Computing power comprehensive scheduling method and system for cloud computing

Abstract

The invention discloses a computing power comprehensive scheduling method and system for cloud computing, and relates to the technical field of computing power scheduling, and the method comprises the steps: determining the task priority of a cloud computing task, preliminarily generating a mapping matrix, and respectively defining task preference and virtual machine preference by using a bilateral preference game mechanism, a stable matching algorithm is used for preliminarily matching the tasks with the virtual machines, idle fragment resources of the virtual machines are determined, a rhododendron behavior scheduling algorithm is used, the scheduling probability of unmatched tasks is calculated, task matching is updated, and an egg replacement mechanism is used for replacement again to match the tasks for local optimization. According to the method, through matching, parasitism and replacement chained control, full-process control from static gaming to dynamic fault tolerance is completed, the task scheduling success rate, the task entering density and the unit utilization efficiency of resources are improved, and the overall scheduling failure rate and the resource vacancy rate of the system are reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of computing power scheduling, and particularly to a comprehensive computing power scheduling method and system for cloud computing. Background Art

[0002] With the expansion of the scale of data centers and the increase in the complexity of computing tasks, traditional static scheduling methods have become difficult to handle problems such as multi-task concurrency, high dynamic load, and resource heterogeneity. Currently, a variety of optimization algorithms have been introduced into the cloud computing scheduling field, including a task grading mechanism based on priority, a bilateral matching algorithm under a game theory model, and a bio-inspired meta-heuristic algorithm, which have improved the intelligence and adaptability of scheduling to a certain extent;

[0003] However, in the real scenario where the multi-objective scheduling requirements are becoming more complex and the computing resources are severely fragmented, the existing technologies still face problems such as coarse scheduling granularity, high resource vacancy rate, large task response delay, and weak system convergence ability. The existing cloud resource scheduling technologies generally have the following deficiencies in task matching strategies and resource management mechanisms: First, the scheduling process often adopts a one-way mapping mechanism, lacking a preference interaction model between tasks and resources, and it is difficult to adapt to the multi-dimensional differences between task computing requirements and resource supply; Second, fragmented resources are not fully utilized, resulting in difficult deployment of medium and low load tasks and a decrease in resource utilization; Third, most scheduling optimization algorithms stay at the initial mapping stage, lacking an iterative optimization and task replacement mechanism, and it is difficult to break through the local optimal bottleneck; Fourth, the secondary scheduling mechanism for edge tasks and failed tasks is insufficient, and the resource recycling and reallocation mechanism is not perfect. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a comprehensive computing power scheduling method for cloud computing to solve the following deficiencies commonly existing in the existing cloud resource scheduling technologies in task matching strategies and resource management mechanisms: First, the scheduling process often adopts a one-way mapping mechanism, lacking a preference interaction model between tasks and resources, and it is difficult to adapt to the multi-dimensional differences between task computing requirements and resource supply; Second, fragmented resources are not fully utilized, resulting in difficult deployment of medium and low load tasks and a decrease in resource utilization; Third, most scheduling optimization algorithms stay at the initial mapping stage, lacking an iterative optimization and task replacement mechanism, and it is difficult to break through the local optimal bottleneck; Fourth, the secondary scheduling mechanism for edge tasks and failed tasks is insufficient, and the resource recycling and reallocation mechanism is not perfect.

[0006] To solve the above technical problems, the present invention provides the following technical solutions:

[0007] In the first aspect, the present invention provides a comprehensive computing power scheduling method for cloud computing, which includes:

[0008] Determine the task priority of cloud computing tasks, initially generate a mapping matrix, use the bilateral preference game mechanism to define task preferences and virtual machine preferences respectively, and initially match tasks and virtual machines using the stable matching algorithm. Determine the idle fragment resources of virtual machines, use the cuckoo behavior scheduling algorithm, calculate the scheduling probability of unmatched tasks and update task matching, and use the egg replacement mechanism to re-replace to match tasks for local optimization;

[0009] Construct a task set to be optimized, use the virtual machine expansion decision mechanism to expand virtual machines, construct a scheduling candidate model, calculate the total objective cost function of different scheduling candidate solutions, and use the crow search algorithm for iterative calculation to output an optimized scheduling candidate solution;

[0010] Update the mapping matrix, record the real-time parameters of virtual machines, and output log data;

[0011] Perform encrypted transmission according to the log data.

[0012] As a preferred solution of the computing power comprehensive scheduling method for cloud computing described in the present invention, wherein: the matching of tasks and virtual machines and the use of the egg replacement mechanism to re-replace to match tasks for local optimization include,

[0013] Divide the computing tasks of cloud computing, and determine the resource consumption of each computing task, including the task deadline, the computing demand value, and the resource consumption value, including the memory occupancy value and the bandwidth usage value;

[0014] Set a virtual machine set, clarify the memory capacity and network bandwidth for each virtual machine, and determine the computing power value. Based on the task deadline and computing demand value, determine the task priority, and randomly assign an initial virtual machine to each task in the resource pool to obtain an initial mapping matrix A, and use the bilateral preference game mechanism to define task preferences and virtual machine preferences respectively;

[0015] Among them, the task preference defines the ratio of the descending order of the virtual machine computing power value to the task computing priority number as the task preference list, and the virtual machine preference defines the ratio of the descending order of the task computing priority number to the task computing demand value as the virtual machine preference list;

[0016] Use the stable matching algorithm Gale-Shapley, let the virtual machine accept in turn according to the proposal ranking, cumulatively calculate the computing demand value of the accepted proposal tasks. When the current cumulative computing demand value is greater than or equal to the computing power value of the virtual machine, reject the proposal task; Cumulatively calculate the resource consumption values of the accepted proposal tasks, including the memory occupancy value and the bandwidth usage value. When the current cumulative memory occupancy value and bandwidth usage value are respectively greater than or equal to the memory capacity and network bandwidth of the virtual machine, reject the proposal task;

[0017] The tasks of the rejected proposals propose the next virtual machine according to the task preference list, and loop to complete the preliminary task matching until the remaining unmatched tasks have clearly exhausted the proposals of all virtual machines in the preference list;

[0018] Define the idle resource fragment set of the virtual machine according to the usage fragment time period of the virtual machine. For the tasks that have not been successfully matched, perform task simulation parasitism judgment. If the computing requirement value, memory occupancy value, and bandwidth usage value of the task to be hatched are respectively less than the total computing power value, memory capacity, and network bandwidth that the fragment segment can provide, it can be used as the simulation object of the corresponding virtual machine fragment segment and combined into the task set to be hatched;

[0019] Use the scheduling algorithm of the cuckoo behavior, calculate the probability of each task to be hatched being successfully scheduled to the idle resource fragment through the probability control mechanism, perform Bernoulli experiments based on the probability value of being scheduled to the idle resource fragment, update the idle resource fragment, and remove the task from the task set to be hatched;

[0020] If the idle resource fragment cannot carry the task to be hatched, use the egg replacement mechanism to implement task replacement, update the task matching status and the idle resource fragment set, and loop to complete the task replacement.

[0021] As a preferred solution of the computing power comprehensive scheduling method for cloud computing described in the present invention, wherein: the virtual machine expansion includes,

[0022] Construct a task set to be optimized, including tasks that have not been successfully bilaterally matched and have not passed the task simulation parasitism judgment, tasks to be hatched that have not been able to perform the egg replacement mechanism, and matching tasks replaced by the egg replacement mechanism;

[0023] Use the virtual machine expansion decision mechanism to determine the increased value of virtual machine expansion based on historical data;

[0024] Mark the virtual machines whose three idle parameters are all greater than the three parameters required by the task to be optimized as candidate virtual machines.

[0025] As a preferred solution of the computing power comprehensive scheduling method for cloud computing described in the present invention, wherein: the construction of the scheduling candidate model and the output of the optimized scheduling candidate solution include,

[0026] Construct a scheduling candidate model according to the scheduling order of the task set to be optimized and the candidate virtual machines, and the virtual machine numbers assigned to the task to be optimized in its candidate virtual machine set;

[0027] Define the total objective cost function of this scheduling solution as the sum of the maximum completion time, the imbalance degree of virtual machine load, and the scheduling failure rate;

[0028] The Crow Search Algorithm (CSA) is used to perform iterative calculations based on the total objective cost function value of each scheduling candidate solution. If the total objective cost function value is less than or equal to the total objective cost function value of the previous iteration, it is memorized and repeated iterations are performed;

[0029] Iteration is stopped until the calculation loss of the total objective cost function value no longer changes significantly during consecutive iterations, and the optimal scheduling candidate solution is output.

[0030] As a preferred solution of the computing power comprehensive scheduling method for cloud computing described in the present invention, wherein: the updating of the mapping matrix includes,

[0031] Based on the optimal scheduling candidate solution, record the order in which each task in the task set to be optimized is assigned to the virtual machine, and write the task-virtual machine allocation relationship into the mapping matrix for updating.

[0032] As a preferred solution of the computing power comprehensive scheduling method for cloud computing described in the present invention, wherein: the recording of the real-time parameters of the virtual machine and outputting of log data includes,

[0033] Based on the updated mapping matrix, update the computing virtual machine parameters including real-time load, memory usage value, and bandwidth status, and record and update the load value, free memory, and free bandwidth of the virtual machine in real time to generate virtual machine real-time log data.

[0034] As a preferred solution of the computing power comprehensive scheduling method for cloud computing described in the present invention, wherein: the encrypted transmission according to the log data includes,

[0035] Use the symmetric encryption AES-256 technology to encrypt and package the log, transmit the log file to the cloud storage through the SFTP method, and perform consistency verification at the receiving end to generate a verification log for storage.

[0036] In a second aspect, the present invention provides a system for a computing power comprehensive scheduling method for cloud computing, including,

[0037] A priority evaluation module that analyzes cloud computing task parameters to determine task priorities;

[0038] A bilateral preference module that constructs a preference sequence for tasks and virtual machines to perform preliminary task-virtual machine matching;

[0039] A fragmentation recognition module that performs a scheduling probability evaluation on unmatched tasks to update the task matching result;

[0040] A virtual machine expansion module that collects all uncompleted scheduling tasks and calculates the virtual machine expansion increment;

[0041] Construction of a scheduling candidate model that generates a candidate solution structure based on the task scheduling order and virtual machine allocation;

[0042] The scheduling optimization module uses the crow search algorithm to iteratively search for the optimal scheduling order and task allocation;

[0043] The resource status synchronization module updates the optimal scheduling result to the mapping matrix and records data in real time;

[0044] The log encryption and transmission module generates encrypted log files, performs secure transmission, and conducts consistency verification.

[0045] In a third aspect, the present invention provides a computer device, including a memory and a processor, where the memory stores a computer program, and: when the computer program is executed by the processor, any step of the computing power comprehensive scheduling method for cloud computing as described in the first aspect of the present invention is implemented.

[0046] In a fourth aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored, and: when the computer program is executed by the processor, any step of the computing power comprehensive scheduling method for cloud computing as described in the first aspect of the present invention is implemented.

[0047] The beneficial effects of the present invention are as follows: Through the chain control of matching, parasitism, and replacement, the full-process control from static game to dynamic fault tolerance is completed, improving the success rate of task scheduling, the task occupancy density, and the unit utilization efficiency of resources, reducing the overall system scheduling failure rate and resource idle rate. By preferentially matching high-priority tasks, using fragments to carry lightweight tasks, and performing local replacement based on the task priority and load ratio, the hierarchical injection of tasks in the resource space is realized, enabling CSA to no longer handle the global task allocation problem, but focus on the task corner scheduling in complex states, enabling CSA to construct scheduling feasibility constraints based on the real resource fragment structure in the scheduling candidate modeling, greatly enhancing the physical deployability of the solution, reducing the number of invalid search solutions, and increasing the effective search density of the algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0049] Figure 1 It is a flowchart of the computing power comprehensive scheduling method for cloud computing in Embodiment 1.

[0050] Figure 2 It is a structural diagram of the computing power comprehensive scheduling system for cloud computing in Embodiment 1.

[0051] Figure 3Schematic diagram of the task matching process for the computing power comprehensive scheduling method of cloud computing in Embodiment 1.

[0052] Figure 4 Schematic diagram of the optimized scheduling process for the computing power comprehensive scheduling method of cloud computing in Embodiment 1. Detailed implementation manners

[0053] To make the above objects, features, and advantages of the present invention more obvious and understandable, the following detailed description of the specific implementation manners of the present invention will be made in conjunction with the accompanying drawings of the specification.

[0054] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without departing from the spirit of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0055] Secondly, the so-called "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation manner of the present invention. The appearances of "in one embodiment" in different places in this specification do not all refer to the same embodiment, nor are they separate or alternative embodiments that are mutually exclusive of other embodiments.

[0056] Embodiment 1, referring to Figures 1 to 4 , is the first embodiment of the present invention. This embodiment provides a computing power comprehensive scheduling method for cloud computing, including the following steps:

[0057] S1. Determine the task priority of the cloud computing task, initially generate a mapping matrix, use the bilateral preference game mechanism to define task preferences and virtual machine preferences respectively, and initially match tasks and virtual machines using the stable matching algorithm. Determine the virtual machine idle fragment resources, use the cuckoo behavior scheduling algorithm, calculate the scheduling probability of unmatched tasks and update task matching, and use the egg replacement mechanism to re-replace to match tasks for local optimization;

[0058] Preferably, when matching tasks and virtual machines and using the egg replacement mechanism to re-replace to match tasks for local optimization, it includes

[0059] Divide the computing tasks of cloud computing, and determine the resource consumption of each computing task, including the task deadline (unit: ms), the computing demand value (the specific computing resources required, unit: MI value), and the resource consumption value (clearly recorded according to the task type when the task is created, including the memory occupancy value (unit: MB), the bandwidth usage value (unit: Kbps));

[0060] Set a set of virtual machines, specify the memory capacity (in MB) and network bandwidth (in Kbps) for each virtual machine, and use the CloudSim-like simulation framework to determine the computing power value of the virtual machine based on the product of the number of CPU cores of the virtual machine, the number of instructions that each CPU core can execute per millisecond, and the virtual machine task scheduling time window;

[0061] Determine the task priority based on the deadline and computing demand value of the task, expressed as:

[0062]

[0063] where T i represents the computing priority number of the i-th task, De represents the deadline of the task, and Co represents the computing demand value of the task;

[0064] Randomly assign an initial virtual machine to each task in the resource pool to obtain an initial mapping matrix A, expressed as:

[0065] A = [a ji b×n ,a ji = {0,1};

[0066] where a ji represents the mapping relationship between the i-th task and the j-th virtual machine. Specifically, if the i-th task is mapped to the j-th virtual machine, then a ji is defined as 1, otherwise 0. At the same time, define the constraint condition that each task is mapped to only one virtual machine. b represents the total number of virtual machines, and n represents the total number of tasks;

[0067] Use the two-sided preference game mechanism to define task preferences and virtual machine preferences respectively;

[0068] Among them, the task preference is defined as (the computing task tends to select a virtual machine with a higher computing power value, and considering the resource requirements of the virtual machine for the computing priority number of the task), the ratio of the descending-ordered virtual machine computing power value to the computing priority number of the task is used as the task preference list, and the virtual machine preference is defined as (the virtual machine tends to be affected by tasks with high task priorities but low resource requirements), the ratio of the descending-ordered computing priority number of the task to the computing demand value of the task is used as the virtual machine preference list;

[0069] Use the stable matching algorithm Gale-Shapley. Select the task with the highest task priority from the set of unmatched tasks, and make task proposals to the virtual machines that have not been applied for by this task according to the task preference list. After the virtual machine receives the task proposal, it sorts the proposed tasks according to the virtual machine preference list, and the virtual machine accepts them in turn according to the proposal sorting;

[0070] ​Cumulatively calculate the computing requirement value for the accepted proposal tasks. When the currently cumulative computing requirement value is greater than or equal to the computing capacity value of the virtual machine, reject the proposal task;

[0071] The cumulative resource consumption value for the accepted proposal tasks includes the memory occupancy value and the bandwidth usage value. When the currently cumulative memory occupancy value and bandwidth usage value are respectively greater than or equal to the memory capacity and network bandwidth of the virtual machine, reject the proposal task;

[0072] The tasks of the rejected proposals propose to the next virtual machine according to the task preference list, and loop to execute to complete the preliminary task matching until the remaining unmatched tasks have clearly exhausted the proposals of all virtual machines in the preference list;

[0073] (The bilateral matching process is a static resource allocation decision based on the task level, which does not cover the resource utilization problem in the time dimension, and it is very likely to have resource fragmentation on the time axis)

[0074] Define the idle resource fragmentation set of the virtual machine according to the fragmented time period of the virtual machine's use, expressed as:

[0075]

[0076] Among them, F j represents the set of available idle resource fragments on the j-th virtual machine, and each element is a resource fragment segment f jk , representing the k-th fragmented resource segment on the j-th virtual machine, and respectively represent the start time and end time of the k-th fragment segment of the virtual machine j, and res jk represents the total computing capacity value that the fragment segment can provide;

[0077] Perform task simulation parasitism judgment for the tasks that have not been successfully matched. If the computing requirement value, memory occupancy value, and bandwidth usage value of the task to be hatched are respectively less than the total computing capacity value, memory capacity, and network bandwidth that the fragment segment can provide, it can be used as the simulation object of the corresponding virtual machine fragment segment and combined into the task set to be hatched;

[0078] Use the Cuckoo Scheduling Algorithm of the cuckoo behavior to calculate the probability that each task to be hatched is successfully scheduled to the idle resource fragment through the probability control mechanism, expressed as:

[0079]

[0080] Among them, P su (i) represents the scheduling probability of the i-th task to be hatched, CR i represents the computing requirement value of the i-th task to be hatched, T ino Indicates the task priority of the i-th task to be incubated;

[0081] A Bernoulli experiment is performed based on the probability value of scheduling to the idle resource fragment. If successful, the corresponding task to be hatched is scheduled to the corresponding idle resource fragment, the idle resource fragment is updated, and the task is removed from the set of tasks to be hatched;

[0082] If the idle resource fragments cannot carry the tasks to be hatched, the egg replacement mechanism is used to compare the computing demand value and computing priority number of the matched tasks. If the computing demand value of the task to be hatched is less than the computing demand value of the matched tasks, and the computing priority number of the task to be hatched is greater than the computing priority number of the matched tasks, the computing capacity value of the idle resource fragments of the virtual machine after the task to be hatched replaces the corresponding matched tasks is calculated;

[0083] If the updated fragment computing power value is greater than the fragment computing power value before the update, the task replacement is implemented, the task matching state and the idle resource fragment set are updated, and the task replacement is completed in a loop.

[0084] The introduction of task priority makes the scheduling scheme no longer just based on the rough allocation of task volume, but is controlled according to the priority level composed of deadline and computing demand, ensuring that high-urgency tasks are scheduled first, thereby improving the overall task completion rate. Through the bilateral preference game between tasks and virtual machines, a matching model of mutual game and mutual concession is established, so that task scheduling is no longer a one-way static allocation, but forms an optimal game structure based on preference evaluation, taking into account the local load control of virtual machine resources and the global perception tendency of individual tasks to resources, thereby improving the robustness and explainability of scheduling behavior, avoiding the tilt caused by centralized resource allocation, and accurately identifying the idle resources in the time period on the virtual machine, so that the subsequent scheduling has the ability to reuse resources based on the time dimension and computing power structure, thereby reducing the resource fragmentation rate and improving the continuity and compressibility of resource use;

[0085] The introduced "simulated parasitism + hatching probability control" mechanism breaks through the rigid logic of traditional matching failure and queuing, so that unmatched tasks have the parasitic opportunity of "penetrating virtual resources with a smaller load". Through the scheduling probability model of task priority and load volume coordinated control, it guides high-priority and low-load tasks to obtain fragmented resources first, improves the fine-grained scheduling ability of the system under fragmented idle resources, and reduces the total task queue length. Through the egg replacement mechanism, a local structural scheduling correction mechanism is constructed, allowing the scheduling system to "interrupt matched tasks with low priority and high load" under the guarantee of resource constraints, so as to achieve a dynamic correction strategy with the best overall benefit.

[0086] Through the chain control of matching, parasitism, and replacement, the full-process control from static game to dynamic fault tolerance is completed, enabling the scheduling system to have a closed loop of "first scheduling → failed task compensation → local deviation correction and optimization", greatly improving the task scheduling success rate, task occupancy density, and unit utilization efficiency of resources, reducing the overall system scheduling failure rate and resource idle rate, especially having strong availability and effectiveness in scenarios with a large task scale, limited number of virtual machines, and significant differences in task resource requirements.

[0087] S2. Construct a task set to be optimized, use the virtual machine expansion decision mechanism to expand the virtual machine, construct a scheduling candidate model, calculate the total objective cost function of different scheduling candidate solutions, use the crow search algorithm for iterative calculation, and output the optimized scheduling candidate solution;

[0088] Preferably, construct a task set to be optimized, including tasks where bilateral matching fails and tasks that do not pass the task simulation parasitism judgment, tasks to be hatched that cannot undergo the egg replacement mechanism, and matching tasks replaced by the egg replacement mechanism;

[0089] Use the virtual machine expansion decision mechanism, and determine the mean and standard deviation of the memory occupancy value, computing requirement value, and bandwidth requirement value of the computing task. Respectively, statistically calculate the comparison differences between the memory occupancy value, computing requirement value, and bandwidth requirement value of the computing task and the actual memory occupancy value, computing requirement value, and bandwidth requirement value of different tasks, and take the maximum value in the comparison difference and the standard deviation as the increased value for virtual machine expansion;

[0090] Based on the increased value of virtual machine expansion, recalculate the remaining idle computing power value, idle memory space, and idle network bandwidth of the virtual machine, and compare the memory occupancy value, computing requirement value, and bandwidth requirement value required by each task in the task set to be optimized. A virtual machine whose three idle parameters are all greater than the three parameters required by the task to be optimized is marked as a candidate virtual machine.

[0091] Through the statistical analysis of historical task resource requirements, the system can dynamically derive the buffer space required for virtual machine configuration in a data-driven manner, thereby realizing resource increment setting according to demand rather than a static ratio, making the expansion behavior more in line with actual needs, effectively avoiding the two extreme situations of resource redundancy and resource shortage in traditional expansion strategies. After expanding the virtual machine and then calculating the remaining available resources, the processing potential of the virtual machine can be released to the greatest extent, providing a scheduling space for tasks near the resource boundary, thereby improving the resource usage compactness.

[0092] Preferably, according to the scheduling order of the task set to be optimized and the candidate virtual machines, and the virtual machine numbers assigned to the tasks to be optimized in their candidate virtual machine sets, construct a scheduling candidate model, expressed as:

[0093] Sk = (π k , φ k );

[0094] Where S k represents the k-th scheduling candidate solution, π k represents the k-th scheduling order of the task to be optimized, and φ k represents the virtual machine number assigned to the task to be optimized in the k-th scheduling order;

[0095] In the construction of the scheduling candidate model, the matching of each task is updated in real time according to the remaining resources of the current virtual machine;

[0096] Define the total objective cost function of this scheduling solution as the sum of the maximum completion time, the imbalance degree of the virtual machine load, and the scheduling failure rate;

[0097] Calculate the total running time required for allocating tasks to each virtual machine according to different task matching orders based on the scheduling candidate solution, which is expressed as:

[0098]

[0099] Where Mk(S k ) represents the maximum completion time of the k-th scheduling candidate solution, m i represents the i-th task to be optimized, τ ok represents the set of tasks that have been successfully assigned virtual machines in this solution, com i represents the computing requirement value of the i-th task to be optimized, CPU j represents the number of cpu cores of the j-th virtual machine, and M j represents the computing capacity value of the j-th virtual machine;

[0100] Statistically calculate the total computing volume of tasks on each virtual machine, calculate the average task load value of all virtual machines, and analyze the imbalance degree of the task load of each virtual machine, which is expressed as:

[0101]

[0102] Where Im(S k ) represents the imbalance degree of the k-th scheduling candidate solution, b represents the total number of virtual machines, represents the total task computing requirement value of the j-th virtual machine in the k-th scheduling candidate solution, represents the average computing requirement of all virtual machines in the current scheduling solution;

[0103] Calculate how many tasks in the current scheduling plan fail to be successfully assigned and calculate the scheduling failure rate, which is expressed as:

[0104]

[0105] Among them, Fr(S k ) means that v represents the total number of tasks to be optimized;

[0106] The Crow Search Algorithm (CSA) is used to perform iterative calculations based on the total objective cost function value of each scheduling candidate solution, including randomly swapping the scheduling order of the scheduling candidate solutions, randomly selecting virtual machines for task allocation from the candidate virtual machine set during iteration, recalculating the total objective cost function value. If the total objective cost function value is less than or equal to the total objective cost function value of the previous iteration, it is memorized and repeated iteration is performed;

[0107] Iteration is stopped until the calculation loss of the total objective cost function value no longer changes significantly during consecutive iterations, and the optimal scheduling candidate solution is output.

[0108] The scheduling candidate model and the crow search optimization process based on the objective cost function constitute a refined and closed-loop intelligent scheduling iteration mechanism, enabling the system to no longer rely on a one-time static matching result, but to continuously search for a globally better solution in two dimensions of task order and virtual machine allocation, thereby enhancing the response ability of the scheduling system to complex resource states. By explicitly modeling the task scheduling order as an optimization variable, the scheduling system can make a structural exploration of the impact of the task entry order on resource consumption, solve the problem of resource imbalance caused by scheduling high-consumption tasks first, and fundamentally reduce the structural obstacle of early matching behavior to the final scheduling success rate;

[0109] In each round of iteration, a random perturbation of the task and candidate virtual machine matching is introduced, and the objective function is synchronously calculated and evaluated in real time, enabling the scheduling behavior to have the self-feedback ability to adapt to the current resource state, being able to achieve fine-tuning distribution of tasks within the critical interval of resource usage, thereby reducing resource concentration, alleviating resource hotspots, and enhancing the overall balance of the system. By simultaneously introducing three indicators of the maximum completion time, load imbalance degree, and failure rate into the objective function, the optimization behavior does not deviate from the overall objective of the system, avoiding non-business-oriented scheduling deviation caused by only maximizing the success rate or load balance as the sole objective;

[0110] Responsible for constructing task pre-screening, resource micro-utilization, and structural repair in the main scheduling process through the "matching-parasitic-replacement" mechanism. By preferentially matching high-priority tasks, using fragments to carry lightweight tasks, and performing local replacement based on task priority and load ratio, hierarchical injection of tasks into the resource space is achieved. This mechanism preferentially guides a large number of tasks with suitable structures and low resource costs into the system, and maximally releases the local resource utilization rate of the system, reserving a controllable, high-quality, and optimizable set of tasks to be processed for the CSA stage. The CSA scheduling candidate model clearly takes "remaining unmatched tasks + replaced tasks + parasitic failure tasks" as input. This structural connection enables CSA to no longer handle global task allocation problems, but focus on task corner scheduling under complex states, with concentrated goals and clear resource characteristics, significantly improving the convergence efficiency and optimizable boundary of the CSA search space. The chain control mechanism performs fine-grained splitting and release in the resource space dimension, enabling CSA to construct scheduling feasibility constraints based on the actual resource fragment structure in scheduling candidate modeling, so as to ensure that each candidate scheduling solution is mapped based on "dynamically feasible resources" during the construction process, greatly enhancing the physical deployability of the solution, reducing the number of invalid search solutions, increasing the effective search density of the algorithm, and jointly improving the success rate, efficiency, and adaptive elasticity of the task scheduling system under complex resource states.

[0111] S3, update the mapping matrix, record the real-time parameters of the virtual machine, and output log data;

[0112] Preferably, based on the optimal scheduling candidate solution, record the order in which each task in the task set to be optimized is assigned to the virtual machine, and write the task-virtual machine allocation relationship into the mapping matrix for update.

[0113] Through the update of the mapping matrix, the synchronous binding of task allocation information and virtual machine resource status is achieved, improving the real-time accuracy of resource use, avoiding resource reuse conflicts, and providing a complete input basis for subsequent log auditing, migration decision-making, and task runtime monitoring.

[0114] Furthermore, based on the updated mapping matrix, update the calculation of virtual machine parameters including real-time load, memory usage value, and bandwidth status, and record and update the load value, free memory, and free bandwidth of the virtual machine in real time to generate virtual machine real-time log data.

[0115] By generating real-time log data, a high-credibility data basis is provided for task runtime monitoring, anomaly detection, resource prediction, and subsequent scheduling optimization, while enhancing the system's perception and response capabilities to sudden loads and resource bottlenecks.

[0116] S4, perform encrypted transmission according to the log data;

[0117] Preferably, the symmetric encryption AES-256 technology is used to encrypt and package the logs, the log files are transmitted to the cloud storage through the SFTP method, and consistency verification is performed at the receiving end to generate verification logs for storage.

[0118] Through consistency verification and verification log records, the auditability and traceability of the log chain are enhanced, providing credible vouchers for subsequent scheduling result traceability, system compliance inspection, and fault review.

[0119] This embodiment also provides a system for the comprehensive computing power scheduling method of cloud computing, including:

[0120] A priority evaluation module that analyzes cloud computing task parameters to determine task priorities;

[0121] A bilateral preference module that constructs preference sequences for tasks and virtual machines for preliminary task-virtual machine matching;

[0122] A fragment identification module that performs scheduling probability evaluation on unmatched tasks and updates task matching results;

[0123] A virtual machine expansion module that collects all unfinished scheduling tasks and calculates the virtual machine expansion increment;

[0124] Scheduling candidate model construction, generating a candidate solution structure based on the task scheduling order and virtual machine allocation;

[0125] A scheduling optimization module that uses the crow search algorithm to iteratively search for the optimal scheduling order and task allocation;

[0126] A resource status synchronization module that updates the optimal scheduling result to the mapping matrix and records data in real time;

[0127] A log encryption transmission module that generates encrypted log files for secure transmission and performs consistency verification.

[0128] This embodiment also provides a computer device applicable to the comprehensive computing power scheduling method of cloud computing, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the comprehensive computing power scheduling method of cloud computing as proposed in the above embodiment.

[0129] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen, and an input device connected via a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a carrier network, NFC (Near Field Communication), or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a button, a trackball, or a touchpad provided on the housing of the computer device, or an external keyboard, a touchpad, or a mouse, etc.

[0130] This embodiment also provides a storage medium, on which a computer program is stored, and when the program is executed by a processor, it implements the method for comprehensive scheduling of computing power for cloud computing as proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, abbreviated as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, abbreviated as EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, abbreviated as EPROM), programmable read-only memory (Programmable Red-Only Memory, abbreviated as PROM), read-only memory (Read-Only Memory, abbreviated as ROM), magnetic memory, flash memory, a magnetic disk, or an optical disc.

[0131] In summary, through the chain control of matching, parasitism, and replacement, the present invention completes the full-process control from static game to dynamic fault tolerance, improves the success rate of task scheduling, the density of task occupancy, and the unit utilization efficiency of resources, reduces the overall system scheduling failure rate and resource idle rate, realizes the hierarchical injection of tasks in the resource space by preferentially matching high-priority tasks, using fragments to carry lightweight tasks, and performing local replacement based on the task priority and load ratio, so that CSA no longer deals with the global task allocation problem, but focuses on the task corner scheduling in complex states, enables CSA to construct scheduling feasibility constraints based on the real resource fragment structure in the scheduling candidate modeling, greatly enhances the physical deployability of the solution, reduces the number of invalid search solutions, and improves the effective search density of the algorithm.

[0132] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention.

Claims

1. A comprehensive computing power scheduling method for cloud computing, characterized in that, Including: Determine the task priority of cloud computing tasks, initially generate a mapping matrix, use the bilateral preference game mechanism to define task preferences and virtual machine preferences respectively, and initially match tasks with virtual machines using the stable matching algorithm. Determine the idle fragment resources of virtual machines, use the cuckoo behavior scheduling algorithm, calculate the scheduling probability of unmatched tasks and update task matching, and use the egg replacement mechanism to re-replace to match tasks for local optimization; Construct a task set to be optimized, use the virtual machine expansion decision mechanism to expand virtual machines, construct a scheduling candidate model, calculate the total objective cost function of different scheduling candidate solutions, and use the crow search algorithm for iterative calculation to output optimized scheduling candidate solutions; Update the mapping matrix, record the real-time parameters of virtual machines, and output log data; Perform encrypted transmission according to the log data.

2. The comprehensive computing power scheduling method for cloud computing according to claim 1, wherein: The matching of tasks and virtual machines, and using the egg replacement mechanism to re-replace to match tasks for local optimization includes: Divide the computing tasks of cloud computing, and determine the resource consumption of each computing task, including the task deadline, computing demand value, and resource consumption value, including memory occupancy value and bandwidth usage value; Set a set of virtual machines, clarify the memory capacity and network bandwidth for each virtual machine, and determine the computing power value. Based on the task deadline and computing demand value of the task, determine the task priority, and randomly assign an initial virtual machine to each task in the resource pool to obtain the initial mapping matrix A, and use the bilateral preference game mechanism to define task preferences and virtual machine preferences respectively; Among them, the task preference defines the ratio of the descending order of the virtual machine computing power value to the task computing priority number as the task preference list, and the virtual machine preference defines the ratio of the descending order of the task computing priority number to the task computing demand value as the virtual machine preference list; Use the stable matching algorithm Gale-Shapley, let the virtual machine accept in turn according to the proposed order, and cumulatively calculate the computing demand value of the accepted proposed tasks. When the current cumulative computing demand value is greater than or equal to the computing power value of the virtual machine, reject the proposed tasks; Cumulatively calculate the resource consumption values of the accepted proposed tasks, including the memory occupancy value and bandwidth usage value. When the current cumulative memory occupancy value and bandwidth usage value are respectively greater than or equal to the memory capacity and network bandwidth of the virtual machine, reject the proposed tasks; The rejected proposed tasks propose to the next virtual machine according to the task preference list, and loop to complete the initial task matching until the remaining unmatched tasks have clearly exhausted the proposals of all virtual machines in the preference list; Define the idle resource fragment set of the virtual machine according to the usage fragment time period of the virtual machine. Perform task simulation parasitism judgment on the tasks that have not been successfully matched. If the computing demand value, memory occupancy value, and bandwidth usage value of the task to be hatched are respectively less than the total computing power value, memory capacity, and network bandwidth that the fragment segment can provide, it can be used as the simulation object of the corresponding virtual machine fragment segment and combined into the task set to be hatched; The scheduling algorithm using the cuckoo behavior calculates the probability of each task to be hatched being successfully scheduled to an idle resource fragment through a probability control mechanism, conducts a Bernoulli experiment based on the probability value of being scheduled to an idle resource fragment, updates the idle resource fragment and removes the task from the set of tasks to be hatched; If the idle resource fragment cannot carry the task to be hatched, the task replacement is implemented using the egg replacement mechanism, and the task matching status and the set of idle resource fragments are updated, and the task replacement is repeatedly executed in a loop.

3. The comprehensive computing power scheduling method for cloud computing according to claim 2, wherein: The virtual machine expansion mentioned above includes, Constructing a set of tasks to be optimized, including tasks with unsuccessful bilateral matching and tasks that fail to pass the task simulation parasitism judgment, tasks to be hatched that cannot implement the egg replacement mechanism, and matching tasks replaced by the egg replacement mechanism; Using the virtual machine expansion decision mechanism to determine the increased value of virtual machine expansion based on historical data; Marking the virtual machines whose three idle parameters are all greater than the three parameters required by the tasks to be optimized as candidate virtual machines.

4. The comprehensive computing power scheduling method for cloud computing according to claim 3, wherein: The construction of the scheduling candidate model and the output of the optimized scheduling candidate solution include, Constructing a scheduling candidate model according to the scheduling order of the set of tasks to be optimized and the candidate virtual machines and the virtual machine numbers assigned to the tasks to be optimized in their candidate virtual machine sets; Defining the total objective cost function of this scheduling solution as the sum of the maximum completion time, the imbalance degree of virtual machine load, and the scheduling failure rate; Using the Cuckoo Search Algorithm (CSA) to perform iterative calculations based on the total objective cost function value of each scheduling candidate solution. If the total objective cost function value is less than or equal to the total objective cost function value of the previous iteration, then memorize and perform repeated iterations; Stop the iteration until the calculation loss of the total objective cost function value no longer changes significantly during consecutive iterations, and output the optimal scheduling candidate solution.

5. The comprehensive computing power scheduling method for cloud computing according to claim 4, characterized in that: The mapping matrix update mentioned above includes, Based on the optimal scheduling candidate solution, recording the order in which each task in the set of tasks to be optimized is assigned to a virtual machine, and writing the task-virtual machine assignment relationship into the mapping matrix for update.

6. The comprehensive computing power scheduling method for cloud computing according to claim 5, wherein: The recording of virtual machine real-time parameters and the output of log data include, Based on the updated mapping matrix, updating and calculating virtual machine parameters including real-time load, memory usage value, and bandwidth status, and real-time recording and updating the load value, free memory, and free bandwidth of the virtual machine to generate virtual machine real-time log data.

7. The comprehensive computing power scheduling method for cloud computing according to claim 6, characterized in that: The encrypted transmission according to the log data includes, Using the symmetric encryption AES-256 technology to encrypt and package the log, transmitting the log file to the cloud storage through the SFTP method, and performing consistency verification at the receiving end to generate a verification log for storage.

8. A system for the comprehensive scheduling method of computing power in cloud computing, based on the comprehensive scheduling method of computing power in cloud computing according to any one of claims 1 to 7, characterized in that: Including, A priority evaluation module that analyzes cloud computing task parameters to determine task priorities; A bilateral preference module that constructs the preference sequences of tasks and virtual machines for preliminary task-virtual machine matching; A fragment recognition module that performs scheduling probability evaluation on unmatched tasks and updates task matching results; A virtual machine expansion module that collects all unscheduled tasks and calculates the virtual machine expansion increment; Construction of a scheduling candidate model, generating a candidate solution structure based on task scheduling order and virtual machine assignment; A scheduling optimization module that uses the Cuckoo Search Algorithm to iteratively search for the optimal scheduling order and task assignment; The resource status synchronization module updates the optimal scheduling result to the mapping matrix and records data in real time. The log encryption and transmission module generates an encrypted log file, performs secure transmission and conducts consistency verification.

9. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the computing power comprehensive scheduling method for cloud computing according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the computing power comprehensive scheduling method for cloud computing according to any one of claims 1 to 7.

Images