Multi-cloud computing resource management method and system

Abstract

The invention relates to the technical field of multi-cloud computing resource management, in particular to a multi-cloud computing resource management method and system. Comprising a resource monitoring module, a strategy making module and a dynamic adjustment module. According to the invention, the resource monitoring module monitors the use state of the computing power resource of each cloud platform through the monitoring service of the cloud platform, and the strategy making module carries out the priority division of tasks through an analytic hierarchy process according to the importance, timeliness and cost effectiveness of the tasks. The dynamic adjustment module takes the computing power state index of the cloud computing resource as an input variable, takes the task priority value as an output variable, dynamically adjusts the priority according to the state of the computing power resource by utilizing a fuzzy algorithm, and dynamically adjusts the priority determined by the strategy making module according to the state of the computing power resource; it is ensured that the priority strategy can adapt to the state of the computing power resources, and the reasonability of computing power resource allocation is ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of multi-cloud computing resource management, and specifically, to a multi-cloud computing resource management method and system. Background Art

[0002] A multi-cloud management platform is an innovative cloud computing management solution designed to help enterprises and organizations efficiently integrate, manage, and optimize cloud computing resources across multiple cloud service providers (such as public clouds, private clouds, etc.). Through a unified interface and intelligent management mechanisms, it breaks down the barriers between cloud services, enabling collaborative allocation of resources, cost control, security assurance, and convenient operation and maintenance in a multi-cloud environment, providing users with a one-stop multi-cloud management experience and greatly enhancing the flexibility and efficiency of cloud computing resource utilization.

[0003] Currently, when allocating computing power resources in a multi-cloud management platform, different tasks have different requirements for computing power resources. If computing power resources are simply evenly distributed, some unimportant or non-urgent tasks will occupy resources, causing important tasks to be delayed and reducing the utilization rate of computing power resources. At the same time, when multiple tasks are running simultaneously, the state of computing power resources will also change. In order to prioritize tasks according to the importance, timeliness, and cost-effectiveness of tasks, improve resource utilization, and dynamically adjust the priority of tasks according to the state of computing power resources to ensure the rationality of computing power resource allocation, therefore, we propose a multi-cloud computing resource management method and system. Summary of the Invention

[0004] The purpose of the present invention is to solve the problem that the multi-cloud management platform cannot effectively allocate computing power resources according to the importance of tasks, reducing the utilization rate of computing power resources. In order to prioritize tasks according to the importance, timeliness, and cost-effectiveness of tasks, improve resource utilization, and dynamically adjust the priority of tasks according to the state of computing power resources to ensure the rationality of resource allocation.

[0005] To achieve the above object, the present invention provides a multi-cloud computing resource management method, including the following steps:

[0006] S1. Use the monitoring service of the cloud platform to monitor the usage status of computing power resources of each cloud platform;

[0007] S2. Use the analytic hierarchy process to prioritize tasks according to the importance, timeliness, and cost-effectiveness of tasks;

[0008] S3. Use the fuzzy algorithm to dynamically adjust the priority according to the state of computing power resources, with the computing power status index of cloud computing resources as the input variable and the task priority value as the output variable.

[0009] Preferably, in S2, the analytic hierarchy process is used to divide the priorities of tasks, and the method steps are as follows:

[0010] S2.1.1. Construct a hierarchical structure model: Determine that the priority division is the target layer, the importance, timeliness, and cost - effectiveness of tasks are the criterion layers, and the tasks that need to be divided into priorities are the scheme layers;

[0011] S2.1.2. Construct a judgment matrix: Use the scaling method to construct the judgment matrix of the criterion layer relative to the target layer and the judgment matrix of the scheme layer relative to the criterion layer;

[0012] S2.1.3. Calculate the weight vector: Calculate the maximum eigenvalue of the judgment matrix and its corresponding eigenvector, and normalize the eigenvector to obtain the weight vector of each criterion in the criterion layer relative to the target layer;

[0013] S2.1.4. Consistency test: For each judgment matrix of the criterion layer and the scheme layer, calculate the consistency index, according to the order of the judgment matrix, look up the average random consistency index, calculate the consistency ratio, and judge the judgment matrix.

[0014] Preferably, in S2.1.3 for calculating the weight vector and normalizing the eigenvector, the normalization formula is:

[0015]

[0016] where w i is the i - th component of the weight vector, i is the serial number of the eigenvector, v i is the i - th vector in the eigenvector, n is the order of the eigenvector, j is the initial serial number of the eigenvector, and v j is the j - th vector in the eigenvector.

[0017] Preferably, in S2.1.4 for the consistency test, the formula for calculating the consistency index is:

[0018]

[0019] where CI is the value of the consistency index, λ max is the maximum eigenvalue of the eigenvector, and n is the order of the eigenvector.

[0020] Preferably, when S2 divides the priorities, a computing power resource allocation strategy that gives priority to priorities and is supplemented by a queue - based algorithm scheduling is adopted. Inside each priority queue, the first - come - first - served algorithm is used, and tasks obtain computing power resources in the order of entering the queue.

[0021] Preferably, S3 dynamically adjusts the priorities according to the status of computing power resources, and the method steps are as follows:

[0022] S3.1.1. Define fuzzy sets and membership functions: Define fuzzy sets for the input variables, and use trapezoidal membership functions to calculate the membership values of each fuzzy set;

[0023] S3.1.2. Build a fuzzy rule base: According to business requirements and resource management experience, formulate fuzzy rules and refine and quantify them;

[0024] S3.1.3. Fuzzy inference: Use the Mamdani inference algorithm to calculate the activation strength of the business priority fuzzy set based on the membership degree of the input fuzzy variable of the resource status and combine with the fuzzy rules to obtain the final business priority fuzzy result;

[0025] S3.1.4. Defuzzification: Use the centroid method to defuzzify the fuzzy set and obtain a definite priority value by calculating the centroid of the fuzzy output.

[0026] Preferably, in the above S3.1.1 for defining fuzzy sets and membership functions, the formula for the trapezoidal membership function is:

[0027]

[0028] where μ(x) is the membership function value, x is the CPU usage rate, a is the lower limit, b is the end point of the rising interval, c is the starting point of the falling interval, and d is the upper limit.

[0029] Preferably, in the above S3.1.3 for fuzzy inference, use the Mamdani inference algorithm and adopt the method of taking the minimum value to determine the activation strength according to the logical connectives in the rules.

[0030] Preferably, after dynamically adjusting the priority in S3 above, establish a conflict resolution mechanism according to resource conflicts and waiting time conflicts to perform resource regulation on the priority policy and queue scheduling policy.

[0031] The second object of the present invention is to provide a multi-cloud computing resource management system, including the multi-cloud computing resource management method described in any one of the above, including a resource monitoring module, a policy formulation module, and a dynamic adjustment module;

[0032] The resource monitoring module monitors the usage status of the computing power resources of each cloud platform through the monitoring service of the cloud platform;

[0033] The policy formulation module divides the priorities of tasks using the analytic hierarchy process according to the importance, timeliness, and cost-effectiveness of the tasks;

[0034] The dynamic adjustment module takes the computing power status index of cloud computing resources as the input variable and the task priority value as the output variable, and uses the fuzzy algorithm to dynamically adjust the priority according to the status of computing power resources.

[0035] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0036] 1. For the multi-cloud computing resource management method and system, the policy formulation module uses the analytic hierarchy process to divide the priorities of tasks according to the importance, timeliness, and cost-effectiveness of the tasks, and adopts a computing power resource allocation strategy with priority as the main and queue algorithm scheduling as the auxiliary. It preferentially allocates computing power resources to tasks with high importance, improves the utilization rate of computing power resources. At the same time, within the same priority, with the queue algorithm as the auxiliary, tasks obtain computing power resources in the order of entry into the queue, accelerating the use process of computing power resources and ensuring the high efficiency of computing power resource utilization;

[0037] 2. The resource monitoring module uses the monitoring service of the cloud platform to monitor the usage status of the computing power resources of each cloud platform. The dynamic adjustment module, based on the usage status of the computing power resources of the cloud platform monitored by the resource monitoring module, takes the computing power status index of cloud computing resources as the input variable and the task priority value as the output variable, and uses the fuzzy algorithm to dynamically adjust the priority determined by the policy formulation module according to the status of computing power resources, ensuring that the priority policy can adapt to the status of computing power resources and guaranteeing the rationality of computing power resource allocation. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 is the method flow block diagram of the present invention;

[0039] Figure 2 is the specific flowchart of S2 of the present invention;

[0040] Figure 3 is the specific flowchart of S3 of the present invention;

[0041] Figure 4 is the system flowchart of the present invention.

[0042] The meanings of the various labels in the figure are as follows:

[0043] 100, resource monitoring module; 200, policy formulation module; 300, dynamic adjustment module. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0044] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0045] Currently, the multi-cloud management platform cannot effectively allocate computing power resources according to the importance of tasks, reducing the utilization rate of computing power resources. In order to be able to prioritize tasks according to the importance, timeliness, and cost-effectiveness of tasks, improve resource utilization, and dynamically adjust the priority of tasks according to the status of computing power resources to ensure the rationality of resource allocation.

[0046] Therefore, the present invention proposes that the resource monitoring module monitors the usage status of the computing power resources of each cloud platform through the monitoring service of the cloud platform, the policy formulation module uses the analytic hierarchy process to prioritize tasks according to the importance, timeliness, and cost-effectiveness of tasks, and the dynamic adjustment module uses the computing power status index of cloud computing resources as the input variable and the task priority value as the output variable, and uses the fuzzy algorithm to dynamically adjust the priority according to the status of computing power resources.

[0047] Embodiment 1 is as follows:

[0048] As Figure 1 shown, one of the purposes of the present invention is to provide a multi-cloud computing resource management method, including the following steps:

[0049] S1. Monitor the usage status of the computing power resources of each cloud platform through the monitoring service of the cloud platform;

[0050] S2. Use the analytic hierarchy process to prioritize tasks according to the importance, timeliness, and cost-effectiveness of tasks;

[0051] S3. Use the fuzzy algorithm to dynamically adjust the priority according to the status of computing power resources, with the computing power status index of cloud computing resources as the input variable and the task priority value as the output variable.

[0052] Each major cloud computing platform provides its own monitoring service. For example, CloudWatch of Amazon AWS is a powerful monitoring tool that can monitor the usage of various AWS resources, and it can provide detailed metrics such as the CPU usage rate, memory usage rate, network I / O, and disk I / O of EC2 instances (cloud servers);

[0053] Microsoft Azure's Monitor service is similar, covering the monitoring of resources such as Azure virtual machines and storage accounts. You can view the performance data of resources in detail, including but not limited to the percentage of processor time, the number of bytes read / written from / to the disk per second, etc.

[0054] As Figure 2 shown, where S2 uses the analytic hierarchy process to prioritize tasks, and its method steps are as follows:

[0055] S2.1.1. Build a hierarchical structure model: Determine that the priority division is the goal layer, the importance, timeliness, and cost - effectiveness of the task are the criterion layers, and the tasks that need to be prioritized are the scheme layers;

[0056] S2.1.2. Construct a judgment matrix: Use the scale method to construct the judgment matrix of the criterion layer relative to the goal layer and the judgment matrix of the scheme layer relative to the criterion layer;

[0057] S2.1.3. Calculate the weight vector: Calculate the maximum eigenvalue of the judgment matrix and its corresponding eigenvector, and normalize the eigenvector to obtain the weight vector of each criterion in the criterion layer relative to the goal layer;

[0058] S2.1.4. Consistency test: For each judgment matrix in the criterion layer and the scheme layer, calculate the consistency index, according to the order of the judgment matrix, find the average random consistency index, calculate the consistency ratio, and judge the judgment matrix;

[0059] Determine the overall goal of priority division, including three main criteria: business importance, timeliness, and cost - effectiveness. Business importance can be further broken down into sub - criteria such as the impact on core business and the impact on customer satisfaction. Timeliness can include sub - criteria such as the urgency of the task deadline and the length of the business cycle. Cost - effectiveness can cover sub - criteria such as the resource input - output ratio and potential economic benefits, that is, each business that needs to be prioritized;

[0060] In the analytic hierarchy process (AHP), the 1 - 9 scale method is used to quantify and compare the relative importance of two elements. The specific scale is as follows:

[0061] 1 means that when two elements are compared, they have the same importance;

[0062] 3 means that when two elements are compared, one element is slightly more important than the other;

[0063] 5 means that when two elements are compared, one element is significantly more important than the other;

[0064] 7 means that when two elements are compared, one element is strongly more important than the other;

[0065] 9 means that when comparing two elements, one element is extremely more important than the other;

[0066] 2, 4, 6, 8 represent the intermediate values of the above adjacent judgments. For example, 2 means that when comparing two elements, one element is between slightly more important and equally important than the other;

[0067] If it is considered that the task importance is slightly more important than the timeliness, then the matrix element corresponding to the task importance and timeliness is assigned a value of 3. If it is considered that the timeliness is significantly more important than the cost - effectiveness, it is assigned a value of 7, etc. For each criterion, compare the relative importance between each task. For example, under the task importance criterion, compare the influence degrees of task 1 and task 2 on the core task. If the influence degree of task 1 is much larger than that of task 2, assign a value of 7. If the influence degrees of both are the same, assign a value of 1.

[0068] Calculate the maximum eigenvalue of the judgment matrix and its corresponding eigenvector. For the above judgment matrix of the criterion layer, through calculation, obtain the maximum eigenvalue and the corresponding eigenvector, and perform normalization processing on the eigenvector to obtain the weight vector of each criterion in the criterion layer relative to the target layer.

[0069] In order to better perform normalization processing on the eigenvector, where, S2.1.3 Calculate the weight vector and perform normalization processing on the eigenvector. The normalization formula is:

[0070]

[0071] where, w i is the i - th component of the weight vector, i is the serial number of the eigenvector, v i is the i - th vector in the eigenvector, n is the order of the eigenvector, j is the initial serial number of the eigenvector, v j is the j - th vector in the eigenvector.

[0072] For example, the judgment matrix of the criterion layer relative to the target layer is as follows:

[0073] Task importance (B1) Timeliness (B2) Cost-effectiveness (B3) Task importance (B1) 1 3 4 Timeliness (B2) 1 / 3 1 2 Cost-effectiveness (B3) 1 / 4 1 / 2 1

[0074] In this matrix, for example, when comparing the task importance (B1) and the timeliness (B2), it is considered that the task importance is "slightly more important" than the timeliness, so it is assigned a value of 3. When comparing the timeliness (B2) and the cost - effectiveness (B3), it is considered that the timeliness is "slightly more important" than the cost - effectiveness, and it is assigned a value of 2;

[0075] The judgment matrix of the scheme layer relative to each criterion in the criterion layer is as follows:

[0076] Under the task importance B1:

[0077]

[0078]

[0079] Here, it means that in terms of business importance, Project A is "slightly more important" than Project B, with an assigned value of 2, and Project A is "significantly more important" than Project C, with an assigned value of 3;

[0080] Under timeliness B2:

[0081] A B C A 1 1 / 2 1 / 3 B 2 1 1 / 2 C 3 2 1

[0082] In terms of timeliness, Project C is "significantly more important" than Project A, with an assigned value of 3; Project B is "slightly more important" than Project A, with an assigned value of 2;

[0083] Under cost - benefit B3:

[0084] A B C A 1 3 1 / 2 B 1 / 3 1 1 / 3 C 2 3 1

[0085] In terms of cost - benefit, Project A is "significantly more important" than Project B, with an assigned value of 3, and Project C is "slightly more important" than Project A, with an assigned value of 2.

[0086] In order to better calculate the consistency index value of the eigenvector, where, for S2.1.4 consistency test, the formula for calculating the consistency index is:

[0087]

[0088] Among them, CI is the consistency index value, λ max is the maximum eigenvalue of the eigenvector, and n is the order of the eigenvector.

[0089] Calculate the maximum eigenvalue λ of the judgment matrix of the criterion layer max , substitute λ max ≈3.039, n = 3 into the formula for the consistency index, and we get:

[0090] CI=(3.039 - 3) / (3 - 1)≈0.0195;

[0091] Look up the average random consistency index RI and calculate the consistency ratio:

[0092] CR = CI / RI≈0.0336<0.1;

[0093] So the consistency of the judgment matrix is acceptable. After calculating the eigenvector and normalizing it, we obtain the weight vector of each criterion in the criterion layer relative to the target layer. Similarly, calculate the weight vector of the scheme layer under each criterion according to the same steps, and conduct a comprehensive priority calculation on it;

[0094] The weight vectors of the solution layer under each criterion are comprehensively calculated with the weight vectors of the criterion layer, and the services are sorted according to the size of the comprehensive priority, thus completing the priority division.

[0095] In order to better implement the computing power resource allocation strategy that gives priority to priorities and supplements with a queue algorithm-based scheduling, where in S2, when dividing priorities, a computing power resource allocation strategy that gives priority to priorities and supplements with a queue algorithm-based scheduling is adopted. Inside each priority queue, the first-come, first-served algorithm is adopted, and tasks obtain computing power resources in the order of their entry into the queue.

[0096] Different queues are divided according to the priority levels of tasks. For example, create a priority 1 queue, a priority 2 queue, and a priority 3 queue. Tasks are assigned to the corresponding queues according to their comprehensive priorities, and tasks in the high-priority queue obtain computing power resources first.

[0097] Inside each priority queue, the first-come, first-served algorithm is adopted. That is, tasks obtain computing power resources in the order of their entry into the queue. This algorithm is simple and easy to understand and implement, and can ensure the fairness of tasks when the task arrival times are relatively uniform. For example, in the priority 2 queue, task A enters the queue before task B, so when there is enough computing power, task A will obtain the computing power resources first for processing.

[0098] Tasks in the queue are scheduled according to the real-time availability of computing resources. For example, when new GPU resources are available, tasks are preferentially selected from the GPU demand queue for allocation, or when there are a large number of idle CPU resources in the server cluster of a certain data center, tasks with a large demand for CPU resources are selected from each priority queue and allocated to the cluster. The availability information of resources can be obtained in real time through a resource monitoring system and fed back to the queue scheduling system to make reasonable scheduling decisions.

[0099] The task enqueue rule should be closely combined with the priority strategy. When a new task arrives, it is first placed in the corresponding priority queue according to the priority calculation. In this process, the accuracy and consistency of the priority should be ensured. For example, if the priority of a task is adjusted due to business changes, it should be promptly moved to the correct priority queue.

[0100] In the process of queue scheduling, priority is always the main consideration factor. Even when the first-come, first-served algorithm is adopted inside the queue, when a high-priority task arrives at the head of the queue, computing power resources should be preferentially allocated. At the same time, when considering resource availability and other scheduling factors, the dominant position of priority cannot be ignored. For example, when scheduling tasks based on resource availability, if two tasks in different queues have a demand for the same resource and the resource can only satisfy one task, then the task with a higher priority is preferentially selected.

[0101] As shown in Figure 3 Figure [3], where S3 dynamically adjusts the priority according to the status of computing power resources, and the method steps are as follows:

[0102] S3.1.1. Define fuzzy sets and membership functions: Define fuzzy sets for input variables, and obtain the membership values of each fuzzy set using trapezoidal membership functions;

[0103] S3.1.2. Construct a fuzzy rule base: Develop fuzzy rules based on business requirements and resource management experience, and refine and quantify them;

[0104] S3.1.3. Fuzzy inference: Using the Mamdani inference algorithm, calculate the activation intensity of the business priority fuzzy set based on the membership degrees of the input resource status fuzzy variables and combine with the fuzzy rules to obtain the final business priority fuzzy result;

[0105] S3.1.4. Defuzzification: Defuzzify the fuzzy set using the centroid method, and obtain a definite priority value by calculating the centroid of the fuzzy output.

[0106] For the status of computing power resources, fuzzy sets such as "resource shortage", "resource medium", and "resource abundance" are defined. For example, for CPU usage, the usage rate between 0 - 30% can be defined as "resource abundance", 30% - 70% as "resource medium", and 70% - 100% as "resource shortage". For each fuzzy set, a membership function needs to be defined to describe the degree to which an element belongs to the fuzzy set. For example, for the "resource shortage" fuzzy set, a trapezoidal membership function can be used. When the CPU usage rate is greater than or equal to 70%, the membership degree starts to linearly increase from 0 to 1 (reaching 1 at a usage rate of 100%);

[0107] To better determine the trapezoidal membership function, where S3.1.1 defines fuzzy sets and membership functions, and the formula for the trapezoidal membership function is:

[0108]

[0109] where μ(x) is the membership function value, x is the CPU usage rate, a is the lower limit, b is the end point of the rising interval, c is the start point of the falling interval, and d is the upper limit.

[0110] It is used to describe the situation where the membership degree of a variable is 1 within a certain interval and gradually changes at both ends of the interval. For example, for the "high load" state of resources, when the CPU usage rate reaches a certain level (such as 70%), it is considered to have entered the high load state, and the membership degree starts to gradually increase from 0. After reaching a relatively high value (such as 90%), the membership degree remains 1, and the trapezoidal membership function is selected.

[0111] Define fuzzy sets for task priorities, such as "high priority", "medium priority", and "low priority". The definitions of these fuzzy sets can be determined based on factors such as the importance and timeliness of the business. Similarly, define membership functions for each task priority fuzzy set. For example, the membership function for a "high priority" task can be determined based on the criticality of the business and the time urgency. For instance, the membership degree of a business that is urgent and crucial to the core business of the enterprise is 1, and other businesses take values between 0 and 1 according to the situation;

[0112] The universe of discourse is the range of values that a fuzzy variable can take. For resource status variables, such as the universe of discourse of CPU usage rate is [0, 100%], and the universe of discourse of memory occupancy rate is [0, 100%]. For business priority variables, the universe of discourse can be the interval [0, 1], where 0 represents the lowest priority and 1 represents the highest priority. By determining the universe of discourse, the range of values of the fuzzy variable can be clarified, providing a basis for subsequent fuzzy reasoning.

[0113] According to business requirements and resource management experience, formulate fuzzy rules. For example, "If resources are scarce and the business has a high demand for resources, then the business priority is reduced"; "If resources are abundant and the business has a certain demand for resources, then the business priority can be appropriately increased". These rules reflect the intuitive relationship between resource status and business priority, and also consider the demand characteristics of the business for resources;

[0114] For each rule, further refine the descriptions of resource status and business priority. For example, for the description of "the business has a high demand for resources", it can be quantified by the occupancy rate of resources when the business is running normally. If a business often has a CPU usage rate exceeding 50% when running normally, it is considered to have a high demand for CPU resources. Through this quantification method, the fuzzy rules are made more operational and convenient for subsequent fuzzy reasoning.

[0115] Using the Mamdani inference algorithm, based on the membership degrees of the input resource status fuzzy variables and combined with the fuzzy rules, calculate the activation intensity of the business priority fuzzy set to obtain the final business priority fuzzy result;

[0116] When specific resource status data is input, for example, the CPU usage rate is 80%, according to the previously defined membership function, calculate its membership degree belonging to the "resource shortage" fuzzy set (assumed to be 0.6). At the same time, assume that the business has a "high demand" for the CPU. According to Rule 1, the activation intensity of the business priority "low priority" is 0.6. Perform such calculations for all eligible rules to obtain the activation intensities of each fuzzy set of business priorities.

[0117] In order to better determine the activation intensity, in which, S3.1.3 Fuzzy Inference, using the Mamdani inference algorithm, according to the logical connectives in the rules, the minimum value is used to determine the activation intensity;

[0118] Calculate the fuzzy output according to the activation intensity. For example, using the max-min composition method, for the fuzzy sets of "high priority", "medium priority", and "low priority" of service priority, the activation intensity of each fuzzy set is combined with the corresponding membership function to obtain a fuzzy output, indicating the distribution of service priority on each fuzzy set;

[0119] Convert the clear input values (such as specific resource status indicators like CPU usage rate, memory occupancy rate, etc.) into membership degrees of fuzzy sets. This is achieved through pre-defined membership functions. For example, for the input variable of CPU usage rate, we may define three fuzzy sets: "low load", "medium load", and "high load", each with a corresponding membership function. If the actual CPU usage rate is 60%, through the membership function calculation, its membership degree for the "medium load" fuzzy set may be 0.7, for the "high load" fuzzy set may be 0.3, and for the "low load" fuzzy set is 0;

[0120] Fuzzy rules are the key part of Mamdani inference. They are usually expressed in the form of "if... then...", describing the relationship between input fuzzy variables and output fuzzy variables (such as service priority). For each fuzzy rule, calculate its activation intensity according to the above method. Suppose there is another rule: "If the CPU usage rate is'medium load' and the memory occupancy rate is'medium occupancy', then the service priority is'medium priority'", and the calculated activation intensity is 0.5 (assumed value). These activation intensities indicate the degree to which the corresponding output fuzzy sets (such as task priority fuzzy sets like "low priority", "medium priority", etc.) are triggered.

[0121] Use the centroid method to defuzzify the fuzzy set, and obtain a definite priority value by calculating the centroid of the fuzzy output;

[0122] Suppose that after fuzzy reasoning, the membership degree of the service priority in the "high priority" fuzzy set is 0.3, in the "medium priority" fuzzy set is 0.5, and in the "low priority" fuzzy set is 0.2. Using the centroid method, according to the definition of the service priority fuzzy set (for example, "high priority" corresponds to the numerical interval [0.7, 1], "medium priority" corresponds to [0.4, 0.7], and "low priority" corresponds to [0, 0.4]), the determined priority value is calculated. For example, the calculation result may be a numerical value between 0 and 1, and this numerical value represents the service priority dynamically adjusted according to the current resource status and service resource requirements. Sort the services and allocate resources according to this priority value to achieve dynamic priority adjustment.

[0123] In order to be able to continue resource regulation for the priority policy and the queue scheduling policy, among them, after S3 dynamically adjusts the priority, according to resource conflicts and waiting time conflicts, a conflict resolution mechanism is established to perform resource regulation on the priority policy and the queue scheduling policy;

[0124] When conflicts occur in resource allocation between the priority policy and the queue scheduling policy, for example, priority preemptive scheduling may cause excessive preemption of resources of low-priority tasks, affecting their normal completion. In this case, a conflict resolution mechanism needs to be established. A resource preemption upper limit can be set, stipulating that high-priority tasks can at most preempt a certain proportion (such as 50%) of the resources of low-priority tasks to ensure that low-priority tasks can be completed within a certain time;

[0125] If a task waits in the queue for too long due to its low priority, it may affect the overall efficiency of the service. A maximum waiting time limit can be set. When the waiting time of a low-priority task exceeds this limit, appropriately increase its priority or adopt other strategies (such as allocating some idle resources to it) to ensure that it can be processed. This can ensure the priority processing of high-priority tasks while avoiding indefinite waiting of low-priority tasks.

[0126] The second object of the present invention is to provide a multi-cloud computing resource management system, including the multi-cloud computing resource management method of any one of the above, as Figure 4 shown, which includes a resource monitoring module 100, a policy formulation module 200, and a dynamic adjustment module 300;

[0127] The resource monitoring module 100 monitors the usage status of the computing power resources of each cloud platform through the monitoring service of the cloud platform;

[0128] The policy formulation module 200 divides the priorities of tasks using the analytic hierarchy process according to the importance, timeliness, and cost-effectiveness of the tasks;

[0129] The dynamic adjustment module 300 takes the computing power status index of cloud computing resources as the input variable and the task priority value as the output variable, and uses the fuzzy algorithm to dynamically adjust the priority according to the status of computing power resources.

[0130] In summary, the working principle of this solution is as follows:

[0131] For this multi-cloud computing resource management method and system, the policy formulation module 200 uses the analytic hierarchy process to divide the priorities of tasks according to the importance, timeliness, and cost-effectiveness of the tasks, and adopts a computing power resource allocation strategy with priority as the main and queue algorithm scheduling as the auxiliary. High-priority tasks are preferentially allocated computing power resources to improve the utilization rate of computing power resources. At the same time, within the same priority, with the queue algorithm as the auxiliary, tasks obtain computing power resources in the order of entry into the queue, accelerating the use process of computing power resources and ensuring the high efficiency of the use of computing power resources. The resource monitoring module 100 uses the monitoring service of the cloud platform to monitor the usage status of the computing power resources of each cloud platform. The dynamic adjustment module 300 dynamically adjusts the priorities determined by the policy formulation module 200 according to the usage status of the computing power resources of the cloud platform monitored by the resource monitoring module 100, taking the computing power status index of cloud computing resources as the input variable and the task priority value as the output variable, and using the fuzzy algorithm, so as to ensure that the priority policy can adapt to the status of computing power resources and ensure the rationality of the allocation of computing power resources.

[0132] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art of this industry should understand that the present invention is not limited by the above embodiments. The above embodiments and the descriptions in the specification are only preferred examples of the present invention and are not used to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. A multi-cloud computing resource management method, characterized in that: The following steps are involved: S1. Use the monitoring service of the cloud platform to monitor the usage status of the computing resources of each cloud platform; S2. Prioritize tasks using the analytic hierarchy process based on their importance, timeliness, and cost-effectiveness; S3. Taking the computing power status indicator of cloud computing resources as the input variable and the task priority value as the output variable, the fuzzy algorithm is used to dynamically adjust the priority according to the status of computing power resources.

2. The multi-cloud computing resource management method according to claim 1, characterized in that: S2 uses the hierarchical analysis method to prioritize tasks, and the method steps are as follows: S2.1.

1. Construct a hierarchical model: determine the priority level as the target level, the importance, timeliness and cost-effectiveness of the task as the criterion level, and the tasks that need to be prioritized as the solution level; S2.1.2, construct judgment matrix: Use the scaling method to construct the judgment matrix of the criterion layer relative to the target layer and the judgment matrix of the solution layer relative to the criterion layer; S2.1.

3. Calculate the weight vector: Calculate the maximum eigenvalue of the judgment matrix and its corresponding eigenvector, normalize the eigenvector, and obtain the weight vector of each criterion in the criterion layer relative to the target layer; S2.1.4, consistency test: For each judgment matrix at the criterion layer and the scheme layer, calculate the consistency index, find the average random consistency index according to the order of the judgment matrix, calculate the consistency ratio, and judge the judgment matrix.

3. The multi-cloud computing resource management method according to claim 2, characterized in that: S2.1.3 calculates the weight vector and normalizes the feature vector, wherein the normalization formula is: Among them, w i is the i-th component of the weight vector, i is the sequence number of the eigenvector, v i is the i-th vector in the eigenvector, n is the order of the eigenvector, j is the initial sequence number of the eigenvector, v j is the jth vector in the eigenvector.

4. The multi-cloud computing resource management method according to claim 2, characterized in that: The formula for calculating the consistency index in S2.1.4 consistency test is: Among them, CI is the consistency index value, λ max is the maximum eigenvalue of the eigenvector, and n is the order of the eigenvector.

5. The multi-cloud computing resource management method according to claim 1, characterized in that: When dividing priorities, S2 adopts a computing power resource allocation strategy that is based on priority and supplemented by queue algorithm scheduling. Within each priority queue, a first-come, first-served algorithm is adopted, and tasks obtain computing power resources in the order they enter the queue.

6. The multi-cloud computing resource management method according to claim 1, characterized in that: S3 dynamically adjusts the priority according to the status of computing resources, and the method steps are as follows: S3.1.

1. Define fuzzy sets and membership functions: Define fuzzy sets for input variables and use trapezoidal membership functions to find the membership values ​​of each fuzzy set; S3.1.

2. Build a fuzzy rule base: formulate fuzzy rules based on business needs and resource management experience, and refine and quantify them; S3.1.3, fuzzy reasoning: using the Mamdani reasoning algorithm, according to the membership of the input resource status fuzzy variables, combined with fuzzy rules, the activation intensity of the business priority fuzzy set is calculated to obtain the final business priority fuzzy result; S3.1.4, Defuzzification: The fuzzy set is defuzzified using the centroid method, and a definite priority value is obtained by calculating the centroid of the fuzzy output.

7. The multi-cloud computing resource management method according to claim 6, characterized in that: S3.1.1 defines fuzzy sets and membership functions, where the formula for the trapezoidal membership function is: Among them, μ(x) is the membership function value, x is the CPU usage, a is the lower limit, b is the end point of the rising interval, c is the starting point of the falling interval, and d is the upper limit.

8. The multi-cloud computing resource management method according to claim 6, characterized in that: The fuzzy reasoning in S3.1.3 uses the Mamdani reasoning algorithm to determine the activation strength by taking the minimum value according to the logical connectives in the rules.

9. The multi-cloud computing resource management method according to claim 1, characterized in that: After dynamically adjusting the priority, S3 establishes a conflict resolution mechanism according to resource conflicts and waiting time conflicts, and performs resource regulation on the priority strategy and queue scheduling strategy.

10. A method for implementing a multi-cloud computing resource management system, comprising the multi-cloud computing resource management method according to any one of claims 1 to 9, characterized in that: It includes a resource monitoring module (100), a policy formulation module (200) and a dynamic adjustment module (300); The resource monitoring module (100) monitors the usage status of computing resources of each cloud platform through the monitoring service of the cloud platform; The strategy formulation module (200) prioritizes tasks using a hierarchical analysis method according to the importance, timeliness and cost-effectiveness of the tasks; The dynamic adjustment module (300) uses the computing power status index of the cloud computing resources as an input variable and the task priority value as an output variable, and uses a fuzzy algorithm to dynamically adjust the priority according to the status of the computing power resources.

Images