling of cross-domain resources. Through SID expansion and dynamic path adjustment, it ensures that computing power requests can be transmitted to the best computing power node via the optimal path, thereby maximizing resource utilization and improving business processing performance.

[0097] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A collaborative resource scheduling method based on a programmable network, characterized in that: include: Acquire all computing resource data of the computer cluster, and perform performance modeling on heterogeneous computing resource data in all the computing resource data to obtain a unified resource performance index; Defining each target computing resource data required by the multi-dimensional resource demand service, and determining matching relationship information between each of the multi-dimensional resource demand services and the current computer cluster based on a linear relationship between different target computing resource data in the current computer cluster; During the deployment process, according to all the matching relationship information, the corresponding selected multi-dimensional resource demand services are deployed to each of the computer clusters respectively; and during the deployment process, the shortest deployment path between each routing node in each of the computer clusters is determined based on the Floyd algorithm; During the actual scheduling process, the computing power scheduling plan generated by the deployment process is executed.

2. The programmable network-based collaborative resource scheduling method according to claim 1, characterized in that: The linear relationship between different target computing resource data in the computer cluster is expressed as: Score(cluster) i )=a cpu X CPU +a gpu X GPU +a mem X Mem +a io X IO Among them, Score(cluster i ) represents the matching relationship information between the multi-dimensional resource demand business and the current computer cluster, and its value range is [0,4], α x is the weight of different target computing resource data to all computing resource data. The x type includes CPU resources, GPU resources, memory resources and IO resources. Memory resources are expressed in Mem. x Indicates the matching relationship between different target computing resource data types and the current computer cluster of the multi-dimensional resource demand business; The matching relationship between the different target computing resource data types and the current computer cluster of the multi-dimensional resource demand business is expressed as: Among them, cluster i (x) represents the weight coefficient of the x-type computing resource data of the ith computer cluster, x max Indicates the highest weight coefficient of type x resource in all current computer clusters.

3. The programmable network-based collaborative resource scheduling method according to claim 1, characterized in that: In the actual scheduling process, executing the computing power scheduling scheme generated by the deployment process includes: Construct an SDN architecture, establish a network topology structure based on the multi-dimensional resources on the SDN architecture, and adopt an SRV6-based computing power orchestration solution generated by the deployment process.

4. The programmable network-based collaborative resource scheduling method according to claim 3, characterized in that: The SDN architecture includes an SDN controller, a southbound interface, a soft routing device and an SDN control plane; The SDN control plane is located on the SDN controller, and the SDN control plane implements routing management and path orchestration based on network programming and path management technology; The soft routing device is connected to the SDN controller via a southbound interface, and is used to execute instructions issued by the SDN controller.

5. The programmable network-based collaborative resource scheduling method according to claim 3, characterized in that: The computing power orchestration scheme generated by the deployment process is executed based on SRV6 and includes: Using SID as a network node identifier, generating a SID list, and dynamically allocating and adjusting the SID according to network status and traffic; The Floyd algorithm is used to calculate the shortest path to obtain a routing path; When the data is sent from the source node, the SID list is combined with SRV6. When the data reaches the intermediate node from the routing path, the content of SRV6 is parsed, the adjusted SID that needs to be processed is identified, and the processed SID is sent to its corresponding next routing node to execute and complete the computing power orchestration plan.

6. A collaborative resource scheduling system based on a programmable network, characterized in that: include: A resource perception module is used to obtain all computing resource data of a computer cluster, and to perform performance modeling on heterogeneous computing resource data in all the computing resource data to obtain a unified resource performance index; A cluster allocation module, used to define each target computing resource data required by the multi-dimensional resource demand business, and determine the matching relationship information between each multi-dimensional resource demand business and the current computer cluster based on the linear relationship between different target computing resource data in the current computer cluster; A cross-cloud routing control module is used to deploy the corresponding selected multi-dimensional resource demand services to each of the computer clusters according to all the matching relationship information during the deployment process; and to determine the shortest deployment path between each routing node in each of the computer clusters based on the Floyd algorithm during the deployment process; The cloud computing power orchestration module is used to execute the computing power orchestration plan generated by the deployment process during the actual scheduling process.

7. The programmable network-based collaborative resource scheduling system according to claim 6, characterized in that: The linear relationship between different target computing resource data in the computer cluster is expressed as: Score(cluster) i )=a cpu X CPU +a gpu X GPU +a mem X Mem +a io X IO Among them, Score(cluster i ) represents the matching relationship information between the multi-dimensional resource demand business and the current computer cluster, and its value range is [0,4], α x is the weight of different target computing resource data to all computing resource data. The x type includes CPU resources, GPU resources, memory resources and IO resources. Memory resources are expressed in Mem. x Indicates the matching relationship between different target computing resource data types and the current computer cluster of the multi-dimensional resource demand business; The matching relationship between the different target computing resource data types and the current computer cluster of the multi-dimensional resource demand business is expressed as: Among them, cluster i (x) represents the weight coefficient of the x-type computing resource data of the ith computer cluster, x max Indicates the highest weight coefficient of type x resource in all current computer clusters.

8. The programmable network-based collaborative resource scheduling system according to claim 6, characterized in that: In the actual scheduling process, the computing power scheduling scheme generated by the deployment process is executed, including Construct an SDN architecture, establish a network topology structure based on the multi-dimensional resources on the SDN architecture, and adopt an SRV6-based computing power orchestration solution generated by the deployment process.

9. The programmable network-based collaborative resource scheduling system according to claim 8, characterized in that: The SDN architecture includes an SDN controller, a southbound interface, a soft routing device and an SDN control plane; The SDN control plane is located on the SDN controller, and the SDN control plane implements routing management and path orchestration based on network programming and path management technology; The soft routing device is connected to the SDN controller via a southbound interface, and is used to execute instructions issued by the SDN controller.

10. The programmable network-based collaborative resource scheduling system according to claim 8, characterized in that: The computing power orchestration scheme generated by the deployment process is executed based on SRV6 and includes: Using SID as a network node identifier, generating a SID list, and dynamically allocating and adjusting the SID according to network status and traffic; The Floyd algorithm is used to calculate the shortest path to obtain a routing path; When the data is sent from the source node, the SID list is combined with SRV6. When the data reaches the intermediate node from the routing path, the content of SRV6 is parsed, the adjusted SID that needs to be processed is identified, and the processed SID is sent to its corresponding next routing node to execute and complete the computing power orchestration plan.

Images