A method for intelligent resource allocation in 5G networks under a cloud computing environment
By acquiring resource adjustment tasks that have not yet been completed in the cloud computing environment, establishing a set of resource adjustments in progress, predicting stable resource quantities, and optimizing resource adjustment schemes, the problem of unstable resource allocation is solved, and resource utilization efficiency and allocation stability are improved.
Patent Information
- Application Number
- CN202611569809.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Priority Date
- 2026-09-22
- Filing Date
- 2026-10-12
- Publication Date
- 2026-11-17
AI Technical Summary
The existing resource allocation methods in cloud computing environments suffer from problems such as repeated resource adjustments, resource over-provisioning, ineffective adjustments, and repeated alternation between expansion and contraction, leading to unstable resource allocation.
By acquiring resource adjustment tasks that have not yet been completed in the cloud computing environment, an in-progress resource adjustment set is established, the stable resource quantity is predicted, duplicate adjustments, reverse adjustments and resource occupation conflicts are identified, an optimized resource adjustment plan is generated, and the plan is continuously updated in conjunction with the execution status of the resource adjustment tasks.
It improves the stability and efficiency of 5G network resource allocation in cloud computing environments, and reduces redundant expansion, over-release and resource over-provisioning caused by resource adjustment execution latency.
Smart Images

Figure CN123124111A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the fields of 5G network resource management and cloud computing, and particularly to a method for intelligent resource allocation of 5G networks in a cloud computing environment. Background Technology
[0002] As 5G networks evolve towards cloudification, virtualization, and service-orientation, more and more 5G network functions are deployed in cloud computing environments in the form of virtual machines, containers, or cloud-native network functions. To meet the dynamic needs of different 5G services in terms of computing power, storage capacity, and network transmission capabilities, cloud computing platforms typically need to perform resource adjustment operations such as adding, releasing, or migrating resources for corresponding service instances based on service load and node resource status.
[0003] Existing resource allocation methods typically determine resource requirements based on currently collected business load and resource occupancy status, and generate new resource adjustment tasks accordingly. However, in real-world cloud computing environments, there is usually a certain execution delay between the generation of a resource adjustment task and the actual configuration and effectiveness of the corresponding resources. For example, resource expansion requires processes such as resource application, instance creation, resource binding, and business takeover; similarly, resource release or migration requires state synchronization and resource unbinding. Therefore, while a resource adjustment task is incomplete, the currently monitored resource status may still remain in its pre-adjustment state.
[0004] During continuous resource scheduling, if subsequent scheduling cycles still directly rely on the current resource status for resource assessment without considering resource adjustment tasks that have been issued but not yet taken effect, it is easy to generate duplicate resource adjustment tasks in the same direction, or generate resource adjustment tasks in the opposite direction before existing resource adjustments are completed. Furthermore, for resources that are in the process of migrating, locking, or releasing, multiple resource adjustment tasks may compete for the same resource, leading to problems such as duplicate resource allocation, resource over-provisioning, invalid adjustments, and even repeated alternations between expansion and contraction.
[0005] To address this, a method for intelligent resource allocation in 5G networks under a cloud computing environment is proposed. Summary of the Invention
[0006] The main objective of this invention is to provide a method for intelligent resource allocation in 5G networks under a cloud computing environment, which can effectively solve the problems in the background technology.
[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows:
[0008] A method for intelligent resource allocation in a 5G network under a cloud computing environment includes the following steps:
[0009] S1. Obtain the service load, resource requirement constraints and current resource occupancy status of multiple 5G service instances in the cloud computing environment, and obtain the current resource status of the cloud computing node carrying the 5G service instances and the resource adjustment tasks that have not yet been completed.
[0010] S2. Based on the adjustment object, resource type, adjustment direction, adjustment resource quantity and task execution status of each resource adjustment task, determine the resource adjustment quantity that has not yet taken effect, and establish a corresponding in-transit resource adjustment set according to the adjustment object and resource type.
[0011] S3. For any target resource type of any 5G service instance, determine the predicted stable resource quantity based on the current actual resource quantity and the expected increase and release of resources in the in-transit resource adjustment set during the target analysis period:
[0012]
[0013] in, For the first The first 5G service instance Predicting stable resource quantities for resource types. This represents the current actual resource quantity. and Add task sets and resource release task sets to the corresponding resources in transit; and These represent the resource adjustments for the corresponding tasks that have not yet taken effect. and These are the expected effectiveness coefficients of the corresponding resource adjustment tasks during the target analysis period;
[0014] S4. Determine the required resource quantity for the target analysis period based on the business load and resource demand constraints, and determine the stable resource deviation based on the required resource quantity and the predicted stable resource quantity:
[0015]
[0016] in, For the first The first 5G service instance for the first The required amount of resources for this type of resource; when When it is determined that a corresponding resource gap exists under the predicted steady state, when At that time, it is determined that there is a corresponding resource surplus under the predicted steady state;
[0017] S5. Generate an initial resource adjustment action based on the stable resource deviation amount, and compare the initial resource adjustment action with the set of resource adjustments in transit according to the adjustment object, resource type and adjustment direction, identify duplicate adjustment, reverse adjustment or resource occupation conflict, and determine the remaining adjustment amount that has not been covered by existing in-transit resource adjustments in the same direction.
[0018] S6. Generate multiple candidate resource adjustment schemes based on the initial resource adjustment action, remaining adjustment amount, and resource adjustment conflict identification results, and determine the target resource adjustment scheme from the candidate resource adjustment schemes that meet the resource demand constraints and have no unresolved resource adjustment conflicts.
[0019] S7. Perform resource adjustments on the corresponding 5G service instances or cloud computing nodes according to the target resource adjustment scheme, and update the current resource status and the set of resource adjustments in transit according to the execution status of the resource adjustment task, so that subsequent resource allocation is based on the updated resource status.
[0020] Furthermore, the resource type includes at least one of computing resources, memory resources, storage resources, and network transmission resources; the resource adjustment task includes at least one of resource addition task, resource release task, and resource migration task, wherein the resource migration task forms corresponding resource release adjustment and resource addition adjustment according to the migration source node and migration target node, respectively.
[0021] Furthermore, in step S3, the task stage identifier, completed resource adjustment amount, and estimated remaining execution time of the resource adjustment task fed back by the resource orchestration execution component are obtained. The estimated effective adjustment amount within the target analysis period is determined based on the task stage identifier, completed resource adjustment amount, and estimated remaining execution time. The estimated effective adjustment coefficient is determined based on the estimated effective adjustment amount and the corresponding uneffective resource adjustment amount. When the estimated remaining execution time of the resource adjustment task is not greater than the remaining time of the target analysis period, the corresponding estimated effective coefficient is determined to be 1.
[0022] Furthermore, in step S4, the stable resource deviation amount is determined for different resource types of the same 5G service instance. Based on each stable resource deviation amount, the resource shortage type and resource surplus type of the 5G service instance in the predicted stable state are determined. The stable resource deviation amount corresponding to the resource type with resource shortage is used as the initial resource increase amount, and the absolute value of the stable resource deviation amount corresponding to the resource type with resource surplus is used as the initial resource release amount.
[0023] Furthermore, in step S5:
[0024] When there are in-transit resource adjustment tasks for the same adjustment object and the same resource type with the same direction as the initial resource adjustment action, and the resource adjustment amount of the in-transit resource adjustment task that has not yet taken effect can cover the adjustment amount corresponding to the initial resource adjustment action, it is identified as a duplicate adjustment;
[0025] When an initial resource adjustment action targeting the same adjustment object and the same resource type is in the opposite direction to an ongoing resource adjustment task, it is identified as a reverse adjustment.
[0026] When the resource corresponding to the initial resource adjustment action is in a locked state or a migration state of an existing in-transit resource adjustment task, it is identified as a resource occupation conflict.
[0027] Furthermore, in step S6, for initial resource adjustment actions with resource adjustment conflicts, candidate resource adjustment schemes are generated by at least one of the following methods: waiting for the corresponding in-transit resource adjustment task to complete, modifying the unexecuted portion of the resource adjustment amount in the corresponding in-transit resource adjustment task, or generating a supplementary resource adjustment task for the remaining adjustment amount. Based on the comparison of the degree of improvement in stable resource deviation, resource adjustment overhead, and expected effective delay of each candidate resource adjustment scheme, the target resource adjustment scheme is determined from the candidate resource adjustment schemes that meet the resource demand constraints.
[0028] Furthermore, the first is determined according to the following formula. Evaluation values of the candidate resource adjustment schemes:
[0029]
[0030] in, For the first The evaluation value of each candidate resource adjustment plan. The improvement in stable resource deviation after implementing the candidate resource adjustment scheme relative to the state before implementation. The resource adjustment overhead required to execute the candidate resource adjustment scheme, The estimated effective time delay for the resource adjustment actions in the candidate resource adjustment schemes to reach the corresponding predicted stable state. , , As corresponding weights; from the candidate resource adjustment schemes that meet the resource demand constraints and have no unresolved resource adjustment conflicts, select the candidate resource adjustment scheme with the largest scheme evaluation value as the target resource adjustment scheme.
[0031] Furthermore, in step S7, when the resource adjustment task is partially completed, the corresponding unused resource adjustment amount is updated according to the execution status feedback; when the resource adjustment task is completed, the corresponding resource adjustment result is updated to the current resource status, and the resource adjustment task is removed from the in-transit resource adjustment set; when the resource adjustment task fails or is canceled, it is removed from the in-transit resource adjustment set, and the impact of its unused resource adjustment amount on the predicted stable resource amount is canceled.
[0032] A 5G network intelligent resource allocation system in a cloud computing environment includes:
[0033] The status acquisition module is used to acquire the service load, resource requirement constraints and current resource occupancy status of multiple 5G service instances, as well as the current resource status of cloud computing nodes and resource adjustment tasks that have not yet been completed.
[0034] The in-transit adjustment determination module is used to determine the amount of resource adjustments that have not yet taken effect and establish an in-transit resource adjustment set based on the adjustment object, resource type, adjustment direction, amount of adjusted resources and task execution status of the resource adjustment task.
[0035] The stable state determination module is used to determine the predicted stable resource quantity based on the current actual resource quantity and the expected increase and release of resources in the in-transit resource adjustment set.
[0036] The resource deviation determination module is used to determine the stable resource deviation based on the resource demand and predicted stable resource quantity of the 5G service instance, and generate an initial resource adjustment action based on the stable resource deviation.
[0037] The conflict identification module is used to compare the initial resource adjustment action with the set of resource adjustments in progress, identify duplicate adjustments, reverse adjustments or resource occupation conflicts, and determine the remaining adjustment amount;
[0038] The resource allocation module is used to generate candidate resource adjustment schemes based on the initial resource adjustment action, the remaining adjustment amount and the resource adjustment conflict identification result, and to determine the target resource adjustment scheme from the candidate resource adjustment schemes that meet the resource demand constraints and have no unresolved resource adjustment conflicts.
[0039] The execution update module is used to perform resource adjustments according to the target resource adjustment plan, and update the current resource status and the set of resource adjustments in progress based on the execution status of the resource adjustment tasks.
[0040] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described.
[0041] Compared with the prior art, the present invention has the following beneficial effects:
[0042] 1. This invention obtains resource adjustment tasks that have not yet been completed, determines the corresponding resource adjustment amounts that have not yet taken effect, and establishes a set of resource adjustments in progress. This enables the resource allocation process to simultaneously consider the current actual resource status and the resource adjustment status that is being executed, avoiding redundant resource decisions based solely on instantaneous resource monitoring results.
[0043] 2. This invention determines the predicted stable resource quantity based on the current actual resource quantity, the expected effective increase in resources, and the resource release quantity, and further calculates the stable resource deviation quantity. This enables the resource demand judgment to be based on the expected resource status after the completion of the existing resource adjustment task, which can reduce the repeated expansion, over-release, and resource over-allocation caused by the execution delay of resource adjustment.
[0044] 3. This invention compares the initial resource adjustment action generated based on stable resource deviation with existing resource adjustment tasks in progress, identifies duplicate adjustments, reverse adjustments, and resource occupation conflicts, and forms a subsequent resource adjustment plan based on the remaining adjustment amount not covered by existing adjustment tasks, thereby reducing conflicting resource control commands and repeated oscillations between expansion and contraction.
[0045] 4. This invention generates and optimizes candidate resource adjustment schemes, and continuously updates the current resource status and the set of resource adjustments in progress in conjunction with the actual execution status of resource adjustment tasks, forming a closed loop between resource status prediction, resource adjustment decision-making and execution feedback, thereby improving the stability of dynamic allocation of 5G network resources, resource utilization efficiency and business resource guarantee capabilities in the cloud computing environment. Attached Figure Description
[0046] Figure 1 This is a schematic diagram of the method flow of the present invention;
[0047] Figure 2 This is a schematic diagram of the system modules of the present invention. Detailed Implementation
[0048] 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.
[0049] Example 1
[0050] like Figure 1-2 As shown, a method for intelligent resource allocation in a cloud computing environment for 5G networks can be deployed in a 5G network resource management platform, a cloud resource orchestration platform, or a resource scheduling and control device that communicates with both.
[0051] The cloud computing environment runs multiple 5G service instances. These 5G service instances can be 5G network function instances deployed using virtual machines, containers, or other virtualization methods. For each 5G service instance, one or more of the following resources—computing resources, memory resources, storage resources, and network transmission resources—can be allocated based on the actual deployment conditions.
[0052] During resource scheduling, the resource monitoring component typically collects the current resource status that has already taken effect. However, there is an execution process between the issuance of control commands and the actual implementation of corresponding resource adjustments, such as adding, releasing, or migrating resources. Therefore, this embodiment further acquires resource adjustment tasks that have been generated but not yet completed, and uses their corresponding inactive resource adjustment amounts as in-transit resource statuses to participate in subsequent resource allocation.
[0053] Specifically, it includes the following steps.
[0054] S1. Obtain 5G service status, current resource status, and resource adjustment tasks.
[0055] Obtain the service load, resource requirement constraints, and current resource occupancy status of multiple 5G service instances in the cloud computing environment, and obtain the current resource status of the cloud computing nodes that support each 5G service instance.
[0056] Among them, the service load is used to reflect the amount of business that the corresponding 5G service instance needs to process during the current or target analysis period. It can be determined by indicators that reflect the service processing pressure, such as the number of service requests, the number of concurrent sessions, data throughput, and packet processing rate.
[0057] Resource requirement constraints are used to represent the resource conditions that must be met for 5G services to operate normally. For example, for computing resources, these may include the minimum amount of computing resources or the minimum processing power; for network transmission resources, these may include the minimum available bandwidth, the maximum allowable transmission latency, or the corresponding quality of service constraints.
[0058] The current resource usage status indicates the state after resources have been actually allocated and are in effect. For example, a 5G service instance currently has 8 virtual computing cores, 16GB of memory, and 2Gbps of network transmission bandwidth.
[0059] Further retrieve resource adjustment tasks that have not yet been completed from the resource orchestration platform.
[0060] The resource adjustment task includes at least one of the following: resource addition task, resource release task, and resource migration task.
[0061] For example, before the current scheduling time, the resource management platform has generated a resource increase task for business instance A to increase computing resources from 8 virtual computing cores to 12 virtual computing cores. However, since the resource application, resource binding, or instance adjustment process has not yet been completed, the current monitoring result still shows 8 virtual computing cores.
[0062] If subsequent resource scheduling still relies directly on these eight virtual computing cores to determine resource requirements, resource increase tasks may be generated again. This embodiment avoids the aforementioned duplicate resource adjustments by incorporating unfinished resource adjustment tasks into the resource status analysis.
[0063] S2. Establish a set of resources in transit for adjustment.
[0064] For each resource adjustment task that has not yet been completed, obtain the corresponding adjustment object, resource type, adjustment direction, amount of adjusted resources, and task execution status.
[0065] The objects to be adjusted can be 5G service instances, virtual network function instances, or corresponding cloud computing nodes; the resource types can include at least one of computing resources, memory resources, storage resources, and network transmission resources.
[0066] The adjustment direction is used to indicate whether the corresponding resource is being added or released.
[0067] For example, the computing resource adjustment task corresponding to business instance A is as follows: the current actual computing resource amount is 8 virtual computing cores; the target computing resource amount is 12 virtual computing cores; the adjustment direction is resource increase; and the resource adjustment amount is 4 virtual computing cores.
[0068] If one virtual computing core has already completed resource configuration and been included in the current actual resource status, then the remaining unadjusted resource adjustment amount is three virtual computing cores.
[0069] The amount of resource adjustments that have not yet taken effect can be represented as:
[0070]
[0071] in, Indicates the first Resource adjustments for each resource adjustment task that has not yet taken effect; This indicates the total amount of resource adjustment corresponding to this resource adjustment task; This indicates the amount of resource adjustments that have been actually completed and included in the current resource status.
[0072] Based on the adjustment object and resource type, the corresponding resource adjustment tasks that have not yet been completed are grouped into a set of resource adjustments in progress.
[0073] For the The first 5G service instance These types of resources can be used to create separate sets of in-transit resources to increase task availability. and collection of resource release tasks in transit .
[0074] For resource migration tasks, resource release adjustments and resource increase adjustments are made according to the source node and the target node, respectively.
[0075] For example, when a business instance migrates from cloud computing node N1 to node N2, the corresponding migration resources are displayed as resources to be released for node N1, and as resources to be added for node N2. This allows resource migration to be uniformly included in the in-transit resource adjustment status for processing.
[0076] S3. Determine the predicted stable resource quantity
[0077] For each target resource type of each 5G service instance, the predicted stable resource quantity is determined based on the current actual resource quantity and the corresponding in-transit resource adjustment set.
[0078] Predicted stable resource quantity refers to the resource quantity that a corresponding business instance is expected to achieve after further considering the resource adjustment tasks that are expected to be completed or partially completed within the target analysis period, based on the current actual resource status.
[0079] No. The first 5G service instance The predicted stable resource quantity corresponding to this type of resource is:
[0080]
[0081] in, For the first The first 5G service instance Predicting stable resource quantities for similar resources; This represents the amount of resources that are currently in effect. Add a task set to the corresponding in-transit resources; For the corresponding set of tasks to release resources in transit; For the first The resource adjustment amount for each resource increase task that has not yet taken effect; For the first Resource adjustments for resource release tasks that have not yet taken effect; and These are the expected effectiveness coefficients for the corresponding resource adjustment tasks during the target analysis period, with values ranging from 0 to 1.
[0082] In this embodiment, the expected effectiveness coefficient is determined based on the actual task execution status reported by the resource orchestration execution component, rather than being set arbitrarily by humans.
[0083] Specifically, it can obtain the task stage identifier, completed resource adjustment amount, uncompleted resource adjustment amount, and estimated remaining execution time for the corresponding resource adjustment task, and determine the estimated actual effective resource adjustment amount within the target analysis period by combining the remaining time of the target analysis period.
[0084] For resource increase tasks that can provide feedback on the progress of continuous resource adjustments, the expected effectiveness coefficient can be further expressed as:
[0085]
[0086] in, For the first The amount of resources that each resource increase task is expected to actually take effect during the target analysis period. This refers to the resource adjustments that have not yet taken effect for this task.
[0087] Expected activation coefficient for resource release task The same method can be used to determine this.
[0088] For example, if a resource addition task still has 4 virtual computing cores that have not yet been implemented, and based on the task stage and remaining execution time feedback from the resource orchestration execution component, it is expected that 3 of these virtual computing cores will be configured before the end of the target analysis period, then:
[0089]
[0090] When the estimated remaining execution time of the resource adjustment task is not greater than the remaining time of the target analysis period, it can be assumed that the corresponding resource adjustment amount that has not yet taken effect can take effect within the target analysis period. In this case, the corresponding expected effective coefficient is determined to be 1.
[0091] In another implementation, for resource adjustment tasks that can only provide feedback on task stages but not continuously provide feedback on resource adjustment amounts, a correspondence between task stages and expected completion ratios can be established based on historical resource adjustment tasks of the same type.
[0092] For example, resource addition tasks can be divided into execution stages such as resource application, resource binding, instance configuration, and business takeover. Based on the remaining execution time of similar tasks in each stage and the actual effective ratio of resources, the expected amount of resources that can be effective in each stage can be determined.
[0093] This ensures that the expected effectiveness factor is derived from the actual task execution data during the resource orchestration process, rather than being set independently of the cloud computing resource's operating status.
[0094] S4. Determine the stable resource deviation.
[0095] Determine the corresponding required resource quantity based on the business load and resource requirement constraints corresponding to the target analysis period. .
[0096] The required resource quantity can be determined by a pre-established correspondence between service load and resource demand, or by the resource demand strategy pre-configured for the corresponding 5G service instance.
[0097] For example, a correspondence between the number of concurrent sessions and the demand for computing resources can be established based on the historical operation data of 5G service instances. While ensuring that the service processing latency meets the preset service requirements, the required amount of computing resources can be determined based on the number of concurrent sessions corresponding to the target analysis period.
[0098] Then, the stable resource deviation is calculated based on the demanded resource quantity and the predicted stable resource quantity:
[0099]
[0100] in, For the first The first 5G service instance The stable resource deviation corresponding to the resource class.
[0101] when This indicates that even after considering adjustments to existing resources in transit, there is still a resource gap during the target analysis period.
[0102] when When, it indicates that there is a corresponding resource surplus under the predicted steady state.
[0103] when When the forecasted stable resource quantity matches the demanded resource quantity, it indicates that the forecasted stable resource quantity matches the demanded resource quantity.
[0104] For example, business instance A currently has 8 virtual computing cores in actual computing resources, and there is already an in-transit resource adjustment task to add 4 virtual computing cores. This task is expected to take effect within the target analysis period.
[0105] at this time:
[0106]
[0107] If the required computing resources for the target analysis period are also 12 virtual computing cores, then:
[0108]
[0109] Therefore, there is no need to generate new resources to add tasks.
[0110] If the workload increases further, causing the required computing resources for the target analysis period to reach 16 virtual computing cores, then:
[0111]
[0112] The system only generates an initial resource increase action for the resource gap of the remaining 4 virtual computing cores, instead of adding 8 more virtual computing cores based on the difference between the current actual resource quantity of 8 and the required resource quantity of 16.
[0113] For the same 5G service instance involving multiple resource types, the corresponding stable resource deviation amount is determined separately.
[0114] For resource types with resource gaps, the stable resource deviation can be used as the initial resource increase; for resource types with resource surplus, the absolute value of the stable resource deviation can be used as the initial resource release.
[0115] S5. Generate initial resource adjustment actions and identify resource adjustment conflicts.
[0116] The initial resource adjustment action is generated based on the stable resource deviation determined in step S4.
[0117] The initial resource adjustment action includes at least the adjustment object, resource type, adjustment direction, and initial adjustment amount.
[0118] Then, based on the adjustment target, resource type, and adjustment direction, the initial resource adjustment action is compared with the corresponding set of resource adjustments in progress.
[0119] This embodiment at least identifies duplicate adjustments, reverse adjustments, and resource usage conflicts.
[0120] Among them, duplicate adjustment refers to: for the same adjustment object and the same resource type, there is already an in-transit resource adjustment task with the same direction as the current initial resource adjustment action, and the resource adjustment amount of the in-transit resource adjustment task that has not yet taken effect can cover the adjustment amount corresponding to the current initial resource adjustment action.
[0121] For example, if the current initial resource adjustment action requires business instance A to add 2 virtual computing cores, but there are still 3 virtual computing cores in the existing resource addition tasks that have not yet taken effect, then the existing resource adjustment in progress can cover the current resource requirements. The current initial resource adjustment action is identified as a duplicate adjustment and no new resource addition task will be generated directly.
[0122] Reverse adjustment refers to an initial resource adjustment action that is opposite in direction to an existing resource adjustment task in progress, targeting the same adjustment object and the same resource type.
[0123] For example, the initial resource adjustment action corresponding to the current resource status is to release 2 virtual computing cores, while the same business instance still has a resource addition task that has not yet been completed. If the addition and release operations are performed simultaneously, the resource adjustment actions may cancel each other out or form a continuous reversal, so it is identified as a reverse adjustment.
[0124] Resource conflict refers to a situation where the resource corresponding to the initial resource adjustment action is currently locked or in a migration state due to an existing resource adjustment task in progress.
[0125] For example, if business instance A is currently migrating from node N1 to node N2, the corresponding computing resources, memory resources, or network connection resources may be locked by the migration task. If the same resource is released, added, or migrated again before the migration is complete, multiple resource control actions may compete for the same resource.
[0126] If existing resource adjustment tasks in the same direction are insufficient to fully cover current resource needs, the remaining adjustment amount can be further determined:
[0127]
[0128] in, The remaining adjustment amount, This is the initial adjustment amount obtained based on the stable resource deviation. This refers to the amount of resource adjustments that can be covered by existing resource adjustment tasks that are in transit in the same direction.
[0129] For example, if 5 additional computing resource units are needed, and there are already resource adjustment tasks in the same direction that can cover 3 of these resource units, then the remaining adjustment amount is 2 resource units, and subsequent resource adjustments will only be made for these 2 resource units.
[0130] S6. Generate candidate resource adjustment plans and determine the target resource adjustment plan.
[0131] One or more candidate resource adjustment schemes are generated based on the initial resource adjustment action, the remaining adjustment amount, and the resource adjustment conflict identification results.
[0132] For initial resource adjustment actions that do not have resource adjustment conflicts, the corresponding resource increase task or resource release task can be generated directly based on the remaining adjustment amount.
[0133] For resource adjustment actions identified as duplicate adjustments, waiting for existing resource adjustment tasks in transit to continue execution can be considered as a candidate solution, without generating new resource adjustment tasks in the same direction.
[0134] In cases where reverse adjustments are possible, different candidate processing methods can be generated based on the current execution stage of the task and the available resources.
[0135] For example, when an existing resource adjustment task is still in the stage where it can be modified or canceled, the amount of resources to be adjusted in the part of the original resource adjustment task that has not yet been executed can be adjusted; when an existing resource adjustment task has entered the stage where it cannot be canceled and takes effect, the resource deviation can be recalculated based on the updated actual resource status after it is completed.
[0136] In cases of resource conflict, the existing task causing the resource lockout can be allowed to complete; or, if other cloud computing nodes that meet the resource requirements exist, the remaining resource requirements can be mapped to other available nodes.
[0137] For example, if business instance A still has a resource gap of 4 virtual computing cores under the predicted stable state, the following candidate solutions can be formed: Candidate solution 1: wait for the existing resource addition task in transit to be completed before adding the remaining 4 virtual computing cores; Candidate solution 2: if the target adjustment amount of the existing resource addition task in transit is allowed to be modified, directly increase the target resource amount corresponding to the task; Candidate solution 3: allocate the remaining computing resource demand to other cloud computing nodes that can support the corresponding 5G business instance.
[0138] To further determine the target resource adjustment scheme from the candidate schemes, the following scheme evaluation function can be used to determine the... Evaluation values of the candidate resource adjustment schemes:
[0139]
[0140] in, For the first The evaluation value of each candidate resource adjustment plan; The improvement in stable resource deviation after implementing the candidate resource adjustment scheme relative to the state before implementation; The resource adjustment overhead required to execute the candidate resource adjustment scheme; The expected effective time delay for the resource adjustment actions in the candidate resource adjustment schemes to reach the corresponding predicted stable state; , , For the corresponding weights.
[0141] The improvement in stable resource deviation can be determined based on the change in the absolute value of stable resource deviation before and after the implementation of the candidate scheme.
[0142] For 5G service instances involving multiple resource types, it can be represented as:
[0143]
[0144] in, This indicates the first [number] candidate resource adjustment plan before implementation. Stable resource deviation; Indicates execution of the first The expected outcome after the first candidate resource adjustment plan is the [number]th [project / plan]. Stable resource deviation; Indicates the first Resource importance coefficient corresponding to a resource class.
[0145] Resource adjustment overhead The amount of resources to be adjusted can be determined based on one or more of the following: the total amount of resources to be adjusted, the number of cloud computing nodes involved, the amount of business migration data, and the number of resource control operations.
[0146] For example, in a resource addition scheme, the resource adjustment overhead can be determined based on the number of virtual computing cores added, the amount of memory resources, and the number of resource reconfiguration operations between nodes; in a migration scheme, the amount of service status data that needs to be transmitted in the network can also be further considered.
[0147] Expected effective date The timeframe is determined based on the estimated execution time of each resource adjustment task in the candidate resource adjustment plan.
[0148] For multiple resource adjustment tasks that can be executed in parallel, the longest estimated completion time can be used as the estimated effective time of the candidate solution; for multiple resource adjustment tasks with sequential execution dependencies, the estimated execution time can be accumulated according to the execution order of each task.
[0149] Before evaluating candidate solutions, those that cannot meet the corresponding resource requirement constraints or still have unresolved resource adjustment conflicts are first eliminated. Then, the evaluation value of the remaining candidate solutions is selected. The largest candidate resource adjustment plan is taken as the target resource adjustment plan.
[0150] By using "meeting resource demand constraints" and "not having unresolved resource adjustment conflicts" as admission criteria for candidate solutions, we can avoid selecting solutions with high resource returns but actual resource conflicts by simply relying on weighted evaluation values.
[0151] S7. Execute the target resource adjustment plan and update the resource status.
[0152] Based on the target resource adjustment plan determined in step S6, corresponding resource control instructions are generated and sent to the cloud resource orchestration execution component.
[0153] Resource control commands can be used to add or release computing resources, adjust memory resources, reconfigure network transmission resources, migrate business instances, and adjust the resource configuration of corresponding virtual machines or containers.
[0154] The resource adjustment tasks corresponding to the target resource adjustment plan are added to the corresponding in-transit resource adjustment set after the control command is issued but before the actual execution is completed.
[0155] The resource orchestration execution component provides feedback on the task execution status to the resource allocation system during the execution of resource adjustment tasks.
[0156] When the resource adjustment task is partially completed, update the actual and non-effective resource adjustment amounts based on the execution status feedback, and re-determine the corresponding expected effectiveness coefficients and predicted stable resource amounts during the subsequent target analysis period.
[0157] When the resource adjustment task is completed, the corresponding resource adjustment result will be updated to the current actual resource status, and the resource adjustment task will be removed from the corresponding in-transit resource adjustment set.
[0158] When a resource adjustment task fails or is canceled, the task is removed from the in-transit resource adjustment set, and the impact of its unused resource adjustments on the subsequent prediction of stable resource quantities is canceled.
[0159] This forms a closed-loop resource allocation process: "acquiring current resource status and in-transit resource status, determining predicted stable resource status, calculating stable resource deviation, generating initial resource adjustment actions, identifying resource adjustment conflicts, generating and optimizing candidate solutions, executing resource adjustments, and updating actual and in-transit statuses."
[0160] Specific application examples
[0161] Taking 5G service instance A deployed on cloud computing node N1 as an example, the target resource type is explained using computing resources as an example.
[0162] At the current scheduling moment, assume that business instance A is currently actually allocated 9 computing resource units.
[0163] Based on the workload and resource requirements of business instance A, it is determined that business instance A requires 14 computing resource units during the target analysis period.
[0164] Meanwhile, a query of the resource orchestration platform revealed that a task to increase computing resources had been generated for business instance A before the current time. This task initially planned to add 4 computing resource units, of which 1 computing resource unit has been configured and included in the current actual resource status. Therefore, the current actual resource quantity has been updated from the original 8 computing resource units to 9 computing resource units.
[0165] Three computing resource units of this resource adjustment task have not yet taken effect.
[0166] The resource orchestration execution component reports that the task is currently in the resource binding phase, and the estimated remaining execution time is less than the remaining time of the target analysis period. Therefore, it can be determined that:
[0167]
[0168] The corresponding predicted stable resource quantity is:
[0169]
[0170] The required resource quantity for the target analysis period is 14, therefore the stable resource deviation is:
[0171]
[0172] Therefore, even though the current actual resource quantity is only 9 computing resource units, considering the existing tasks in progress, the expected stable resource quantity can reach 12 computing resource units. Thus, the actual unmet resource gap is only 2 computing resource units.
[0173] The system generates an initial resource adjustment action that adds 2 computing resource units, instead of directly adding 5 computing resource units based on the difference between 14 and the current 9.
[0174] The initial resource adjustment action is then compared with the set of resource adjustments in progress.
[0175] Since the three computing resource units corresponding to the original tasks that were added in transit but had not yet taken effect have been included in the predicted stable resource quantity, the two computing resource units form a new remaining adjustment quantity.
[0176] The system can generate different candidate solutions, such as waiting for the original resource addition task to complete before adding 2 computing resource units, directly modifying the target resource amount of the original resource addition task, or allocating the additional 2 computing resource units to other available cloud computing nodes.
[0177] For each candidate solution, determine the amount of stable resource deviation improvement, resource adjustment overhead, and expected effective delay, and exclude solutions with unresolved resource adjustment conflicts or those that cannot meet business resource requirements.
[0178] Then, the evaluation value of each valid candidate solution is calculated according to the following evaluation function, and the candidate solution with the highest evaluation value is selected as the target resource adjustment solution:
[0179]
[0180] For example, if the original resource increase task allows adjustment of the target resource amount during execution, and modifying the original task has less resource control overhead and a shorter expected effective time delay compared to re-establishing the resource increase task, the original task can be modified to increase its remaining adjustment target from 3 computing resource units to 5 computing resource units.
[0181] While the modified resource adjustment task is still incomplete, it will continue to be recorded as a resource adjustment task in progress.
[0182] After all resource adjustment tasks are completed, the current actual computing resource quantity of business instance A is updated to 14 computing resource units, and the corresponding resource adjustment task is removed from the in-transit resource adjustment set.
[0183] In the next resource scheduling cycle, the system redetermines the predicted stable resource quantity based on the updated actual resource status and the adjustment status of the remaining resources in transit, thereby continuously completing the dynamic allocation of 5G network resources.
[0184] Furthermore, this embodiment is not limited to a single computing resource. When the same 5G service instance involves computing resources, memory resources, storage resources, and network transmission resources, corresponding in-transit resource adjustment sets, predicted stable resource quantities, and stable resource deviation quantities can be established for different resource types, and a resource adjustment scheme can be jointly determined by combining the resource demand constraints of each resource type.
[0185] For different cloud computing platforms and 5G service deployment environments, the specific mapping relationship between service load and required resource quantity, the method for determining the expected effective resource quantity, and the weights in the scheme evaluation function can be set or adjusted according to the actual service operation characteristics. As long as the resource adjustment status that has not yet been completed is included in the subsequent resource status prediction during the resource allocation process, and the subsequent resource adjustment scheme is determined based on the predicted stable resource status, the technical solution of this invention can be adopted.
[0186] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.
Claims
1. A method for intelligent resource allocation of 5G network in cloud computing environment, characterized in that, Includes the following steps: S1. Obtain the service load, resource requirement constraints and current resource occupancy status of multiple 5G service instances in the cloud computing environment, and obtain the current resource status of the cloud computing node carrying the 5G service instances and the resource adjustment tasks that have not yet been completed. S2. Based on the adjustment object, resource type, adjustment direction, adjustment resource quantity and task execution status of each resource adjustment task, determine the resource adjustment quantity that has not yet taken effect, and establish a corresponding in-transit resource adjustment set according to the adjustment object and resource type. S3. For any target resource type of any 5G service instance, determine the predicted stable resource quantity based on the current actual resource quantity and the expected increase and release of resources in the in-transit resource adjustment set during the target analysis period: wherein, is the predicted stable resource amount of the first type of resource of the i-th 5G service instance, is the predicted stable resource amount of the first type of resource of the i-th 5G service instance, is the current actual resource amount of the first type of resource of the i-th 5G service instance, is the current actual resource amount of the first type of resource of the i-th 5G service instance, and are respectively a corresponding in-transit resource increase task set and an in-transit resource release task set; and are respectively a resource adjustment amount of a corresponding task that has not yet taken effect, and are respectively a predicted effective coefficient of a corresponding resource adjustment task within the target analysis period. S4. Determine the required resource quantity for the target analysis period based on the business load and resource demand constraints, and determine the stable resource deviation based on the required resource quantity and the predicted stable resource quantity: in, For the first The first 5G service instance for the first The required amount of resources for this type of resource; when When it is determined that a corresponding resource gap exists under the predicted steady state, when At that time, it is determined that there is a corresponding resource surplus under the predicted steady state; S5. Generate an initial resource adjustment action based on the stable resource deviation amount, and compare the initial resource adjustment action with the set of resource adjustments in transit according to the adjustment object, resource type and adjustment direction, identify duplicate adjustment, reverse adjustment or resource occupation conflict, and determine the remaining adjustment amount that has not been covered by existing in-transit resource adjustments in the same direction. S6. Generate multiple candidate resource adjustment schemes based on the initial resource adjustment action, remaining adjustment amount, and resource adjustment conflict identification results, and determine the target resource adjustment scheme from the candidate resource adjustment schemes that meet the resource demand constraints and have no unresolved resource adjustment conflicts. S7. Perform resource adjustments on the corresponding 5G service instances or cloud computing nodes according to the target resource adjustment scheme, and update the current resource status and the set of resource adjustments in transit according to the execution status of the resource adjustment task, so that subsequent resource allocation is based on the updated resource status.
2. The intelligent resource allocation method for 5G networks in a cloud computing environment according to claim 1, characterized in that, The resource types include at least one of computing resources, memory resources, storage resources, and network transmission resources; the resource adjustment tasks include at least one of resource addition tasks, resource release tasks, and resource migration tasks, wherein the resource migration tasks form corresponding resource release adjustments and resource addition adjustments according to the migration source node and the migration target node, respectively.
3. The intelligent resource allocation method for 5G networks in a cloud computing environment according to claim 1, characterized in that, In step S3, the task stage identifier, completed resource adjustment amount, and estimated remaining execution time of the resource adjustment task fed back by the resource orchestration execution component are obtained. The estimated effective adjustment amount within the target analysis period is determined based on the task stage identifier, completed resource adjustment amount, and estimated remaining execution time. The estimated effective coefficient is determined based on the estimated effective adjustment amount and the corresponding resource adjustment amount that has not yet taken effect. When the estimated remaining execution time of the resource adjustment task is not greater than the remaining time of the target analysis period, the corresponding estimated effectiveness coefficient will be determined as 1.
4. The intelligent resource allocation method for 5G networks in a cloud computing environment according to claim 1, characterized in that, In step S4, the stable resource deviation amount is determined for different resource types of the same 5G service instance. Based on each stable resource deviation amount, the resource shortage type and resource surplus type of the 5G service instance in the predicted stable state are determined. The stable resource deviation amount corresponding to the resource type with resource shortage is used as the initial resource increase amount, and the absolute value of the stable resource deviation amount corresponding to the resource type with resource surplus is used as the initial resource release amount.
5. The intelligent resource allocation method for 5G networks in a cloud computing environment according to claim 1, characterized in that, In step S5: When there are in-transit resource adjustment tasks for the same adjustment object and the same resource type with the same direction as the initial resource adjustment action, and the resource adjustment amount of the in-transit resource adjustment task that has not yet taken effect can cover the adjustment amount corresponding to the initial resource adjustment action, it is identified as a duplicate adjustment; When an initial resource adjustment action targeting the same adjustment object and the same resource type is in the opposite direction to an ongoing resource adjustment task, it is identified as a reverse adjustment. When the resource corresponding to the initial resource adjustment action is in a locked state or a migration state of an existing in-transit resource adjustment task, it is identified as a resource occupation conflict.
6. The intelligent resource allocation method for 5G networks in a cloud computing environment according to claim 1, characterized in that, In step S6, for initial resource adjustment actions with resource adjustment conflicts, candidate resource adjustment schemes are generated by at least one of the following methods: waiting for the corresponding in-transit resource adjustment task to complete, modifying the unexecuted portion of the resource adjustment amount in the corresponding in-transit resource adjustment task, or generating a supplementary resource adjustment task for the remaining adjustment amount. The target resource adjustment scheme is determined from the candidate resource adjustment schemes that meet the resource demand constraints by comparing the degree of improvement in stable resource deviation, resource adjustment overhead, and expected effective delay of each candidate resource adjustment scheme.
7. The intelligent resource allocation method for 5G networks in a cloud computing environment according to claim 6, characterized in that, The first is determined according to the following formula. Evaluation values of the candidate resource adjustment schemes: in, For the first The evaluation value of each candidate resource adjustment plan. The improvement in stable resource deviation after implementing the candidate resource adjustment scheme relative to the state before implementation. The resource adjustment overhead required to execute the candidate resource adjustment scheme, The estimated effective time delay for the resource adjustment actions in the candidate resource adjustment schemes to reach the corresponding predicted stable state. , , As corresponding weights; from the candidate resource adjustment schemes that meet the resource demand constraints and have no unresolved resource adjustment conflicts, select the candidate resource adjustment scheme with the largest scheme evaluation value as the target resource adjustment scheme.
8. The intelligent resource allocation method for 5G networks in a cloud computing environment according to claim 1, characterized in that, In step S7, when the resource adjustment task is partially completed, the corresponding unused resource adjustment amount is updated according to the execution status feedback; when the resource adjustment task is completed, the corresponding resource adjustment result is updated to the current resource status, and the resource adjustment task is removed from the in-transit resource adjustment set; when the resource adjustment task fails or is canceled, it is removed from the in-transit resource adjustment set, and the impact of its unused resource adjustment amount on the predicted stable resource amount is canceled.
9. A 5G network intelligent resource allocation system under a cloud computing environment, characterized in that, include: The status acquisition module is used to acquire the service load, resource requirement constraints and current resource occupancy status of multiple 5G service instances, as well as the current resource status of cloud computing nodes and resource adjustment tasks that have not yet been completed. The in-transit adjustment determination module is used to determine the amount of resource adjustments that have not yet taken effect and establish an in-transit resource adjustment set based on the adjustment object, resource type, adjustment direction, amount of adjusted resources and task execution status of the resource adjustment task. The stable state determination module is used to determine the predicted stable resource quantity based on the current actual resource quantity and the expected increase and release of resources in the in-transit resource adjustment set. The resource deviation determination module is used to determine the stable resource deviation based on the resource demand and predicted stable resource quantity of the 5G service instance, and generate an initial resource adjustment action based on the stable resource deviation. The conflict identification module is used to compare the initial resource adjustment action with the set of resource adjustments in progress, identify duplicate adjustments, reverse adjustments or resource occupation conflicts, and determine the remaining adjustment amount; The resource allocation module is used to generate candidate resource adjustment schemes based on the initial resource adjustment action, the remaining adjustment amount, and the resource adjustment conflict identification results, and to determine the target resource adjustment scheme from the candidate resource adjustment schemes that meet the resource demand constraints and have no unresolved resource adjustment conflicts. The execution update module is used to perform resource adjustments according to the target resource adjustment plan, and update the current resource status and the set of resource adjustments in progress based on the execution status of the resource adjustment tasks.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method described in any one of claims 1 to 8.
