Network slicing and intra-chip scheduling methods and apparatus

Abstract

This application relates to a network slicing and intra-slice scheduling method and apparatus. The method includes: acquiring task information for multiple computing tasks; wherein the task information includes at least task type, data volume, computational load, and maximum tolerable end-to-end latency; classifying and sorting the multiple computing tasks according to the task information to obtain multiple sequentially arranged task sets; allocating network slices to each task set sequentially according to the task information and available resources of the edge computing system to obtain slicing results; wherein the slicing results include a source node, at least one destination node, and at least one transmission path for each network slice; and determining an intra-slice resource scheduling strategy for each network slice based on the task information and the slicing results. This method enables network slicing and intra-slice scheduling to address the diverse business characteristics and differentiated computing-network collaborative service requirements of smart grid applications.

Description

Technical Field

[0001] This application relates to the field of network technology, and in particular to a network slicing and intra-slice scheduling method and apparatus. Background Technology

[0002] With the increasing demand for communication in the power industry and the rapid development of 5G communication technology, a large number of smart grid applications have emerged, such as intelligent inspection, distribution network differential protection, and distribution automation. These applications have different requirements for computing and latency services. For example, control-related businesses such as precise load control, distribution network differential protection, and distribution automation require the decision-making center to make rapid decisions based on power grid field data and to issue remote control commands to distribution equipment, requiring high-performance computing and extremely low latency services. Drone inspections and robot inspections collect massive amounts of monitoring data, which also require rapid computing and processing to assist the monitoring and management center in making control decisions. Furthermore, precise load control and distribution network differential protection belong to the power production control area, located in power safety zones I and II, while drone inspections and robot inspections belong to the management information area, located in power safety zones III, IV, and the Internet area. Based on the power business security requirements, businesses in different security zones need physical or logical resource isolation.

[0003] Network slicing is a key technology in 5G. It uses virtualization to isolate a physical network into multiple isolated logical networks, providing customization capabilities to meet industry requirements. Furthermore, network slices can operate independently and have independent lifecycle management. However, over-configuring network and computing resources for network slices will lead to resource waste, while under-configuration will fail to meet the low-latency, high-performance computing requirements of power services.

[0004] Therefore, how to perform network slicing and intra-slice scheduling to address the diverse business characteristics and differentiated computing-network collaborative service needs of smart grid applications has become an urgent technical problem to be solved. Summary of the Invention

[0005] Therefore, it is necessary to provide a network slicing and intra-slice scheduling method and apparatus to address the above-mentioned technical problems, which can perform network slicing and intra-slice scheduling for the diverse business characteristics and differentiated computing network collaborative service requirements of smart grid applications.

[0006] In a first aspect, this application provides a network slicing and intra-slice scheduling method, the method comprising:

[0007] Obtain task information for multiple computing tasks; the task information shall include at least the task type, data volume, computational load, and maximum tolerable end-to-end latency.

[0008] Based on the task information, multiple computing tasks are classified and sorted to obtain multiple sequentially arranged task sets.

[0009] Based on the task information and the available resources of the edge computing system, network slices are allocated to each task set in sequence to obtain the slicing results; wherein, the slicing results include the source node, at least one destination node and at least one transmission path of each network slice;

[0010] Based on the task information and slicing results, determine the intra-slice resource scheduling strategy for each network slice.

[0011] In one embodiment, network slices are allocated sequentially to each task set based on task information and the available resources of the edge computing system to obtain slicing results, including:

[0012] Starting from the first position of the sorting of multiple task sets, for each task set, the wireless subnet access points corresponding to each computing task in the task set are classified and sorted to obtain multiple ordered sub-task sets.

[0013] Based on the task information and the available resources of the edge computing system, network slices are allocated to each subtask set in sequence to obtain the slice results.

[0014] In one embodiment, network slices are allocated sequentially to each subtask set based on task information and available resources of the edge computing system to obtain slicing results, including:

[0015] Based on task information and the available resources of the edge computing system, allocate network slices to the first subtask set among multiple subtask sets;

[0016] The available resources of the edge computing system are updated based on the allocated network slices, and network slices are allocated to the next set of subtasks based on the updated available resources and task information.

[0017] Return to the step of updating the available resources of the edge computing system based on the allocated network slices, until the allocation is complete and the slice results are obtained.

[0018] In one embodiment, a network slice is allocated to the first sub-task set among multiple sub-task sets based on task information and the available resources of the edge computing system, including:

[0019] For the first set of subtasks, a set of candidate computing nodes is allocated based on the available resources of the edge computing system; wherein, the set of candidate computing nodes includes multiple edge computing nodes that serve as destination nodes;

[0020] The candidate end-to-end transmission path set between each source-destination node pair is determined based on the candidate computing node set, as well as the equivalent bandwidth corresponding to the candidate end-to-end transmission path set; wherein, the source node in the source-destination node pair is the wireless subnet access point corresponding to the subtask set, and the transmission path in the candidate end-to-end transmission path set has the maximum equivalent bandwidth, and the equivalent bandwidth of the transmission path is determined based on the available bandwidth of each link in the transmission path.

[0021] The candidate network slice set is determined based on the candidate end-to-end transmission path set and the equivalent bandwidth of each transmission path;

[0022] Assign network slices to the subtask set based on task information, the candidate network slice set, and the maximum tolerable end-to-end latency of the subtask set.

[0023] In one embodiment, allocating network slices to the sub-task set based on task information, a candidate network slice set, and the maximum tolerable end-to-end latency of the sub-task set includes:

[0024] Sort the edge computing nodes in the candidate network slice set from shortest to longest according to the transmission path between the source and destination node pairs;

[0025] The first edge computing node after sorting is taken as the destination node, and the wireless subnet access point corresponding to the subtask set is taken as the source node. The computational and data volume of each computing task in the subtask set is determined according to the task information, and the candidate network resources and candidate computing resources corresponding to the destination node are used to determine the target latency corresponding to the subtask set. Among them, the candidate network resources include the equivalent bandwidth of the transmission path between the source and destination nodes, and the candidate computing resources include the computing resources that the destination node can allocate to the subtask set.

[0026] If the target latency corresponding to the subtask set is not greater than the maximum tolerable end-to-end latency corresponding to the subtask set, network slices are formed based on the candidate network resources and candidate computing resources corresponding to the destination node, and the network slices are allocated to the subtask set.

[0027] If the target latency corresponding to the subtask set is greater than the maximum tolerable end-to-end latency corresponding to the subtask set, starting from the first edge computing node, with at least two edge computing nodes as destination nodes and the wireless subnet access point corresponding to the subtask set as the source node, network slices are formed based on the computational and data volume of each computing task in the subtask set, the candidate network resources and candidate computing resources corresponding to at least two destination nodes, and the network slices are allocated to the subtask set.

[0028] In one embodiment, the intra-slice resource scheduling strategy for each network slice is determined based on task information and slicing results, including:

[0029] For each network slice corresponding to a set of subtasks in the slice result, the scheduling strategy for on-slice computing resources is determined based on the computational amount of each computing task in the set of subtasks.

[0030] The scheduling strategy for on-chip network resources is determined based on the amount of data in each computation task within the subtask set.

[0031] In one embodiment, the scheduling strategy for on-chip computing resources is determined based on the computational workload of each computing task in the subtask set, including:

[0032] When a network slice includes a destination node, on-slice computing resources are allocated to each computing task based on the computational workload of each task.

[0033] When a network slice includes multiple destination nodes, on-slice computing resources are allocated to each computing task based on the computing resources of each destination node and the computing workload of each computing task.

[0034] In one embodiment, the scheduling strategy for on-chip network resources is determined based on the data volume of each computing task in the subtask set, including:

[0035] When the same destination node within a network slice includes a transmission path, allocate on-slice network resources to each computing task based on the data volume of each computing task.

[0036] When a single destination node within a network slice includes multiple transmission paths, intra-slice network resources are allocated to each computing task based on the network resources of the transmission paths between the source and destination nodes and the data volume of each computing task.

[0037] In one embodiment, the method further includes:

[0038] The on-chip resource scheduling policy is sent to the software-defined network controller, so that the software-defined network controller generates network slicing instructions according to the on-chip resource scheduling policy and sends the network slicing instructions to each network transmission node and each edge computing node to execute the network slicing process.

[0039] The on-chip resource scheduling strategy is sent to multiple IoT terminals so that each IoT terminal can send computing tasks to the edge computing nodes on the chip for processing through the allocated network slices according to the on-chip resource scheduling strategy.

[0040] The on-chip resource scheduling strategy is sent to multiple network transmission nodes so that each network transmission node can schedule the on-chip computing tasks according to the on-chip resource scheduling strategy.

[0041] The on-chip resource scheduling strategy is sent to multiple edge computing nodes so that each edge computing node can process the on-chip computing tasks according to the on-chip resource scheduling strategy and return the task results to the corresponding IoT terminal.

[0042] Secondly, this application also provides a network slicing and intra-slice scheduling apparatus, comprising:

[0043] The information acquisition module is used to acquire task information for multiple computing tasks; the task information includes at least the task type, data volume, computational volume, and maximum tolerable end-to-end latency.

[0044] The task classification module is used to classify and sort multiple computing tasks based on task information, resulting in multiple ordered task sets.

[0045] The network slicing module is used to allocate network slices to each task set in sequence according to task information and available resources of the edge computing system, and obtain slicing results; wherein, the slicing results include the source node, at least one destination node and at least one transmission path of each network slice;

[0046] The strategy determination module is used to determine the intra-slice resource scheduling strategy for each network slice based on task information and slicing results.

[0047] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method as described in any one of the first aspects.

[0048] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method as described in any one of the first aspects.

[0049] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the method as described in any one of the first aspects.

[0050] The aforementioned network slicing and intra-slice scheduling method and apparatus involve an edge computing controller acquiring task information from multiple computing tasks; classifying and sorting these tasks based on the task information to obtain multiple sequentially arranged task sets; allocating network slices to each task set sequentially based on the task information and available resources of the edge computing system, resulting in slicing results; and determining the intra-slice resource scheduling strategy for each network slice based on the task information and slicing results. By allocating network slices based on the task type and tolerable end-to-end latency of smart grid applications, the edge computing controller can not only provide isolated, latency-based differential edge computing services for different types of smart grid applications, but also provide diversified services for massive tasks with limited network slices. This approach avoids increasing the complexity of network management while providing finely differentiated latency protection services for smart grid applications. Attached Figure Description

[0051] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0052] Figure 1 This is a diagram illustrating the application environment of network slicing and intra-slice scheduling methods in one embodiment;

[0053] Figure 2 This is a flowchart illustrating the network slicing and intra-slice scheduling method in one embodiment;

[0054] Figure 3 This is a flowchart illustrating the steps of sequentially allocating network slices to each task set in one embodiment.

[0055] Figure 4 This is a flowchart illustrating the steps of allocating network slices to each set of subtasks in one embodiment.

[0056] Figure 5 This is a flowchart illustrating the step of allocating network slices to the first set of subtasks in one embodiment;

[0057] Figure 6 This is a flowchart illustrating the steps of allocating network slices to a set of subtasks in one embodiment.

[0058] Figure 7 This is a flowchart illustrating the network slicing and intra-slice scheduling method in another embodiment;

[0059] Figure 8 This is a schematic diagram of the transmission path in one embodiment;

[0060] Figure 9 This is a structural block diagram of network slicing and on-chip scheduling in one embodiment;

[0061] Figure 10 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0062] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0063] It should be noted that the terms "first," "second," etc., used in this application can be used to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from the second element. The terms "comprising" and "having," and any variations thereof, used in this application, are intended to cover non-exclusive inclusion. The term "multiple" used in this application refers to two or more. The term "and / or" used in this application refers to one of the embodiments, or any combination of multiple embodiments.

[0064] Network slicing is a key technology in 5G. It uses virtualization to isolate a physical network into multiple isolated logical networks, providing customization capabilities to meet industry requirements. Furthermore, network slices can operate independently and have independent lifecycle management. However, over-configuring network and computing resources for network slices will lead to resource waste, while under-configuration will fail to meet the low-latency, high-performance computing requirements of power services.

[0065] To address the aforementioned issues, this application provides a network slicing and intra-slice scheduling method that can meet the diverse business characteristics and differentiated computing-network collaborative service needs of smart grid applications.

[0066] The network slicing and intra-slice scheduling methods provided in this application can be applied to, for example... Figure 1 The application environment shown includes multiple IoT terminals 101, multiple wireless subnet access points 102, an edge computing system 103, and a Software-Defined Networking (SDN) controller 104. The IoT terminals 101 can communicate with the wireless subnet access points 102, and the edge computing system 103 can communicate with both the wireless subnet access points 102 and the SDN controller 104. The IoT terminals 101 may include, but are not limited to, various power inspection robots, FTUs (Feeder Terminal Units), DTUs (Distribution Terminal Units), and TTUs (Transformer Terminal Units). The wireless subnet access points 102 may include, but are not limited to, various wireless gateways and base stations. The edge computing system 103 includes an edge computing controller and multiple edge computing nodes. The edge computing controller and edge computing nodes may be independent physical servers or server clusters composed of multiple physical servers. In some embodiments, the application environment may also include multiple network transmission nodes, such as routers and gateways. Wireless subnet access point 102 can transmit data to edge computing nodes through network transmission nodes.

[0067] In one exemplary embodiment, such as Figure 2 As shown, a network slicing and intra-slice scheduling method is provided, which is applied to... Figure 1 Taking the edge computing controller in the example, the process can include the following steps:

[0068] Step 201: Obtain task information for multiple computing tasks.

[0069] The task information includes at least the task type, data volume, computational load, and maximum tolerable end-to-end latency; data volume is measured in bits or bytes, and computational load is measured in computational resource requirements. The task information may also include the terminal identifier of the IoT terminal. For example, multiple task information items may include { , , ..., , ..., , ...}, the task type is Data volume is The computational workload is The maximum tolerable end-to-end delay is and terminal identifier as Where n represents the identifier of the wireless subnet (wireless subnet access point) where the IoT terminal is located, and k represents the identifier of the IoT terminal or computing task in the wireless subnet.

[0070] IoT terminal collection consisting of multiple IoT terminals The system receives computing tasks submitted by smart grid users (i.e., the application layer) and sends the task information to the edge computing controller. The edge computing controller can then obtain task information for multiple computing tasks.

[0071] Step 202: Classify and sort multiple computing tasks according to task information to obtain multiple ordered task sets.

[0072] Multiple computing tasks are classified according to task type and maximum tolerable end-to-end latency to obtain multiple task sets, and the multiple task sets are sorted in descending order of maximum tolerable end-to-end latency level.

[0073] Each task set has the same task type and the same maximum tolerable end-to-end latency level, where a higher latency level corresponds to a lower tolerable end-to-end latency. The task set at the head of the queue among multiple task sets has the highest latency level. Multiple task sets can be represented as { , , ..., , ..., Let the natural number M represent the number of categories in the multiple task sets, then Let represent the corresponding set, and let Let represent the end-to-end latency of the m-th task set. Then, the m-th task set can be represented as ( , ,{ , ..., ...}), that is, the set of tasks of this type has the same task type. The same maximum tolerable end-to-end delay level, i.e., satisfying .

[0074] Step 203: Based on the task information and the available resources of the edge computing system, network slices are allocated to each task set in sequence to obtain the slice results.

[0075] The available resources of the edge computing system include the task types and available computing resources supported by each edge computing node, the equivalent bandwidth of the transmission path, etc. The slicing result includes the source node, at least one destination node, and at least one transmission path for each network slice. The source node is the wireless subnet access point that generates the computing task, and the destination node is the edge computing node.

[0076] For multiple task sets, starting from the head of the queue, network slices are allocated sequentially to each task set based on a delay-aware slicing strategy. This delay-aware slicing strategy assigns computational tasks from the same wireless subnet within the same type of task to the same network slice. A network slice contains one source node, one or more destination nodes, and one or more transmission paths.

[0077] Step 204: Based on the task information and slicing results, determine the intra-slice resource scheduling strategy for each network slice.

[0078] Based on the computational load, data volume, and maximum tolerable end-to-end latency of each computation task in the task set, as well as the source node, destination node, and transmission path of the network slice allocated to the task set, the computational resources and network resources in the network slice are allocated to each computation task.

[0079] For example, the decision results of network slicing and on-chip resource scheduling strategies can be expressed as:

[0080] Assigned to computing tasks Network slicing is The on-chip resource scheduling strategy for computational tasks within a network slice is as follows: Where n represents the identifier of the wireless subnet where the IoT terminal is located, k represents the identifier of the computing task in the wireless subnet, and T K The 'i' represents the task type, 'i' represents the number of network slices corresponding to multiple task sets, and 'j' represents the number of network slices corresponding to multiple sub-task sets within each task set.

[0081] In the above embodiments, the edge computing controller, based on the task type of the smart grid application and the method of allocating network slices with tolerable end-to-end latency, can not only provide isolated, latency-based differential edge computing services for different types of smart grid applications, but also provide diversified services for massive tasks through limited network slices. This does not increase the complexity of network management, and can provide finely differentiated latency protection services for smart grid applications.

[0082] In one exemplary embodiment, such as Figure 3 As shown, in the above embodiment, "allocating network slices to each task set sequentially based on task information and available resources of the edge computing system to obtain slice results" may include the following steps:

[0083] Step 301: Starting from the first position of the sorting of multiple task sets, for each task set, classify and sort it according to the wireless subnet access point corresponding to each computing task in the task set to obtain multiple ordered sub-task sets.

[0084] After sorting the tasks in descending order of maximum tolerable end-to-end latency, starting with the first task set, we first classify and sort them, then assign network slices to each sorted sub-task set. Next, we classify and sort the second task set, then assign network slices to each sorted sub-task set. This process is repeated for each task set, resulting in a sliced ​​network.

[0085] The classification process includes: classifying tasks according to the access points of the wireless subnets corresponding to the computation tasks, i.e., computation tasks within the same sub-task set originate from the same wireless subnet, and forming the source node set of the sub-task set with the access points of the wireless subnets corresponding to the sub-task set as the source nodes. The set of multiple subtasks obtained from classification is represented as ( , , , { , , ..., }), ( , , , { , , ..., }).

[0086] The sorting process includes: for each subtask set, summing the computational costs of multiple computational tasks in the subtask set to obtain the total computational cost of the subtask set; and sorting the multiple subtask sets in descending order of total computational cost.

[0087] Step 302: Based on the task information and the available resources of the edge computing system, network slices are allocated to each subtask set in sequence to obtain the slice results.

[0088] For each subtask set, computing resources and network resources in the network slice are allocated to each computing task based on the computational amount, data amount, and maximum tolerable end-to-end latency of each computing task in the subtask set, as well as the source node, destination node, and transmission path of the network slice allocated to the subtask set.

[0089] In the above embodiments, the method of coordinating and configuring network slices on demand with the constraint of satisfying the maximum tolerable end-to-end latency can avoid both the waste of resources caused by over-configuration of network and computing resources and the failure to meet the latency requirements of tasks due to insufficient resource configuration.

[0090] In one exemplary embodiment, such as Figure 4 As shown, in the above embodiment, "allocating network slices to each sub-task set sequentially based on task information and available resources of the edge computing system to obtain slice results" may include the following steps:

[0091] Step 401: Assign a network slice to the first subtask set among multiple subtask sets based on task information and available resources of the edge computing system.

[0092] Starting with the first subtask set after sorting, computing resources and network resources in the network slice are allocated to each computing task based on the computational amount, data amount, and maximum tolerable end-to-end latency of each computing task in the first subtask set, as well as the source node, destination node, and transmission path of the network slice allocated to the first subtask set.

[0093] Step 402: Update the available resources of the edge computing system according to the allocated network slices, and allocate network slices for the next set of subtasks according to the updated available resources and task information.

[0094] The updated available resources are obtained by subtracting the network and computing resources allocated to the first subtask set from the available resources of the edge computing system.

[0095] For example, let Indicates the transmission path One of the links in, This indicates the available bandwidth of the link. This represents the bandwidth allocated to the nth sub-task set within the m-th task set. After updating the available resources, the available bandwidth of this link is represented as... Similarly, let Indicates the destination node The computing resources allocated to this set of subtasks. If the destination node has sufficient computing resources to support the set of subtasks, then after updating the available resources, the computing resources can be represented as follows: Where l represents the identifier of the destination node, T K Indicates the task type.

[0096] After updating the available resources, the computing resources and network resources in the network slice are allocated to each computing task based on the computational amount, data amount, and maximum tolerable end-to-end latency of each computing task in the second subtask set, as well as the source node, destination node, and transmission path of the network slice allocated to the second subtask set.

[0097] Step 403: Return to the step of updating the available resources of the edge computing system based on the allocated network slices, until the allocation is completed and the slice results are obtained.

[0098] After allocating network slices for the second set of subtasks, update the available resources as described in the example above, and then allocate network slices for the third set of subtasks. Continue in this manner until all slices are allocated and the desired results are obtained.

[0099] In the above embodiments, network slices are allocated according to task type and maximum tolerable end-to-end latency. This allows for the provision of diverse services for a large number of tasks with limited network slices, without increasing the complexity of network management, while providing refined latency protection services for smart grid applications.

[0100] In one exemplary embodiment, such as Figure 5 As shown, in the above embodiment, "allocating a network slice to the first sub-task set among multiple sub-task sets based on task information and available resources of the edge computing system" may include the following steps:

[0101] Step 501: For the first set of subtasks, allocate a set of candidate computing nodes based on the available resources of the edge computing system.

[0102] The candidate computing node set includes multiple edge computing nodes that serve as destination nodes.

[0103] For the first subtask set, edge computing nodes in the edge computing system that support the task types of that subtask set and have available computing resources are assigned to the candidate computing node set for that subtask set. For example, let... Indicates the task type that supports the first sub-task set in the edge computing system. And there is a set of candidate computing nodes with available computing resources. This represents the available computing resources of the candidate computing node set, where, This indicates the edge computing nodes that support this set of subtasks. Available computing resources.

[0104] Step 502: Determine the set of candidate end-to-end transmission paths between each source-destination node pair and the equivalent bandwidth corresponding to the set of candidate end-to-end transmission paths based on the set of candidate computing nodes.

[0105] Among them, source and destination node pairs This includes a source node s and a destination node d. The source node s is the wireless subnet access point corresponding to the subtask set, and the destination node d is the edge computing node in the candidate computing node set. The transmission path in the candidate end-to-end transmission path set has the maximum equivalent bandwidth, which is determined based on the available bandwidth of each link in the transmission path.

[0106] For this set of subtasks, multiple source-destination node pairs can be determined based on the candidate computation node set and the source node. And determine each source-destination node pair The set of candidate end-to-end transmission paths between and the equivalent bandwidth corresponding to this set. Assume the tasks in the subtask set originate from the wireless subnet access point. Then the source node The destination node d is the set of candidate computing nodes. Edge computing nodes in, i.e. Candidate end-to-end transmission path set Transmission path in , ... It is the path with the maximum equivalent bandwidth between the source and destination node pairs, i.e. = =…= = and That is, the equivalent bandwidth of the transmission path within the candidate end-to-end transmission path set is greater than or equal to the equivalent bandwidth of the transmission path outside the candidate end-to-end transmission path set.

[0107] make Indicates the transmission path Among them, l1, l2...l k Let s be the network transmission nodes in this transmission path. s to l1 is one link, l1 to l2 is another link, and the reciprocal of the equivalent bandwidth of the transmission path is the sum of the reciprocals of the available bandwidths of each link in the transmission path, as shown in the following formula:

[0108]

[0109] in, This represents the transmission path between the source node s and the destination node d. The equivalent bandwidth, , , , This indicates the available bandwidth of each link in the transmission path.

[0110] Step 503: Determine the candidate network slice set based on the candidate end-to-end transmission path set and the equivalent bandwidth of each transmission path.

[0111] Assume that the set of subtasks comes from the source node. That is, the nth wireless subnet, and the service type is ,make , represents a destination node of the subtask set, let Indicates source and destination node pair ( The set of candidate end-to-end transmission paths, This represents the equivalent bandwidth corresponding to the set of candidate end-to-end transmission paths. This represents the maximum computing resources that destination node d can be allocated to this set of subtasks. Denotes a set of subtasks, and the set of candidate network slices for that set of subtasks. It can be represented as ,in, It is a computational task Task information.

[0112] Step 504: Assign network slices to the subtask set based on the task information, the candidate network slice set, and the maximum tolerable end-to-end latency of the subtask set.

[0113] Based on the computational load, data volume, and maximum tolerable end-to-end latency of each computational task in the subtask set, and based on the source node, destination node, and transmission path of the network slices allocated to the subtask set in the candidate network slice set, candidate computational resources and candidate network resources corresponding to the candidate network slice set are determined. If the candidate computational resources and candidate network resources meet the maximum tolerable end-to-end latency requirement, then computational resources and network resources from the network slices are allocated to each computational task according to the candidate network slice set.

[0114] Understandably, after allocating network slices to the first set of subtasks, each subsequent set of subtasks can be allocated network slices in accordance with the example above.

[0115] In the above embodiments, by allocating network slices according to the maximum tolerable end-to-end delay, a refined delay guarantee service can be provided for smart grid applications, avoiding the inability to meet the delay requirements of tasks due to insufficient resource allocation.

[0116] In one exemplary embodiment, such as Figure 6 As shown, in the above embodiment, "allocating network slices to the sub-task set based on the task information, the candidate network slice set, and the maximum tolerable end-to-end latency of the sub-task set" may include the following steps:

[0117] Step 601: Sort the edge computing nodes in the candidate network slice set from shortest to longest according to the transmission path between the source and destination node pairs.

[0118] Based on the candidate network slice set The edge computing nodes in the process determine the source-destination node pair and the transmission path between the source-destination node pair; according to the transmission path from shortest to longest, the edge computing node corresponding to the shortest transmission path is placed at the head of the queue.

[0119] For example, let m represent the current task set and n represent the current subtask set within the current task set, and let... Represents the set of candidate network slices The system can support a set of candidate computing nodes for this subtask set. This represents the number of hops in the transmission path between the source and destination node pairs corresponding to each edge computing node. Assume... If the above is true, then the multiple edge computing nodes in the candidate computing node set are arranged in ascending order of transmission path length, resulting in the following arrangement: That is, edge computing nodes It is located at the head of the candidate computing node set.

[0120] Step 602: Using the first edge computing node after sorting as the destination node and the wireless subnet access point corresponding to the subtask set as the source node, determine the computational and data volume of each computing task in the subtask set based on the task information, and determine the target latency corresponding to the subtask set based on the candidate network resources and candidate computing resources corresponding to the destination node.

[0121] Among them, candidate network resources include the equivalent bandwidth of the transmission path between source and destination node pairs, and candidate computing resources include the computing resources that the destination node can allocate to the set of subtasks.

[0122] Starting with the first edge computing node in the candidate computing node set, assuming the destination node is an edge computing node. ,make This indicates the total amount of data in the set of subtasks. This represents the total computational cost of the set of subtasks, where, , Let z represent the data volume and computational volume of the i-th computational task in the subtask set, respectively, and let z be the number of computational tasks in the subtask set. This indicates the source node corresponding to the set of subtasks. to destination node The equivalent bandwidth, This represents the maximum tolerable end-to-end latency for this set of subtasks. Indicates the destination node The computing resources that can be allocated to the set of subtasks are determined based on the total data volume and total computational cost of the subtask set, and the source node. to destination node Equivalent bandwidth, destination node Allocable computing resources determine target latency .

[0123] Step 603: If the target latency corresponding to the subtask set is not greater than the maximum tolerable end-to-end latency corresponding to the subtask set, form a network slice based on the candidate network resources and candidate computing resources corresponding to the destination node, and allocate the network slice to the subtask set.

[0124] if If true, then edge computing nodes are used. For the destination node, the wireless subnet access point corresponding to the set of subtasks As the source node, based on the source node to destination node Equivalent bandwidth, destination node Allocable computing resources form network slices And assign network slices to a set of subtasks. This is the transmission path of the network slice. It is the equivalent bandwidth of the transmission path allocated to this set of subtasks. It is the destination node The computing resources allocated to this set of subtasks. It is a collection of subtasks that share resources within the film.

[0125] The equivalent bandwidth allocated to this set of subtasks by the transmission path is the candidate network resource, and the destination node The computing resources allocated to this set of subtasks are candidate computing resources. The combination of candidate network resources and candidate computing resources is the minimum combination of network resources and computing resources that can satisfy the maximum tolerable end-to-end latency requirement of the set of subtasks, that is, , satisfy: and , , , , among which, natural numbers In other words, reducing the number of candidate network resources or candidate computational resources by a random number will not satisfy the maximum tolerable end-to-end latency requirement of the subtask set.

[0126] Step 604: If the target latency corresponding to the subtask set is greater than the maximum tolerable end-to-end latency corresponding to the subtask set, starting from the first edge computing node, with at least two edge computing nodes as destination nodes and the wireless subnet access point corresponding to the subtask set as the source node, network slices are formed based on the computational and data volume of each computing task in the subtask set, the candidate network resources and candidate computing resources corresponding to at least two destination nodes, and the network slices are allocated to the subtask set.

[0127] Starting with the first edge computing node in the candidate computing node set, at least two edge computing nodes are extracted as destination nodes. The wireless subnet access point corresponding to the subtask set is used as the source node. This yields candidate network resources and candidate computing resources between the source and multiple destination nodes. Candidate network resources include the equivalent bandwidth of the transmission path between each source-destination node pair, and candidate computing resources include the computing resources allocated to the subtask set by each destination node. The combination of candidate network resources and candidate computing resources is the minimum combination among multiple destination sequential combination methods that can satisfy the maximum tolerable end-to-end latency requirement of the subtask set. That is, when any destination node is removed, the combination of candidate network resources and candidate computing resources cannot satisfy the maximum tolerable end-to-end latency requirement of the subtask set.

[0128] For example, let Let represent the destination node to be extracted. , These represent transmissions to the destination node. , The amount of data, , These represent the required destination nodes. , The computational load of the processing, among which, , , , These represent the source nodes corresponding to the set of subtasks. to destination node , The equivalent bandwidth, , Representing the destination node respectively , Computational resources available for allocation to this set of subtasks. If the equivalent bandwidth... , and computing resources , The combination of these parameters can satisfy the maximum tolerable end-to-end latency requirement of this set of subtasks, i.e. and Furthermore, when any destination node is reduced, the combination of equivalent bandwidth and computing resources cannot meet the maximum tolerable end-to-end latency requirement of the subtask set, i.e. and Then it can be based on the source node Destination node , Candidate end-to-end path set , The equivalent bandwidth allocated to the set of candidate end-to-end paths for this set of subtasks. , Destination node , Computational resources allocated to this set of subtasks , Determine a network slice, which can be represented as: .

[0129] In the above embodiments, the computing nodes for network slices are preferentially selected from edge computing nodes that are closest to the data source and can meet the computing power requirements of the task. This can reduce network load and lower the network transmission latency of computing tasks.

[0130] In an exemplary embodiment, the above embodiment of "determining the intra-slice resource scheduling strategy for each network slice based on task information and slicing results" includes: for the network slice corresponding to each sub-task set in the slicing results, determining the intra-slice computing resource scheduling strategy based on the computing volume of each computing task in the sub-task set; and determining the intra-slice network resource scheduling strategy based on the data volume of each computing task in the sub-task set.

[0131] For each network slice corresponding to a set of subtasks, on-slice computing resources within the network slice can be allocated to each computing task based on the computational workload of each task within the subtask set. For example, if the computational workload of each computing task is... Then the computing resources of the destination node will be allocated to each computing task according to the amount of computing required.

[0132] For each network slice corresponding to a subtask set, the on-slice network resources within the network slice can be allocated to each computational task based on the data volume of each computational task in the subtask set. For example, if the data volume of each computational task in the subtask set is... Then the equivalent bandwidth of the transmission path between the source and destination nodes is allocated to each computing task according to the amount of data.

[0133] In the above embodiments, the on-chip resource scheduling strategy fairly schedules on-chip tasks based on the amount of computation, the amount of data, the resource size of edge computing nodes, and the equivalent bandwidth of the transmission path, which can improve the fairness of on-chip task services with the same task type and the same end-to-end latency level.

[0134] In an exemplary embodiment, the "determining the scheduling strategy of on-chip computing resources based on the computational workload of each computing task in the subtask set" in the above embodiment may include the following:

[0135] 1) When a network slice includes a destination node, allocate on-slice computing resources to each computing task based on the computational workload of each computing task.

[0136] When a network slice includes a destination node, the allocation weight of each computing task is determined based on the computational workload of each computing task, and intra-slice computing resources are allocated to each computing task according to the allocation weight of each computing task and the computing resources of the destination node.

[0137] For example, suppose the network slice is The network slice contains a destination node. Subtask set The computational workload of each computational task in the process is For the k-th ( There are ) computational tasks, and the weights are allocated according to the computational workload. The computing resources allocated to the k-th computational task for this network slice are: The scheduling strategy for on-chip computing resources can be expressed as: Where n is the identifier of the wireless subnet access point, and z is the number of computation tasks.

[0138] 2) When a network slice includes multiple destination nodes, allocate on-slice computing resources to each computing task based on the computing resources of each destination node and the computing workload of each computing task.

[0139] When a network slice includes multiple destination nodes, the computational tasks in the subtask set are allocated to the destination nodes for computation based on the computational resource weights of those destination nodes. The computational resource weight is the ratio of the computational resources of the destination node to the sum of the computational resources of all destination nodes within the slice. Then, computational tasks with the same destination node are assigned weights based on their computational complexity, and the on-slice computational resources of the same destination node are allocated to these tasks according to their assigned weights.

[0140] Network slicing is A network slice contains more than one destination node. , Subtask set The computational tasks are assigned to the destination node. , The computational complexity is respectively , ,in, , Destination nodes , The computing resources allocated to this set of subtasks. , Destination nodes , The weight of computational resources, This represents the total computational cost allocated to this network slice. Based on the task allocation strategy described above, it is assumed that the task is allocated to the destination node. The set of subtasks is The computational workload for each computational task is Assigned to the destination node The set of subtasks is The computational workload for each computational task is Then the destination node Assigned to task ( The computing resources are Destination node Assigned to task ( The computing resources are The scheduling strategy for on-chip computing resources can be expressed as: .

[0141] In the above embodiments, the on-chip resource scheduling strategy fairly schedules on-chip tasks based on the amount of computing power and the resource size of edge computing nodes, which can improve the fairness of on-chip computing services with the same task type and the same end-to-end latency level.

[0142] In an exemplary embodiment, the "determining the scheduling strategy of on-chip network resources based on the data volume of each computing task in the subtask set" in the above embodiment may include the following:

[0143] 1) When the same destination node within a network slice includes a transmission path, allocate on-slice network resources to each computing task based on the data volume of each computing task.

[0144] When the same destination node within a network slice includes a transmission path, the allocation weight of each computing task is determined based on the data volume of each computing task, and the intra-slice network resources are allocated to each computing task according to the allocation weight of each computing task and the equivalent bandwidth of the transmission path.

[0145] For example, let Represents network slices Internal source node to destination node The set of transmission paths, assuming and That is, the transmission path The link is l1, l2...l k As a network transmission node, let Indicate the equivalent bandwidth of the transmission path, let Indicates allocation to the destination node The set of subtasks This represents the amount of data in each computation task within the subtask set. The allocation weight for each computation task is determined based on its data volume. Then the transmission path Assigned to task ( The equivalent bandwidth is The scheduling strategy for its on-chip network resources can be expressed as follows: Where h is the destination node. The number of tasks to be calculated.

[0146] 2) When the same destination node within a network slice includes multiple transmission paths, allocate intra-slice network resources to each computing task based on the network resources of the transmission paths between the source and destination nodes and the data volume of each computing task.

[0147] When a single destination node within a network slice includes multiple transmission paths, the computational tasks assigned to that destination node within the subtask set are split into subtask sets according to the weights of each transmission path between the source and destination node pairs, and transmitted simultaneously across multiple paths. The weight of a transmission path is the ratio of its equivalent bandwidth to the sum of the equivalent bandwidths of all transmission paths between the source and destination node pairs within the slice. Then, the allocation weight is determined based on the data volume of the computational tasks along the same transmission path, and finally, the on-chip bandwidth resources for that transmission path are allocated to each computational task according to the allocation weight.

[0148] Assume the set of transmission paths to the same destination node is and , That is, the set of transmission paths contains 2 transmission paths. , Transmission path The link is Transmission path The link is ,make , Representing paths , The equivalent bandwidth is then split across the transmission path. The amount of data transmitted is Split into transmission path The amount of data transmitted is ,in, , These are the transmission paths , The weight, From the source node Unload to the destination node The amount of data. Based on the task splitting strategy described above, assuming it is allocated to the transmission path... The set of subtasks is ( Then the transmission path Assigned to this set of subtasks ( The equivalent bandwidth is Assigned to transmission path The set of subtasks is Then the transmission path Assign computational tasks to this set of subtasks ( The equivalent bandwidth is Its on-chip computing resource scheduling strategy can be expressed as: i is the path assigned to the transmission path The identifier of the computation task, where y is the path to the transmission path. The number of computational tasks, where j is the number of tasks assigned to the transmission path. The identifier of the computation task, h is the number of computation tasks assigned to the same destination node.

[0149] The intra-chip resource scheduling strategy for network slicing includes the intra-chip compute resource scheduling strategy and the intra-chip network resource scheduling strategy. The intra-chip resource scheduling strategy can be expressed as follows: .

[0150] In the above embodiments, the on-chip resource scheduling strategy fairly schedules on-chip tasks based on the data volume and the equivalent bandwidth of the transmission path, which can improve the fairness of on-chip transmission services with the same task type and the same end-to-end latency level.

[0151] In one exemplary embodiment, such as Figure 7 As shown, embodiments of this application may further include the following steps:

[0152] Step 205: The on-chip resource scheduling policy is sent to the software-defined network controller, so that the software-defined network controller generates network slicing instructions according to the on-chip resource scheduling policy and sends the network slicing instructions to each network transmission node and each edge computing node to execute the network slicing process.

[0153] Step 206: Send the on-chip resource scheduling strategy to multiple IoT terminals so that each IoT terminal can send computing tasks to the edge computing nodes on the chip for processing through the allocated network slices according to the on-chip resource scheduling strategy.

[0154] Step 207: Send the on-chip resource scheduling policy to multiple network transmission nodes so that each network transmission node can schedule the on-chip computing tasks according to the on-chip resource scheduling policy.

[0155] Step 208: The on-chip resource scheduling strategy is sent to multiple edge computing nodes so that each edge computing node can process the on-chip computing tasks according to the on-chip resource scheduling strategy and return the task results to the corresponding IoT terminal.

[0156] After determining the on-chip resource scheduling strategy, the edge computing controller sends the strategy to the IoT terminal, network transmission node, edge computing node, and software-defined network controller. The software-defined network controller then generates network slicing instructions based on the on-chip resource scheduling strategy and sends these instructions to the network transmission node and edge computing node.

[0157] The IoT terminal sends computing tasks to the network transmission node according to the on-chip resource scheduling strategy. The network transmission node then transmits the computing tasks to the next network transmission node or edge computing node according to the network slicing instructions and the on-chip resource scheduling strategy.

[0158] Edge computing nodes receive computing tasks, allocate corresponding computing resources to the tasks according to on-chip resource scheduling and network slicing instructions, and process the tasks using these resources. Afterward, the edge computing nodes transmit the task results back to the IoT terminal via a transmission path.

[0159] In the above embodiments, the edge computing controller works in collaboration with IoT terminals, network transmission nodes, edge computing nodes, and software-defined network controllers to provide isolated, latency-based differential edge computing services for different types of smart grid applications.

[0160] In one exemplary embodiment, a network slicing and intra-slice scheduling method is provided, which is applied to... Figure 1 Taking the edge computing controller in the example, the process can include the following steps:

[0161] Step 1: Obtain task information for multiple computing tasks.

[0162] The task information includes at least the task type, data volume, computational load, and maximum tolerable end-to-end latency.

[0163] Step 2: Classify and sort multiple computing tasks according to the task information to obtain multiple ordered task sets.

[0164] Step 3: Starting from the first position of the sorting of multiple task sets, for each task set, classify and sort them according to the wireless subnet access point corresponding to each computation task in the task set to obtain multiple ordered sub-task sets.

[0165] Step 4: For the first set of subtasks, allocate a set of candidate computing nodes based on the available resources of the edge computing system.

[0166] Step 5: Determine the set of candidate end-to-end transmission paths between each source-destination node pair and the equivalent bandwidth corresponding to the set of candidate end-to-end transmission paths based on the set of candidate computing nodes.

[0167] Step 6: Determine the candidate network slice set based on the candidate end-to-end transmission path set and the equivalent bandwidth of each transmission path.

[0168] Step 7: Sort the edge computing nodes in the candidate network slice set from shortest to longest according to the transmission path between the source and destination node pairs.

[0169] Step 8: Using the first sorted edge computing node as the destination node and the wireless subnet access point corresponding to the subtask set as the source node, determine the total computation and total data volume corresponding to the subtask set based on the task information, and determine the target latency corresponding to the subtask set based on the total computation, total data volume, candidate network resources and candidate computing resources corresponding to the edge computing node.

[0170] Step 9: If the target latency corresponding to the subtask set is not greater than the maximum tolerable end-to-end latency corresponding to the subtask set, form a network slice based on the candidate network resources and candidate computing resources corresponding to the edge computing node, and allocate the network slice to the subtask set.

[0171] Step 10: If the target latency corresponding to the subtask set is greater than the maximum tolerable end-to-end latency corresponding to the subtask set, starting from the first edge computing node, with at least two edge computing nodes as destination nodes and the wireless subnet access point corresponding to the subtask set as the source node, network slices are formed based on the total computing volume, total data volume, candidate network resources and candidate computing resources corresponding to at least two edge computing nodes, and the network slices are allocated to the subtask set.

[0172] Step 11: Update the available resources of the edge computing system according to the allocated network slices, and allocate network slices for the next set of subtasks according to the updated available resources and task information.

[0173] Step 12: Return to the step of updating the available resources of the edge computing system based on the allocated network slices, until the allocation is complete and the slice results are obtained.

[0174] Step 13: For each network slice corresponding to each subtask set in the slice result, determine the scheduling strategy of on-slice computing resources based on the computational load of each computing task in the subtask set.

[0175] 1) When a network slice includes a destination node, allocate on-slice computing resources to each computing task based on the computational workload of each task;

[0176] 2) When a network slice includes multiple destination nodes, allocate on-slice computing resources to each computing task based on the computing resources of each destination node and the computing workload of each computing task.

[0177] Step 14: Determine the scheduling strategy for on-chip network resources based on the data volume of each computing task in the subtask set.

[0178] 1) When the same destination node within a network slice includes a transmission path, allocate on-slice network resources to each computing task based on the data volume of each computing task;

[0179] 2) When the same destination node within a network slice includes multiple transmission paths, allocate intra-slice network resources to each computing task based on the network resources of the transmission paths between the source and destination nodes and the data volume of each computing task.

[0180] Step 15: Send the on-chip resource scheduling policy to the software-defined network controller, so that the software-defined network controller can generate network slicing instructions according to the on-chip resource scheduling policy and send the network slicing instructions to each network transmission node and each edge computing node to execute the network slicing process.

[0181] Step 16: Send the on-chip resource scheduling policy to multiple IoT terminals so that each IoT terminal can send computing tasks to the edge computing nodes on the chip for processing through the allocated network slices according to the on-chip resource scheduling policy.

[0182] Step 17: Send the on-chip resource scheduling policy to multiple network transmission nodes so that each network transmission node can schedule the on-chip computing tasks according to the on-chip resource scheduling policy.

[0183] Step 18: Send the on-chip resource scheduling strategy to multiple edge computing nodes so that each edge computing node can process the on-chip computing tasks according to the on-chip resource scheduling strategy and return the task results to the corresponding IoT terminal.

[0184] Assume the system contains two wireless subnets, and the corresponding wireless subnet access points are... , Wireless subnet access point The corresponding wireless subnet contains 5 IoT terminals, and the task information they generate is as follows: , , , , Wireless subnet access point The corresponding wireless subnet contains 3 IoT terminals, and the task information they generate is as follows: , , The edge computing system contains four edge computing nodes. , , , The task types and available computing resources supported by edge computing nodes are as follows: : , : , : , : , To edge computing nodes The transmission path and corresponding link bandwidth are , To edge computing nodes The transmission path and corresponding link bandwidth are , To edge computing nodes The transmission path and corresponding link bandwidth are , To edge computing nodes The transmission path and corresponding link bandwidth are , To edge computing nodes The transmission path and corresponding link bandwidth are , To edge computing nodes The transmission path and corresponding link bandwidth are , To edge computing nodes The transmission path and corresponding link bandwidth are , To edge computing nodes The transmission path and corresponding link bandwidth are .

[0185] The edge computing controller receives data from multiple IoT terminals. The sent task information categorizes all computational tasks into three types based on task type and maximum tolerable end-to-end latency. These tasks are then sorted in descending order of latency level, resulting in multiple task sets ordered in descending order of latency level. , , );

[0186] For the first task set Network slice allocation: Based on the data source of the computation task (wireless subnet access point), the task set can be divided into two sub-task sets. and Since the computational cost of the first subtask set is greater than that of the second subtask set, the network slices of the first subtask set are determined first.

[0187] according to Figure 8 The task types supported by each edge computing node in the mid-edge computing system determine the candidate computing node set for the first sub-task set. The candidate network slice set for this subtask set is shown in Table 1:

[0188] Table 1

[0189]

[0190] The edge computing nodes in the candidate network slice set are arranged in ascending order of the transmission path between the source and destination node pairs. The order of the edge computing nodes is as follows: ,by For the destination node, with The combination of resources for the transmission path can satisfy the end-to-end latency requirements of the first subtask set, i.e. This can form network slices. The network slice is then allocated to the subtask set. Subsequently, based on the data volume and computational load of each computational task in the first subtask set, the on-chip resource scheduling strategy for each computational task is determined as follows: , Therefore, the average end-to-end latency of the first subtask set is 0.92s.

[0191] Determining the network slice for the second subtask set: Based on the edge computing system and the slicing results of the first subtask set, the candidate computing node set for the second subtask set is as follows: The candidate network slice set for this subtask set is shown in Table 2:

[0192] Table 2

[0193]

[0194] The edge computing nodes in the candidate network slice set are arranged in ascending order of the transmission path between the source and destination node pairs. The order of the edge computing nodes is as follows: .by For the destination node, with The target latency corresponding to the resource combination of the transmission path is This can meet the end-to-end latency requirements of the subtask set, forming a network slice. Based on the data volume and computational volume of each computational task in this subtask set, the on-chip resource scheduling strategy for each computational task is determined as follows: , Therefore, the end-to-end average latency of the second subtask set is 0.86s.

[0195] Similarly, a second set of tasks can be obtained. Network Slicing Since there is only one computing task in this network slice, the computing task exclusively enjoys the on-slice resources, and the end-to-end latency of the computing task is 1.96s.

[0196] Determine the first sub-task set in the third task set. Network Slicing If there is only one computing task in the network slice, then the computing task has exclusive access to the slice's resources, and the end-to-end latency of the computing task is 3.0s.

[0197] Determine the second sub-task set within the third task set. Network Slicing The network slice has only one computing task, which exclusively uses the slice's resources, and the end-to-end latency of the computing task is 2.9 seconds.

[0198] In the above embodiments, logical isolation between different types of tasks and tasks of the same type but from different sources can be achieved, providing strict end-to-end latency guarantees for all types of tasks. Moreover, the computing power of network slices is preferentially matched to the edge computing nodes closest to the data source that can meet the computing power requirements of the tasks, which not only improves the data privacy of smart grid applications, but also reduces network load and data network transmission latency. Furthermore, the on-chip resource scheduling strategy fairly schedules on-chip tasks from the perspective of network resources and computing resources, so that on-chip tasks can obtain the same level of end-to-end latency service.

[0199] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.

[0200] Based on the same inventive concept, this application also provides a network slicing and intra-chip scheduling apparatus for implementing the network slicing and intra-chip scheduling methods described above. The solution provided by this apparatus is similar to the implementation scheme described in the above methods; therefore, the specific limitations in one or more network slicing and intra-chip scheduling apparatus embodiments provided below can be found in the limitations of the network slicing and intra-chip scheduling methods described above, and will not be repeated here.

[0201] In one exemplary embodiment, such as Figure 9 As shown, a network slicing and intra-slice scheduling device is provided, including:

[0202] The information acquisition module 701 is used to acquire task information of multiple computing tasks; wherein, the task information includes at least the task type, data volume, computation volume and maximum tolerable end-to-end latency;

[0203] The task classification module 702 is used to classify and sort multiple computing tasks according to task information to obtain multiple ordered task sets;

[0204] The network slicing module 703 is used to allocate network slices to each task set in sequence according to task information and available resources of the edge computing system to obtain slicing results; wherein, the slicing results include the source node, at least one destination node and at least one transmission path of each network slice;

[0205] The strategy determination module 704 is used to determine the intra-slice resource scheduling strategy for each network slice based on task information and slicing results.

[0206] In one embodiment, the network slicing module 703 is specifically used to start from the first position of the sorting of multiple task sets, and for each task set, classify and sort them according to the wireless subnet access points corresponding to each computing task in the task set to obtain multiple sequentially arranged sub-task sets; and allocate network slices to each sub-task set in sequence according to the task information and the available resources of the edge computing system to obtain slicing results.

[0207] In one embodiment, the network slicing module 703 is specifically used to allocate a network slice to the first subtask set among multiple subtask sets according to task information and available resources of the edge computing system; update the available resources of the edge computing system according to the allocated network slice, and allocate a network slice to the next subtask set according to the updated available resources and task information; return to the step of updating the available resources of the edge computing system according to the allocated network slice, until the allocation is completed and the slicing result is obtained.

[0208] In one embodiment, the network slicing module 703 is specifically configured to, for the first subtask set, allocate a candidate computing node set based on the available resources of the edge computing system; wherein the candidate computing node set includes multiple edge computing nodes serving as destination nodes; determine a candidate end-to-end transmission path set between each source-destination node pair and the equivalent bandwidth corresponding to the candidate end-to-end transmission path set based on the candidate computing node set; wherein the source node in the source-destination node pair is the wireless subnet access point corresponding to the subtask set, the transmission path in the candidate end-to-end transmission path set has the maximum equivalent bandwidth, and the equivalent bandwidth of the transmission path is determined based on the available bandwidth of each link in the transmission path; determine a candidate network slice set based on the candidate end-to-end transmission path set and the equivalent bandwidth of each transmission path; and allocate network slices to the subtask set based on the task information, the candidate network slice set, and the maximum tolerable end-to-end delay of the subtask set.

[0209] In one embodiment, the network slicing module 703 is specifically used to sort the edge computing nodes in the candidate network slice set according to the transmission path between the source and destination nodes from shortest to longest; taking the first edge computing node after sorting as the destination node and the wireless subnet access point corresponding to the subtask set as the source node, the computational and data volumes of each computing task in the subtask set are determined according to the task information, and the candidate network resources and candidate computing resources corresponding to the destination node are used to determine the target latency corresponding to the subtask set; wherein, the candidate network resources include the equivalent bandwidth of the transmission path between the source and destination nodes, and the candidate computing resources include the computing resources that the destination node can allocate to the subtask set; in the subtask set corresponding to If the target latency is not greater than the maximum tolerable end-to-end latency corresponding to the sub-task set, network slices are formed based on the candidate network resources and candidate computing resources corresponding to the destination node, and the network slices are allocated to the sub-task set. If the target latency corresponding to the sub-task set is greater than the maximum tolerable end-to-end latency corresponding to the sub-task set, starting from the first edge computing node, with at least two edge computing nodes as destination nodes and the wireless subnet access point corresponding to the sub-task set as the source node, network slices are formed based on the computational and data volume of each computing task in the sub-task set, and the candidate network resources and candidate computing resources corresponding to at least two destination nodes, and the network slices are allocated to the sub-task set.

[0210] In one embodiment, the strategy determination module 704 is specifically used to determine the scheduling strategy of on-chip computing resources based on the computational amount of each computing task in the sub-task set for the network slice corresponding to each sub-task set in the slicing result; and to determine the scheduling strategy of on-chip network resources based on the data amount of each computing task in the sub-task set.

[0211] In one embodiment, the strategy determination module 704 is specifically used to allocate on-chip computing resources to each computing task based on the computing workload of each computing task when the network slice includes one destination node; and to allocate on-chip computing resources to each computing task based on the computing resources of each destination node and the computing workload of each computing task when the network slice includes multiple destination nodes.

[0212] In one embodiment, the strategy determination module 704 is specifically used to allocate intra-slice network resources to each computing task based on the data volume of each computing task when the same destination node within a network slice includes one transmission path; and to allocate intra-slice network resources to each computing task based on the network resources of the transmission path between the source and destination nodes and the data volume of each computing task when the same destination node within a network slice includes multiple transmission paths.

[0213] In one embodiment, the device further includes:

[0214] The first strategy sending module is used to send the on-chip resource scheduling strategy to the software-defined network controller, so that the software-defined network controller generates network slicing instructions according to the on-chip resource scheduling strategy and sends the network slicing instructions to each network transmission node and each edge computing node to execute the network slicing process.

[0215] The second strategy sending module is used to send the on-chip resource scheduling strategy to multiple IoT terminals, so that each IoT terminal can send the computing task to the on-chip edge computing node for processing through the allocated network slice according to the on-chip resource scheduling strategy.

[0216] The third strategy sending module is used to send the on-chip resource scheduling strategy to multiple network transmission nodes, so that each network transmission node can schedule the on-chip computing tasks according to the on-chip resource scheduling strategy.

[0217] The fourth strategy sending module is used to send the on-chip resource scheduling strategy to multiple edge computing nodes, so that each edge computing node can process the on-chip computing tasks according to the on-chip resource scheduling strategy and return the task results to the corresponding IoT terminal.

[0218] Each module in the aforementioned network slicing and on-chip scheduling device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the operations corresponding to each module.

[0219] In one exemplary embodiment, a computer device is provided, which may be an edge computing controller, and its internal structure diagram may be as follows: Figure 10 As shown, this computer device includes a processor, memory, input / output interfaces (I / O), and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores network slicing and on-chip scheduling data. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When executed by the processor, the computer program implements a network slicing and on-chip scheduling method.

[0220] Those skilled in the art will understand that Figure 10The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0221] In one exemplary embodiment, a non-transitory computer-readable storage medium including instructions is also provided, such as a memory including instructions that can be executed by a processor of an electronic device to perform the above-described method. For example, the non-transitory computer-readable storage medium may be a ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, and optical data storage device, etc.

[0222] In one exemplary embodiment, a computer program product is also provided, which, when executed by a processor, can implement the above-described methods. The computer program product includes one or more computer instructions. When these computer instructions are loaded and executed on a computer, some or all of the above-described methods can be implemented, wholly or partially, according to the processes or functions described in the embodiments of this application.

[0223] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0224] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0225] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0226] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A network slicing and intra-slice scheduling method, characterized in that, The method includes: Obtain task information for multiple computing tasks; wherein the task information includes at least the task type, data volume, computational load, and maximum tolerable end-to-end latency; Based on the task information, the multiple computing tasks are classified and sorted to obtain multiple sequentially arranged task sets. Based on the task information and the available resources of the edge computing system, network slices are allocated sequentially to each task set to obtain slicing results; wherein, the slicing results include the source node, at least one destination node, and at least one transmission path of each network slice; Based on the task information and the slicing results, the intra-slice resource scheduling strategy for each network slice is determined.

2. The method according to claim 1, characterized in that, The process of allocating network slices to each task set sequentially based on the task information and the available resources of the edge computing system to obtain the slicing results includes: Starting from the first position of the sorting of the multiple task sets, for each task set, the wireless subnet access point corresponding to each computing task in the task set is classified and sorted to obtain multiple sequentially arranged subtask sets. Based on the task information and the available resources of the edge computing system, network slices are allocated to each of the sub-task sets in sequence to obtain the slice results.

3. The method according to claim 2, characterized in that, The step of allocating network slices to each of the subtask sets sequentially based on the task information and the available resources of the edge computing system to obtain the slicing results includes: Based on the task information and the available resources of the edge computing system, a network slice is allocated to the first subtask set among the multiple subtask sets; The available resources of the edge computing system are updated based on the allocated network slices, and network slices are allocated to the next set of subtasks based on the updated available resources and the task information. Return to the step of updating the available resources of the edge computing system based on the allocated network slices, until the allocation is completed and the slice result is obtained.

4. The method according to claim 3, characterized in that, The step of allocating a network slice to the first sub-task set among multiple sub-task sets based on the task information and the available resources of the edge computing system includes: For the first set of subtasks, a set of candidate computing nodes is allocated based on the available resources of the edge computing system; wherein, the set of candidate computing nodes includes multiple edge computing nodes that serve as destination nodes; Based on the candidate computing node set, a set of candidate end-to-end transmission paths between each source-destination node pair is determined, along with the equivalent bandwidth corresponding to the set of candidate end-to-end transmission paths; wherein, the source node in the source-destination node pair is the wireless subnet access point corresponding to the subtask set, and the transmission path in the set of candidate end-to-end transmission paths has the maximum equivalent bandwidth, the equivalent bandwidth of the transmission path being determined based on the available bandwidth of each link in the transmission path; The candidate network slice set is determined based on the candidate end-to-end transmission path set and the equivalent bandwidth of each transmission path; Network slices are allocated to the subtask set based on the task information, the candidate network slice set, and the maximum tolerable end-to-end latency of the subtask set.

5. The method according to claim 4, characterized in that, The step of allocating network slices to the sub-task set based on the task information, the candidate network slice set, and the maximum tolerable end-to-end latency of the sub-task set includes: The edge computing nodes in the candidate network slice set are sorted from shortest to longest according to the transmission path between the source and destination node pairs. Using the first edge computing node after sorting as the destination node and the wireless subnet access point corresponding to the subtask set as the source node, the computational and data volumes of each computing task in the subtask set are determined based on the task information, and the candidate network resources and candidate computing resources corresponding to the destination node are used to determine the target latency corresponding to the subtask set; wherein, the candidate network resources include the equivalent bandwidth of the transmission path between the source and destination node pairs, and the candidate computing resources include the computing resources that the destination node can allocate to the subtask set; If the target latency corresponding to the subtask set is not greater than the maximum tolerable end-to-end latency corresponding to the subtask set, a network slice is formed based on the candidate network resources and candidate computing resources corresponding to the destination node, and the network slice is allocated to the subtask set. If the target latency corresponding to the sub-task set is greater than the maximum tolerable end-to-end latency corresponding to the sub-task set, starting from the first edge computing node, with at least two edge computing nodes as destination nodes and the wireless subnet access point corresponding to the sub-task set as the source node, a network slice is formed based on the computational and data volume of each computing task in the sub-task set, the candidate network resources and candidate computing resources corresponding to the at least two destination nodes, and the network slice is allocated to the sub-task set.

6. The method according to claim 2, characterized in that, The step of determining the intra-slice resource scheduling strategy for each network slice based on the task information and the slicing results includes: For each network slice corresponding to each set of subtasks in the slice result, the scheduling strategy of on-slice computing resources is determined according to the computing volume of each computing task in the set of subtasks. The scheduling strategy for on-chip network resources is determined based on the data volume of each computing task in the subtask set.

7. The method according to claim 4, characterized in that, The step of determining the scheduling strategy for on-chip computing resources based on the computational load of each computing task in the subtask set includes: In the case where the network slice includes a destination node, the on-slice computing resources are allocated to each computing task according to the computational workload of each computing task; In the case where the network slice includes multiple destination nodes, the on-slice computing resources are allocated to each computing task according to the computing resources of each destination node and the computing workload of each computing task.

8. The method according to claim 4, characterized in that, The step of determining the scheduling strategy for on-chip network resources based on the data volume of each computing task in the subtask set includes: When the same destination node within the network slice includes a transmission path, the intra-slice network resources are allocated to each computing task according to the data volume of each computing task. In the case where the same destination node within the network slice includes multiple transmission paths, the intra-slice network resources are allocated to each computing task based on the network resources of the transmission paths between the source and destination nodes and the data volume of each computing task.

9. The method according to any one of claims 1-8, characterized in that, The method further includes: The on-chip resource scheduling policy is sent to the software-defined network controller, so that the software-defined network controller generates network slicing instructions according to the on-chip resource scheduling policy and sends the network slicing instructions to each of the network transmission nodes and each of the edge computing nodes to execute the network slicing process. The on-chip resource scheduling strategy is sent to multiple IoT terminals, so that each IoT terminal sends the computing task to the edge computing node on the chip for processing through the allocated network slice according to the on-chip resource scheduling strategy. The on-chip resource scheduling strategy is sent to multiple network transmission nodes so that each network transmission node schedules on-chip computing tasks according to the on-chip resource scheduling strategy. The on-chip resource scheduling strategy is sent to multiple edge computing nodes, so that each edge computing node processes the on-chip computing tasks according to the on-chip resource scheduling strategy and returns the task results to the corresponding IoT terminal.

10. A network slicing and intra-slice scheduling device, characterized in that, The device includes: An information acquisition module is used to acquire task information for multiple computing tasks; wherein, the task information includes at least the task type, data volume, computational volume, and maximum tolerable end-to-end latency; The task classification module is used to classify and sort multiple computing tasks according to the task information to obtain multiple sequentially arranged task sets. The network slicing module is used to allocate network slices to each of the task sets in sequence according to the task information and the available resources of the edge computing system, and obtain slicing results; wherein, the slicing results include the source node, at least one destination node and at least one transmission path of each network slice; The strategy determination module is used to determine the intra-slice resource scheduling strategy for each network slice based on the task information and the slicing results.

Images