Protection component virtual resource scheduling method and system based on strategy selection
By triggering the scheduling strategy based on application layer requirements and resource pool performance state in virtual resource scheduling, combined with the improvement of Harris Eagle algorithm and random integer planning model, the problems of single metrics and unbalanced load in the existing scheduling model are solved, and the resources are improved Utilization and reduce costs.
Patent Information
- Application Number
- CN202410905694.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-08
- Publication Date
- 2024-11-19
AI Technical Summary
In the existing scheduling model, the scheduling scheme considers single indicators, unbalanced load, and inadequate scheduling strategies, resulting in low resource utilization and high cost.
Based on the requirements changes of the upper application layer and the resource pool performance status of the virtualization layer, determine whether the resource scheduling strategy is triggered. If triggered, virtual resources will be conducted from the overall survival of the resource pool performance, the rationality of the underlying device settings, or the resource load balancing. The scheduling of the scheduling system uses improved Harris Hawk algorithm and random integer programming model for resource allocation and migration.
It improves resource utilization, reduces resource costs, realizes efficient allocation and transfer of virtual resources, and solves the problems of load imbalance and policy mismatch.
Smart Images

Figure CN118981371A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of virtual resource scheduling, and in particular to a protection component virtual resource scheduling method and system based on policy selection. Background Art
[0002] In recent years, the continuous penetration of new generation communication technologies represented by the Internet has pointed out the direction for the traditional power industry to form a new business model of interconnection, digital twins, and virtual and real integration. As a key area, resource management improves the integrated dispatching of data, improves the operation and management level of power companies, and plays an increasingly important role in the construction of flexible and efficient new power business system forms. In order to further solve the resource dispatching control and deployment generated by different power protection components, it is necessary to carry out pilot demonstrations of projects such as "virtual power plants" to the greatest extent, optimize resource allocation, and realize the pooling of protection measures resources through distributed storage and data synchronization technology, improve resource utilization, elasticity and flexibility, and realize automated interaction and dispatching management of resources.
[0003] At present, in terms of resource scheduling, the "cloud-edge-end" computing system structure is often used, involving the Internet of Things layer, edge computing layer and cloud computing layer. It basically includes static scheduling model, dynamic scheduling model, load balancing scheduling model, queue scheduling model, predictive scheduling model, etc. The static scheduling model is usually used for deterministic workloads and can be planned and scheduled when the system starts. The dynamic scheduling model can make up for the defects of the static scheduling model and allocate and schedule resources according to real-time resource requirements and system status. The load balancing scheduling model ensures that resources in the system are evenly distributed and avoids resource bottlenecks. The queue scheduling model is based on the concept of task queues, where tasks are queued for execution. Resource allocation is based on the priority of tasks in the queue, resource requirements and other strategies. The predictive scheduling model uses historical data and prediction algorithms to predict future resource requirements in order to better allocate resources. However, the scheduling schemes in the aforementioned scheduling models all have problems such as single indicators, unbalanced loads, and incompatible scheduling strategies. Summary of the invention
[0004] In order to solve the problems of single indicator, unbalanced load, and incompatibility of scheduling strategies in the existing scheduling model, the present invention proposes a virtual resource scheduling method for protection components based on strategy selection, including:
[0005] Determine whether to trigger resource scheduling policies based on the demand changes of the upper application layer and the performance status of the resource pool of the virtualization layer;
[0006] If the resource scheduling policy is triggered, virtual resources are scheduled based on the overall survivability of the resource pool performance, the rationality of the underlying device settings, or the resource load balancing. Otherwise, virtual resources are not scheduled.
[0007] Optionally, scheduling virtual resources from the perspective of overall survivability of resource pool performance or rationality of underlying device settings or resource load balancing includes:
[0008] From the perspective of the overall survivability of the resource pool performance, the allocation of virtual resources is determined by increasing or decreasing security device nodes based on the changes in the upper application layer's demand for virtual resources and the pre-built target constraint model of the resource pool management node consumption cost;
[0009] From the perspective of the rationality of the underlying device settings, based on the historical data of virtualized device load demand, the amount of reserved resources, the pre-built random integer programming model, and the pre-built weighted combination prediction model, the number of reservable virtual resources and abstracted underlying device models is determined to allocate virtual resources;
[0010] In terms of resource load balancing, the load balancing strategy of the improved Harris Eagle algorithm is used to migrate virtualized resources;
[0011] Among them, the random integer programming model is constructed with the objective function and constraints constructed by minimizing the virtualization cost and the maintenance cost suffered due to the attack;
[0012] The weighted combination forecasting model is constructed by using the improved double exponential smoothing method for basic forecasting and the incentive backup forecasting based on the time series forecasting method based on weighted moving average.
[0013] Optionally, the method of determining to increase or decrease security device nodes and allocate virtual resources based on the change in demand for virtual resources by the upper application layer and the target constraint model of the consumption cost of the resource pool management node built in advance from the aspect of the overall survivability of the resource pool performance includes:
[0014] When the demand for the upper application layer increases, the security device node with the lowest consumption cost is determined through the pre-built management node minimum consumption model, and the security device node with the lowest consumption cost is added to expand the virtual resources;
[0015] When the demand of the upper application layer decreases, the security device node with the highest consumption cost is determined through the pre-built maximum consumption model of the management node, and the security device node with the highest consumption cost under the resource pool management node is shut down;
[0016] Wherein, the target constraint model of the resource pool management node consumption cost includes: a management node minimum consumption model and a management node maximum consumption model;
[0017] The minimum consumption model of the management node and the maximum consumption model of the management node are both determined based on the energy consumption cost of a single security device node and the indicator utilization of the currently selected security device node.
[0018] Optionally, it also includes: based on business security and business importance, combining the entropy method to determine whether to adopt a fine-grained resource increase strategy for security device nodes.
[0019] Optionally, based on business security and business importance, the entropy method is combined to determine whether to adopt a fine-grained resource increase strategy for security device nodes, including:
[0020] Use the entropy method to design the critical value of resource pool performance;
[0021] Divide resource pool performance levels based on business security and business importance;
[0022] When the resource pool performance level is lower than the lower performance threshold of the resource pool performance threshold, the current resource pool service is in a good state;
[0023] When the resource pool performance level is between the lower performance threshold and the upper performance threshold of the resource pool performance threshold, the current resource pool service is in normal status;
[0024] When the resource pool performance level is greater than the upper performance threshold of the resource pool performance threshold, the current resource pool service is in a very poor state, and a fine-grained resource increase strategy is adopted for the security device node. Optionally, the rationality of the underlying device settings is determined based on the historical data of the virtualized device load demand, the reserved resource amount, the pre-built random integer programming model, and the pre-built weighted combination prediction model to determine the number of abstracted underlying device models and allocate virtual resources, including:
[0025] The demand forecast of virtualized equipment is obtained based on the historical data of virtualized equipment load demand and the pre-built weighted combination forecasting model;
[0026] When the demand forecast is greater than the reserved resource amount, determining the number of underlying device models based on the reserved resource amount combined with a pre-built random integer programming model;
[0027] The random integer programming model is constructed with the goal of minimizing the cost of the underlying device model and the constraints of the upper application layer service adjustment request.
[0028] Optionally, the random integer programming model is shown in the following formula:
[0029]
[0030] in, Indicates the number of reserved safety device models. represents the 0-1 decision variable in the resource reservation phase, P represents the number of abstractable security device models, p is the sequence number of the abstractable security device model, T represents the resource adjustment phase, t represents the sequence number of the time, represents the number of decision variables assigned to the upper application layer, represents the decision variable assigned to the upper application layer, represents the number of decision variables available for virtual resource demand planning, represents the decision variables available for virtual resource demand planning, Indicates the maximum resource capacity range that the set resource pool can accept, N0 indicates the total number of decision variables, represents the decision variable number, r represents the total number of decision variables allocated to the upper application layer, and R represents the total number of decision variables in the reserved stage.
[0031] Optionally, the number of the abstracted device models is calculated as follows:
[0032]
[0033] in, represents a 0-1 decision variable in the reservation phase, Indicates the number of reserved virtualized device models, and demand indicates the current virtual resource requirements.
[0034] Optionally, the demand forecast of the virtualized device is obtained based on the historical data of the virtualized device load demand and a pre-built weighted combination forecasting model, including:
[0035] Based on the time series of historical data of virtualized device load demand, the basic prediction of multiple virtualized devices is carried out using the improved double exponential smoothing method to obtain the basic prediction amount.
[0036] When the burst resource allocation of virtualized devices is insufficient, based on the fit of the historical data of the virtualized device load demand, the time series prediction method of weighted moving average is used to perform incentive backup prediction and obtain the incentive backup prediction amount;
[0037] The basic forecast amount and the incentive-based backup forecast amount are weightedly combined to obtain the demand forecast amount of the virtualized device.
[0038] Optionally, from the aspect of resource load balancing, the load balancing strategy of the improved Harris Eagle algorithm is adopted to migrate virtualized resources, including:
[0039] Decoding is performed based on the coded information of the communication port between the virtualized resource pool and the underlying device, and matching is performed to find the current resource pool management node corresponding to the underlying device;
[0040] Determine the sequence of virtualized devices to be migrated based on CPU and memory utilization;
[0041] The improved Harris Eagle optimization algorithm is used to reallocate the virtualized devices in the list to be migrated to reasonable resource pool management nodes and determine the locations of the virtualized devices.
[0042] Optionally, the improved Harris Hawk optimization algorithm is used to reallocate the virtualized devices in the to-be-migrated list to a reasonable resource pool management node to determine the location of the virtualized devices, including:
[0043] Step 1: Use the virtualized devices to be migrated in the sequence table of virtualized devices to be migrated as the Harris Hawk population and set the maximum number of iterations T m ;
[0044] Step 2: Calculate the fitness of each individual based on the fitness objective function, and sort the Harris hawk population according to the fitness;
[0045] Step 3: Select the individual with the best fitness as prey and update the escape energy factor E;
[0046] Step 4: Determine the search strategy to be adopted according to the change trend of the escape energy factor E, and update the Harris hawk population based on the search strategy, increase the number of iterations by 1, and return to step 3;
[0047] Step 5: Determine whether the number of iterations reaches the maximum number of iterations T m , if yes, go to step 6, otherwise, return to step 4;
[0048] Step 6: Determine the virtual device location by sorting the virtual devices in the latest Harris Hawk population;
[0049] Wherein, the fitness function is determined by the idea of Pareto optimality.
[0050] Optionally, the search strategy adopted according to the change trend of the escape energy factor E includes:
[0051] If the escape energy factor E>1, enter the search phase to analyze the change rate of the current escape energy factor E;
[0052] If E < 1, it enters the local arrest stage, and the comprehensive escape energy factor E and the convergence trend of individual fitness determine the prey's update position.
[0053] Optionally, if E < 1, the local arrest phase is entered, and the prey update position is determined by combining the escape energy factor E and the individual fitness convergence trend, including:
[0054] Strategy 1: If E>=0.5, R>=0.5, enter soft encirclement, the prey has physical strength but is eventually captured;
[0055] Strategy 2: If E < 0.5, R > = 0.5, it enters a hard encirclement, the prey does not have enough energy and is captured directly;
[0056] Strategy 3: If E>=0.5, R<0.5, E is decreasing rapidly, enter a fast dive soft encirclement, and the prey has the energy to escape. Use the escape energy factor formula to flexibly update the position and implement the encirclement strategy;
[0057] Strategy 4: If E<0.5, R<0.5, enter a fast dive hard encirclement. The prey does not have enough energy, and the prey is captured directly by continuously narrowing the encirclement.
[0058] Optionally, if the escape energy factor E>1, the normal search phase is entered to analyze the change speed of the current escape energy factor E, including:
[0059] If the change rate of the current escape energy factor E is greater than the set threshold, the original position update formula is used to update the individual position; otherwise, the forced search phase is entered, and the search space is dynamically adjusted according to the existing position update formula and the position of the search agent is updated.
[0060] Optionally, the escape energy factor formula is expressed as:
[0061]
[0062] In the formula, E0 represents the initial escape energy, t0 represents the initial value of energy in the range of (-1,1), and T m represents the maximum number of iterations, and st represents the number of times the prey recovers its physical strength during the escape process.
[0063] On the other hand, the present invention also provides a protection component virtual resource scheduling system based on policy selection, comprising:
[0064] A judgment module is used to determine whether to trigger a resource scheduling strategy based on the demand changes of the upper application layer and the performance status of the resource pool of the virtualization layer;
[0065] The resource scheduling module is used to schedule virtual resources from the perspective of the overall survivability of resource pool performance, the rationality of underlying device settings, or resource load balancing if the resource scheduling policy is triggered.
[0066] Optionally, the resource scheduling module includes:
[0067] The first resource scheduling submodule is used to determine the increase or decrease of security device nodes and the allocation of virtual resources from the perspective of the overall survivability of the resource pool performance, based on the changes in the demand for virtual resources by the upper application layer and the target constraint model of the consumption cost of the resource pool management node built in advance;
[0068] The second resource scheduling submodule is used to determine the number of reservable virtual resources and abstract underlying device models from the perspective of rationality of underlying devices, based on historical data of virtualized device load demand, reserved resource quantity, pre-built random integer programming model, and pre-built weighted combination prediction model, and allocate virtual resources;
[0069] The third resource scheduling submodule is used to migrate virtualized resources from the perspective of resource load balancing by adopting a load balancing strategy of an improved Harris Hawk algorithm.
[0070] Optionally, the first resource scheduling submodule includes:
[0071] The demand increase adjustment unit is used to determine the security device node with the lowest consumption cost through the pre-built management node minimum consumption model when the demand of the upper application layer increases, and to increase the security device node with the lowest consumption cost to expand the virtual resources;
[0072] The demand reduction adjustment unit is used to determine the security device node with the highest consumption cost through the pre-built maximum consumption model of the management node when the demand of the upper application layer decreases, and shut down the security device node with the highest consumption cost under the resource pool management node;
[0073] Wherein, the target constraint model of the resource pool management node consumption cost includes: a management node minimum consumption model and a management node maximum consumption model;
[0074] The management node minimum consumption model and the management node maximum consumption model are both determined based on the energy consumption cost of a single security device node and the indicator utilization of the currently selected security device node.
[0075] Optionally, the first resource scheduling submodule further includes a fine-grained resource adjustment unit configured to:
[0076] Use the entropy method to design the critical value of resource pool performance;
[0077] Divide resource pool performance levels based on business security and business importance;
[0078] When the resource pool performance level is lower than the lower performance threshold of the resource pool performance threshold, the current resource pool service is in a good state;
[0079] When the resource pool performance level is between the lower performance threshold and the upper performance threshold of the resource pool performance threshold, the current resource pool service is in normal status;
[0080] When the resource pool performance level is greater than the upper performance threshold of the resource pool performance threshold, the current resource pool service is in a very poor state, and a fine-grained resource increase strategy is adopted for the security device node.
[0081] Optionally, the second resource scheduling submodule is specifically used to:
[0082] The demand forecast of virtualized equipment is obtained based on the historical data of virtualized equipment load demand and the pre-built weighted combination forecasting model;
[0083] When the demand forecast is greater than the reserved resource amount, determining the number of underlying device models based on the reserved resource amount combined with a pre-built random integer programming model;
[0084] The random integer programming model is constructed with the goal of minimizing the cost of the underlying device model and the constraints of the upper application layer service adjustment request.
[0085] Optionally, the random integer programming model is shown in the following formula:
[0086]
[0087] in, Indicates the number of reserved safety device models. represents the 0-1 decision variable in the resource reservation phase, P represents the number of abstractable security device models, p is the sequence number of the abstractable security device model, T represents the resource adjustment phase, t represents the sequence number of the time, represents the number of decision variables assigned to the upper application layer, represents the decision variable assigned to the upper application layer, represents the number of decision variables available for virtual resource demand planning, represents the decision variables available for virtual resource demand planning, Indicates the maximum resource capacity range that the set resource pool can accept, N0 indicates the total number of decision variables, represents the decision variable number, r represents the total number of decision variables allocated to the upper application layer, and R represents the total number of decision variables in the reserved stage.
[0088] Optionally, the number of the abstracted device models is calculated as follows:
[0089]
[0090] in, represents a 0-1 decision variable in the reservation phase, Indicates the number of reserved virtualized device models, and demand indicates the current virtual resource requirements.
[0091] Optionally, the specific implementation steps of obtaining the demand forecast of the virtualized device based on the historical data of the virtualized device load demand combined with the pre-built weighted combination forecasting model in the second scheduling submodule include:
[0092] Based on the time series of historical data of virtualized equipment load demand, the improved double exponential smoothing method is used to make basic predictions of virtualized equipment and obtain basic prediction quantities.
[0093] When the burst resource allocation of virtualized devices is insufficient, based on the fit of the historical data of the virtualized device load demand, the time series prediction method of weighted moving average is used to perform incentive backup prediction.
[0094] The basic forecast amount and the incentive backup forecast amount are weightedly combined to obtain the demand forecast amount of the virtualized device.
[0095] Optionally, the third resource scheduling submodule includes:
[0096] A matching unit, configured to decode the coding information of the communication port between the virtualized resource pool and the underlying device, and match and search for a current resource pool management node corresponding to the underlying device;
[0097] A sequence determination unit, used to determine a sequence of virtualized devices to be migrated based on CPU and memory utilization;
[0098] The position determination unit is used to use the improved Harris Hawk optimization algorithm to reallocate the virtualized devices in the sequence of virtualized devices to be migrated to the resource pool management node and determine the position of the virtualized devices.
[0099] Optionally, the position determination unit is specifically used to:
[0100] Step 1: Use the virtualized devices to be migrated in the sequence table of virtualized devices to be migrated as the Harris Hawk population and set the maximum number of iterations T m ;
[0101] Step 2: Calculate the fitness of each individual based on the fitness objective function, and sort the Harris hawk population according to the fitness;
[0102] Step 3: Select the individual with the best fitness as prey and update the escape energy factor E;
[0103] Step 4: Determine the search strategy to be adopted according to the change trend of the escape energy factor E, and update the Harris hawk population based on the search strategy, increase the number of iterations by 1, and return to step 3;
[0104] Step 5: Determine whether the number of iterations reaches the maximum number of iterations T m , if yes, go to step 6, otherwise, return to step 4;
[0105] Step 6: Determine the virtual device location by sorting the virtual devices in the latest Harris Hawk population;
[0106] Wherein, the fitness function is determined by the idea of Pareto optimality.
[0107] Optionally, the specific implementation steps of the search strategy adopted according to the change trend of the escape energy factor E in the position determination unit include:
[0108] If the escape energy factor E>1, the normal search phase is entered to analyze the change speed of the current escape energy factor E;
[0109] If E < 1, it enters the local arrest stage, and the comprehensive escape energy factor E and the convergence trend of individual fitness determine the prey's update position.
[0110] Optionally, if E<1 in the position determination unit, the local arrest phase is entered, and the prey update position is determined by combining the escape energy factor E and the individual fitness convergence trend. The specific implementation steps include:
[0111] Strategy 1: If E>=0.5, R>=0.5, enter soft encirclement, the prey has physical strength but is eventually captured;
[0112] Strategy 2: If E < 0.5, R > = 0.5, it enters a hard encirclement, the prey does not have enough energy and is captured directly;
[0113] Strategy 3: If E>=0.5, R<0.5, E is decreasing rapidly, enter a fast dive soft encirclement, and the prey has the energy to escape. Use the escape energy factor formula to flexibly update the position and implement the encirclement strategy;
[0114] Strategy 4: If E<0.5, R<0.5, enter a fast dive hard encirclement. The prey does not have enough energy, and the prey is captured directly by continuously narrowing the encirclement.
[0115] Optionally, the escape energy factor formula is expressed as:
[0116]
[0117] In the formula, E0 represents the initial escape energy, t0 represents the initial value of energy in the range of (-1,1), and Tm represents the maximum number of iterations, and st represents the number of times the prey recovers its physical strength during the escape process.
[0118] Optionally, if the escape energy factor E>1 in the position determination unit, the normal search phase is entered to analyze the change speed of the current escape energy factor E, including:
[0119] If the change rate of the current escape energy factor E is greater than the set threshold, the original position update formula is used to update the individual position; otherwise, the forced search phase is entered, and the search space is dynamically adjusted according to the existing position update formula and the position of the search agent is updated.
[0120] In another aspect, the present application further provides a computing device, comprising: at least one processor and a memory;
[0121] The memory is used to store one or more programs;
[0122] When the one or more programs are executed by the at least one processor, a protection component virtual resource scheduling method based on policy selection as described above is implemented.
[0123] On the other hand, the present application also provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed, it implements a protection component virtual resource scheduling method based on policy selection as described above.
[0124] Compared with the prior art, the present invention has the following beneficial effects:
[0125] The present invention provides a virtual resource scheduling method for protection components based on policy selection, including determining whether to trigger a resource scheduling policy based on demand changes of an upper application layer and the operating status of a resource pool performance status node of a virtualization layer; when the resource scheduling policy is triggered, different scheduling policies are adopted to allocate and transfer virtual resources according to the virtualization cost of security device resources and the flexibility of business request requirements. The present invention adopts different scheduling policies according to the virtualization cost of security device resources and the flexibility of business request requirements, solving the problems of single indicator consideration, unbalanced load, and incompatibility of scheduling policies in existing scheduling models. BRIEF DESCRIPTION OF THE DRAWINGS
[0126] Figure 1 A flow chart of a virtual resource scheduling method for protection components based on policy selection according to the present invention;
[0127] Figure 2 The virtual resource scheduling flow chart of a specific embodiment of the present invention. DETAILED DESCRIPTION
[0128] The embodiment of the present invention provides a virtual resource scheduling method for protection components based on policy selection to solve the problem that the scheduling scheme in the existing scheduling model considers a single indicator and the load is unbalanced. The technical solution proposed by the present invention is as follows:
[0129] Embodiment 1:
[0130] A virtual resource scheduling method for protection components based on policy selection, such as Figure 1 As shown, including:
[0131] Step S1: Determine whether to trigger a resource scheduling policy based on the demand changes of the upper application layer and the resource pool performance status of the virtualization layer;
[0132] Step S2: When the resource scheduling policy is triggered, different scheduling policies are adopted to allocate and transfer virtual resources according to the virtualization cost of security device resources and the flexibility of service request requirements.
[0133] Before step S1, the method further includes: monitoring resource or service adjustment requests proposed by the upper application layer.
[0134] Step S1: Based on the demand changes of the upper application layer and the resource pool performance status of the virtualization layer, determine whether to trigger the resource scheduling strategy, specifically including:
[0135] In the virtual resource scheduling method of the protection component based on policy selection, first, the system continuously monitors various business activities of the upper application layer and the dynamic changes in resource requirements caused by them to determine the demand changes of the upper application layer. Various business activities and the dynamic changes in resource requirements caused by them include but are not limited to the growth or reduction of resource requirements such as computing power, storage space, and network bandwidth.
[0136] At the same time, the performance status of the resource pool at the virtualization layer is deeply explored, and the key performance indicators involved include but are not limited to the core performance parameters such as the overall CPU utilization, memory usage, I / O throughput, and network transmission rate of the resource pool. This step aims to fully understand the actual operation status of the resource pool and identify whether there are resource bottlenecks, excessive loads, or idle waste.
[0137] Furthermore, the system carefully analyzes the working status of the resource pool management nodes, evaluates their resource scheduling capabilities and efficiency under current and potential business pressures, and whether they face performance challenges when coordinating and managing the entire resource pool.
[0138] At the same time, the operating status of security device nodes is also closely monitored to ensure that security devices can operate efficiently, provide necessary security protection, and be ready to make corresponding resource adjustments based on business needs and security situation.
[0139] After completing all the above data collection and analysis, the system compares the changes in upper-level application demand with the actual performance status of the resource pool according to the preset decision rules and threshold standards to determine whether the conditions for triggering the resource scheduling policy have been met. Once there is a significant imbalance between resource demand and supply, the performance of the resource pool is insufficient to support business development, or the security device resources are insufficient to ensure the level of security protection, the resource scheduling policy will be triggered immediately.
[0140] Step S2: If the resource scheduling policy is triggered, the virtual resources are scheduled from the perspective of the overall survivability of the resource pool performance or the rationality of the underlying device settings or the resource load balancing, specifically including:
[0141] From the perspective of the overall survivability of the resource pool performance, the allocation of virtual resources is determined by increasing or decreasing security device nodes based on the changes in the upper application layer's demand for virtual resources and the pre-built target constraint model of the resource pool management node consumption cost;
[0142] From the perspective of the rationality of the underlying device settings, based on the historical data of virtualized device load demand, the amount of reserved resources, the pre-built random integer programming model, and the pre-built weighted combination prediction model, the number of reservable virtual resources and abstracted underlying device models is determined to allocate virtual resources;
[0143] In terms of resource load balancing, the load balancing strategy of the improved Harris Eagle algorithm is used to migrate virtualized resources;
[0144] Among them, the random integer programming model is constructed with the objective function and constraints constructed by minimizing the virtualization cost and the maintenance cost suffered due to the attack;
[0145] The weighted combination forecasting model is constructed by using the improved double exponential smoothing method for basic forecasting and the incentive backup forecasting based on the time series forecasting method based on weighted moving average.
[0146] Furthermore, from the perspective of the overall survivability of the resource pool performance, based on the changes in the upper application layer's demand for virtual resources combined with the pre-built target constraint model of the resource pool management node consumption cost, the increase or decrease of security device nodes is determined to allocate virtual resources, including:
[0147] When the demand for the upper application layer increases, the security device node with the lowest consumption cost is determined through the pre-built management node minimum consumption model, and the security device node with the lowest consumption cost is added to expand the virtual resources;
[0148] When the demand of the upper application layer decreases, the security device node with the highest consumption cost is determined through the pre-built maximum consumption model of the management node, and the security device node with the highest consumption cost under the resource pool management node is shut down;
[0149] Wherein, the target constraint model of the resource pool management node consumption cost includes: a management node minimum consumption model and a management node maximum consumption model;
[0150] The minimum consumption model of the management node and the maximum consumption model of the management node are both determined based on the energy consumption cost of a single security device node and the indicator utilization of the currently selected security device node.
[0151] Furthermore, it also includes: based on business security and business importance, combining the entropy method to determine whether to adopt a fine-grained resource increase strategy for the security device node.
[0152] Furthermore, the method of determining whether to adopt a fine-grained resource increase strategy for security device nodes based on business security and business importance in combination with an entropy method includes:
[0153] Use the entropy method to design the critical value of resource pool performance;
[0154] Divide resource pool performance levels based on business security and business importance;
[0155] When the resource pool performance level is lower than the lower performance threshold of the resource pool performance threshold, the current resource pool service is in a good state;
[0156] When the resource pool performance level is between the lower performance threshold and the upper performance threshold of the resource pool performance threshold, the current resource pool service is in normal status;
[0157] When the resource pool performance level is greater than the upper performance threshold of the resource pool performance threshold, the current resource pool service is in a very poor state, and a fine-grained resource increase strategy is adopted for the security device node.
[0158] Furthermore, from the aspect of the rationality of the underlying device settings, based on the historical data of the virtualized device load demand, the reserved resource amount, the pre-built random integer programming model, and the pre-built weighted combination prediction model, the number of abstracted underlying device models is determined to allocate virtual resources, including:
[0159] The demand forecast of virtualized equipment is obtained based on the historical data of virtualized equipment load demand and the pre-built weighted combination forecasting model;
[0160] When the demand forecast is greater than the reserved resource amount, the number of underlying security device models is determined based on the reserved resource amount combined with a pre-built random integer programming model;
[0161] The random integer programming model is constructed with the goal of minimizing the cost of the underlying device model and the constraints of the upper application layer service adjustment request.
[0162] Furthermore, the random integer programming model is shown in the following formula:
[0163]
[0164] in, Indicates the number of reserved safety device models. represents the 0-1 decision variable in the resource reservation phase, P represents the number of abstractable security device models, p is the sequence number of the abstractable security device model, T represents the resource adjustment phase, t represents the sequence number of the time, represents the number of decision variables assigned to the upper application layer, represents the decision variable assigned to the upper application layer, represents the number of decision variables available for virtual resource demand planning, represents the decision variables available for virtual resource demand planning, Indicates the maximum resource capacity range that the set resource pool can accept, N0 indicates the total number of decision variables, represents the decision variable number, r represents the total number of decision variables allocated to the upper application layer, and R represents the total number of decision variables in the reserved stage.
[0165] Furthermore, the number of the underlying device models is calculated as follows:
[0166]
[0167] in, represents a 0-1 decision variable in the reservation phase, Indicates the number of reserved virtualized device models, and demand indicates the current virtual resource requirements.
[0168] Furthermore, the historical data of the virtualized device load demand is combined with a pre-built weighted combination prediction model to obtain the demand forecast of the virtualized device, including:
[0169] Based on the sequential nature of the historical data of virtualized device load demand, the improved double exponential smoothing method is used to make basic predictions of virtualized devices and obtain basic prediction quantities.
[0170] When the burst resource allocation of virtualized devices is insufficient, based on the fit of the historical data of the virtualized device load demand, the time series prediction method of weighted moving average is used to perform incentive backup prediction and obtain the incentive backup prediction amount;
[0171] The basic forecast amount and the incentive backup forecast amount are weightedly combined to obtain the demand forecast amount of the virtualized device.
[0172] Furthermore, from the aspect of resource load balancing, the load balancing strategy of the improved Harris Hawk algorithm is adopted to migrate virtualized resources, including:
[0173] Decoding is performed based on the coded information of the communication port between the virtualized resource pool and the underlying device, and matching is performed to find the current resource pool management node corresponding to the underlying device;
[0174] Determine the sequence of virtualized devices to be migrated based on CPU and memory utilization;
[0175] The improved Harris Hawk optimization algorithm is used to reallocate the virtualized devices in the to-be-migrated column to the resource pool management nodes and determine the locations of the virtualized devices.
[0176] Furthermore, the improved Harris Hawk optimization algorithm is used to reallocate the resource pool management node of the virtualized device in the sequence of virtualized devices to be migrated to determine the location of the virtualized device, including:
[0177] Step 1: Use the virtualized devices to be migrated in the sequence table of virtualized devices to be migrated as the Harris Hawk population and set the maximum number of iterations T m ;
[0178] Step 2: Calculate the fitness of each individual based on the fitness objective function, and sort the Harris hawk population according to the fitness;
[0179] Step 3: Select the individual with the best fitness as prey and update the escape energy factor E;
[0180] Step 4: Determine the search strategy to be adopted according to the change trend of the escape energy factor E, and update the Harris hawk population based on the search strategy, increase the number of iterations by 1, and return to step 3;
[0181] Step 5: Determine whether the number of iterations reaches the maximum number of iterations T m , if yes, go to step 6, otherwise, return to step 4;
[0182] Step 6: Determine the virtual device location by sorting the virtual devices in the latest Harris Hawk population;
[0183] Wherein, the fitness function is determined by the idea of Pareto optimality.
[0184] Furthermore, the search strategy adopted according to the change trend of the escape energy factor E includes:
[0185] If the escape energy factor E is less than 1, the normal search phase is entered to analyze the change rate of the current escape energy factor E;
[0186] If E < 1, it enters the local arrest stage, and the comprehensive escape energy factor E and the convergence trend of individual fitness determine the prey's update position.
[0187] Furthermore, if E < 1, the local arrest phase is entered, and the prey update position is determined by combining the escape energy factor E and the individual fitness convergence trend, including:
[0188] Strategy 1: If E>=0.5, R>=0.5, enter soft encirclement, the prey has physical strength but is eventually captured;
[0189] Strategy 2: If E < 0.5, R > = 0.5, it enters a hard encirclement, the prey does not have enough energy and is captured directly;
[0190] Strategy 3: If E>=0.5, R<0.5, E is decreasing rapidly, enter a fast dive soft encirclement, and the prey has the energy to escape. Use the escape energy factor formula to flexibly update the position and implement the encirclement strategy;
[0191] Strategy 4: If E<0.5, R<0.5, enter a fast dive hard encirclement. The prey does not have enough energy, and the prey is captured directly by continuously narrowing the encirclement.
[0192] Furthermore, the escape energy factor formula is expressed as:
[0193]
[0194] Where E0 represents the initial escape energy, which is a random value between [-1,1], t0 represents the initial energy value between (-1,1), and T m represents the maximum number of iterations, and st represents the number of times the prey recovers its physical strength during the escape process.
[0195] Furthermore, if the escape energy factor E>1, the normal search phase is entered to analyze the change speed of the current escape energy factor E, including:
[0196] If the change rate of the current escape energy factor E is greater than the set threshold, the original position update formula is used to update the individual position; otherwise, the forced search phase is entered, and the search space is dynamically adjusted according to the existing position update formula and the position of the search agent is updated.
[0197] The comparison step S2 specifically includes the following contents:
[0198] When the resource scheduling policy is triggered, virtual resources are optimized and scheduled with the goal of optimizing resource utilization and reducing resource costs. That is, different scheduling policies are adopted according to the virtualization cost of security device resources and the flexibility of business request requirements to achieve efficient allocation and transfer of virtual resources;
[0199] Furthermore, from the two aspects of overall performance improvement of resource pool and adaptive allocation of virtual resources, different resource adaptive scheduling strategies are adopted according to the differentiated business requirements of the upper application layer.
[0200] Strategy 1: Considering the overall survivability of resource pool performance, adopt an adaptive resource scheduling strategy oriented to the overall survivability of resource pool, consider the imbalance between resource pool device nodes and resource demand allocation, and adaptively increase or decrease the use of virtualized devices.
[0201] Strategy 2: Considering the cost overhead of abstract processing of underlying devices and the reasonable setting of the number of devices, a multi-threaded virtual resource scheduling method that considers virtual resource reservation, prediction and integration is adopted to reduce energy consumption by adding additional device nodes or shutting down redundant device nodes.
[0202] Strategy 3: In the process of resource load balancing, from the perspective of improving the adaptability between the abstract device model and virtual resources and improving resource utilization, a load balancing strategy based on the improved Harris Eagle algorithm is proposed. On the premise of timely checking the management performance of key resource pool management nodes, the mapping relationship between resource pool management nodes (physical hosts) and virtualization security devices (sub-security device nodes) is determined through the Pareto optimal solution, thereby realizing the dynamic migration of virtualization resources.
[0203] Optionally, Strategy 1 is an adaptive resource scheduling strategy for the overall survivability of the resource pool. When the demand for virtual resources at the upper application layer increases or decreases, an elastic resource adaptive scheduling algorithm is designed to implement the expansion and release operations of virtualized device resources using the resource pool performance critical value as an evaluation condition.
[0204] When the demand for the upper application layer increases, the security device node with the lowest consumption cost is determined through the pre-built management node minimum consumption model, and the number of security device nodes with the lowest consumption cost (actual security devices) under the resource pool management node (master server) is increased to achieve coarse-grained expansion of virtualized device resources. If the actual performance value of the resource pool is greater than the upper limit critical value at this time, a fine-grained resource increase strategy is adopted for the security device node.
[0205] The minimum consumption model of the management node is:
[0206]
[0207] The target constraint model of the consumption cost of the resource pool security device node is:
[0208]
[0209] Among them, v jIndicates whether the current j-th security device node is selected. It is a 0-1 variable. m represents the total number of security device nodes managed by the resource pool management node. j is the security device node serial number. represents the performance utilization of the u-th indicator of the j-th security device node, c(p) represents the management cost of the resource scheduling stage, c(v) represents the energy consumption cost of a single security device node, J represents the set of security device nodes, u is the indicator sequence number, and n is the total number of indicators.
[0210] When the demand of the upper application layer decreases, that is, the resource pool performance value is less than the lower critical value, the security device node with the highest consumption cost is determined through the pre-built management node maximum consumption model, and the security device node with the highest consumption cost under the resource pool management node is closed, that is, the above objective function is converted to the maximum value.
[0211] The maximum consumption model of the management node is:
[0212]
[0213] The target constraint model of the consumption cost of the resource pool security device node is:
[0214]
[0215] In the formula, v j Indicates whether the current j-th security device node is selected, which is a 0-1 variable. represents the performance utilization of the u-th indicator of the j-th security device node, c(p) represents the management cost of the resource scheduling stage, c(v) represents the energy consumption cost of a single security device node, m represents the total number of security device nodes managed by the resource pool management node, J represents the security device node set, and j is the security device node serial number.
[0216] Optionally, the resource pool performance critical value is designed in combination with the entropy method, and when the resource demand of the upper application layer changes, the resource pool performance critical value changes.
[0217] Based on business security and business importance, resource pool performance levels are divided. If the level values P1 and P2 are lower than the lower performance critical value of the collected virtual resources, it means that the current resource pool service is in a good state; when it is between the lower performance critical value and the upper performance critical value, it means that it is in the current normal state; when it is greater than the upper performance critical value, it means that it is in a very poor state, and virtualized device resources of different granularities are allocated accordingly.
[0218] The following is a further introduction to business security and business importance:
[0219] The business security here is defined according to the sensitivity of the data involved in the business. It is divided into: low security, medium security and high security. Business data is not sensitive and is not disclosed to the public, which is low security; business data involves some user personal information or financial information, which is medium security; business data involves highly sensitive corporate financial information or network information, which is high security.
[0220] The importance of the business here is defined based on the duration of the business during the operation, the operation time, the operation loss, etc. For example, if there are no failures, interruptions, etc. during the operation, the business is defined as low importance; if failures occur during the operation, causing certain business losses, the business is defined as medium importance; if failures occur during the operation, causing certain business losses and affecting the user experience, resulting in a reduction in the size of the user population, the business is defined as high importance.
[0221] The resource pool performance levels are divided accordingly: low-security and low-importance businesses are matched with some resources that guarantee basic performance; medium-security and medium-importance businesses choose resources with a certain degree of privacy protection; high-security and high-importance businesses choose equipment resources with security guarantees and reliable data center architecture resources.
[0222] The critical value ΔP of resource pool performance is calculated as follows:
[0223]
[0224] Where P1 represents the probability evaluation before the upper application layer demand changes, and P2 represents the probability evaluation after the upper application layer demand changes. It indicates the optimal correlation degree determined by correlation analysis method before the upper application layer demand changes. It represents the optimal correlation degree determined by correlation analysis method after the upper application layer demand changes, μ max Indicates the maximum degree of association.
[0225] Furthermore, the strategy 2 is a multi-threaded virtual resource optimization strategy that considers resource reservation, prediction and integration, and the specific process is as follows:
[0226] Step 1: Considering the security device model that has been abstracted, a hybrid linear programming model is designed to determine the amount of virtual resources to be reserved and the number of abstract device models. Here, a reservation plan and demand strategy are formulated according to the demand of the upper application layer to reduce the cost consumed by security device virtualization.
[0227] For the abstract security device model, CPU and memory utilization are used as the main measurement factors. If the predicted amount is greater than the actual reserved resources, a random integer programming model is constructed with the minimum cost as the goal to determine the security device model with the minimum cost and the reservable virtual resources.
[0228] The random integer programming model is:
[0229]
[0230] in, Indicates the number of reserved safety device models. represents the 0-1 decision variable in the resource reservation phase, P represents the number of abstractable security device models, p is the sequence number of the abstractable security device model, T represents the resource adjustment phase, t represents the sequence number of the time, represents the number of decision variables assigned to the upper application layer, represents the decision variable assigned to the upper application layer, represents the number of decision variables available for virtual resource demand planning, represents the decision variables available for virtual resource demand planning, Indicates the maximum resource capacity range that the set resource pool can accept, N0 indicates the total number of decision variables, represents the decision variable number, r represents the total number of decision variables allocated to the upper application layer, and R represents the total number of decision variables in the reserved stage.
[0231] Step 2: Based on the amount of virtual resources reserved in step 1, and while satisfying the constraints of the current upper-layer application layer service adjustment request, the number of abstract device models is determined using a random integer programming model.
[0232]
[0233] in, represents a 0-1 decision variable in the reservation phase, Indicates the number of reserved virtualized device models, and demand indicates the current virtual resource requirements.
[0234] Step 3: Considering the resource reservation and demand forecast of virtualized security devices, a weighted combined forecasting model is proposed, which mainly includes basic forecasting and incentive backup forecasting, to quickly and accurately predict resource demand and realize the expansion or release of virtual resources.
[0235] Based on the time series of virtualized equipment load demand, the basic prediction model uses an improved double exponential smoothing method to predict the weight coefficient of the next stage based on the principle of minimum error. in, Represents the predicted weight coefficient value.
[0236]
[0237] In the formula, yt+T represents the forecast value for period t, represents the exponential smoothing value of the tth period, It represents the quadratic exponential smoothing value of the tth period, and α is the actual weight coefficient value, which can be set according to the actual data.
[0238] Considering the fitting of virtualized device load demand, the incentive backup prediction aims to predict the problem of insufficient resource allocation when virtualized devices have sudden charge load problems, and mainly uses the time series prediction method EMA of weighted moving average to adjust.
[0239]
[0240] Where P represents the number of abstractable security device models. represents the predicted value based on the double exponential smoothing method, represents the predicted value based on the EMA method, x t It represents the actual value obtained based on the EMA method in the tth period, ES represents the double exponential smoothing method, and EMA represents the exponential moving average.
[0241] Optionally, Strategy 3 proposes a dynamic load balancing strategy for resource pool nodes, and improves the Harris Eagle optimization algorithm to optimize the virtual resource loss rate and performance of different devices to determine the resource pool management node (main server) and the security device node (actual security device) (purpose: accurately locate the virtual security device position), and the specific process is as follows:
[0242] Step 1: Decode based on the virtualized resource pool communication port encoding information and match to find the current resource pool management node.
[0243] Step 2: Determine the sequence of virtualized devices to be migrated. Using CPU and memory utilization, select virtualized devices with high idle rates and low loads as the migration object sequence. Here, the virtualized device idle rate is defined as:
[0244]
[0245] in, represents the utilization of the u-th performance indicator of the j-th virtualized device, v j Indicates whether the current j-th virtualization device is selected. It is a 0-1 variable. m represents the total number of virtualization devices belonging to the current resource pool management node. J represents the virtualization device set. j represents the serial number of the virtualization device belonging to the current resource pool management node. u represents the serial number of the performance indicator. n represents the total number of performance indicators.
[0246] Step 3: Determine the location of virtualized devices. Reallocate the virtualized devices in the list to be migrated to reasonable resource pool management nodes to achieve load balancing of virtual resources. This is mainly achieved by using the improved Harris Eagle optimization algorithm, and setting the fitness function through the Pareto optimal idea to achieve management mapping from virtualized devices to resource pool management nodes.
[0247] Specifically, the fitness function is set to optimize resource utilization and reduce the energy consumption of the resource pool, and to find the Pareto optimal solution and the weak Pareto optimal solution based on the Pareto principle. The Pareto optimal solution and the weak Pareto optimal solution are determined from the two aspects of full load and zero load of the resource pool.
[0248] The full load indicates an ideal state in which the current resource pool management node has fully allocated virtual resources. The Pareto optimal solution λ under the full load state good It can be expressed as:
[0249]
[0250] Where w1 represents the weight of the ratio of the maximum energy consumption of the system to the number of abstractable security device models, P + represents the maximum energy consumption of the system, P represents the number of abstractable security device models, and w2 represents The weight of Indicates the maximum usage cost of the system, U rest Indicates the number of available safety device models.
[0251] The zero load state indicates that the resource pool does not occupy any virtual resources, and a weak Pareto optimal solution can be obtained. bad for:
[0252]
[0253] Where P - Indicates the minimum energy consumption of the system, represents the minimum system usage overhead, λ bad represents a weak Pareto optimal solution.
[0254] Based on the above two solutions, the final fitness objective function fitness is:
[0255]
[0256] Optionally, an improved Harris Hawk algorithm is proposed in which the escape energy factor E is adjusted and updated according to the current number of iterations, and the search strategy to be adopted is determined according to the change trend of the escape energy factor.
[0257] If E>1, enter the search phase. Specifically, analyze the change speed of the current escape energy factor. If it decreases rapidly, it means that the prey consumes a lot of energy during the escape process, and the original position update formula is used to update the individual position; if it changes slowly, it means that the prey is in the energy recovery stage, and it can be forced to enter the search phase. According to the following formula, the position of the search agent is updated to dynamically adjust the search space.
[0258]
[0259] Among them, X max and X min are the individuals farthest and closest to the average position of the population, respectively; Rand is a random number between (0,1); ub and lb are the upper and lower limits, respectively; X(t) represents the improved Harris Hawk algorithm; q is used to randomly select the strategy to be used, which is a random number between 0 and 1.
[0260] If E < 1, it enters the local arrest stage, and the prey's update position is determined by the comprehensive escape energy factor and the individual fitness convergence trend. It can be divided into the following four situations:
[0261] Strategy 1: If E>=0.5, R>=0.5, enter soft encirclement, that is, the prey has physical strength but will eventually be captured.
[0262] Strategy 2: If E < 0.5, R > = 0.5, it enters a hard encirclement, that is, the prey does not have sufficient energy and is captured directly.
[0263] Strategy 3: If E>=0.5, R<0.5, E decreases rapidly and enters a rapid dive soft encirclement, that is, the prey has the physical strength to escape. The escape energy factor formula is used to flexibly update the position and implement an encirclement strategy.
[0264] Strategy 4: If E < 0.5, R < 0.5, enter a fast dive hard encirclement, that is, the prey does not have sufficient energy, and the prey can be directly captured by continuously narrowing the encirclement range.
[0265] Optionally, the escape energy factor formula is expressed as:
[0266]
[0267] In the formula, t0 represents the initial energy value between (-1,1), T m represents the maximum number of iterations, st represents a random number in the range of (20,100), which determines the number of times the prey recovers its physical strength during the escape process, E0 represents the initial escape energy, which is a random value between [-1,1], and E represents the escape energy factor.
[0268] The present invention proposes a multi-stage resource optimization method based on dynamic security and resource elastic configuration to determine the reasonable type and quantity of virtualized devices, which can not only protect the system security and reduce the possibility of being attacked, but also adjust the resource amount according to business requests to reduce the waste of virtual resources. Specifically, the resource reservation amount of the same type of virtualized devices is calculated based on the random integer programming model; secondly, for various allocation methods of virtualized resources, considering the possible attacks faced by virtualized devices and the dynamic demand changes based on business adjustments, the appropriate virtualized resource allocation plan is solved based on the principle of minimizing the equipment virtualization cost and maintenance cost.
[0269] Embodiment 2:
[0270] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.
[0271] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.
[0272] In view of the resource conflict management problem existing in the existing resource scheduling technology, an embodiment of the present invention provides a virtual resource scheduling method for protection components based on policy selection, including:
[0273] Request monitoring (Monitor) module: monitors resource or service adjustment requests proposed by the upper application layer, analyzes service priorities, obtains resource adjustment requests proposed by the upper application layer, and establishes a resource adjustment request model.
[0274] Analysis module: Combines the demand changes of the upper application layer and the performance status of the resource pool of the virtualization layer to analyze the operating status of the resource pool management node and the security device node to determine whether to trigger the resource scheduling strategy;
[0275] Resource Dispatch Module: If triggered, virtual resources are optimally dispatched with the goal of optimizing resource utilization and reducing resource costs. That is, different dispatch strategies are adopted according to the virtualization cost of security device resources and the flexibility of business request requirements to achieve efficient allocation and transfer of virtual resources.
[0276] Specifically, in the embodiment of the present invention, the target system is an industrial cyber-physical system, and the software components in the industrial cyber-physical system are dynamically combined across heterogeneous surface source devices and can perform various functions. Figure 2 The present invention is described in detail. The embodiment of the present invention includes the following steps S100-S400;
[0277] S100: Obtain a resource adjustment request from an upper security application layer, and establish a resource adjustment request model according to the resource adjustment request.
[0278] Specifically, step S100 includes steps S110-S130:
[0279] S110: Obtain resource adjustment request of upper security application layer
[0280] Specifically, a resource adjustment request of the upper security application layer is obtained. The information sequence of such resource adjustment request includes the name of the resource to be adjusted, the number of instructions for requesting the service, the performance parameters of the requested service, the memory required for the requested service, the CPU resources required for the requested service, the network bandwidth utilization required for the requested service, etc.
[0281] S120: According to the resource adjustment request service name, the matching resource pool management node, security device node and sequence number of the resource to be adjusted are retrieved to form a resource pool node sequence and a resource sequence.
[0282] Specifically, the resource pool management node that manages the current resource to be adjusted is queried in the corresponding resource pool by category through the name of the resource to be adjusted, a sequence of resource pool management nodes is obtained, and all current resource adjustment requests are taken as a resource adjustment set.
[0283] S130: Constructing a resource adjustment request triplet for the upper application layer according to the resource pool node sequence and resource sequence
[0284] FHFW={U,S,C}. The resource pool node sequence can be expressed as:
[0285] C={M,c1,c2,...,c i ,...,c m}
[0286] Where M represents the only resource pool management node, m represents the total number of adapted security device nodes in the cloud network environment; c i represents the i-th security device node, FHFW represents the resource adjustment request triplet, C represents the resource pool node sequence, and S represents the resource sequence.
[0287] The underlying virtualization security device generates a series of virtual resources when running.
[0288] The expression of virtualized device sequence can be expressed as: U = {u1,u2,...,u j ,...,u n}
[0289] Among them, U represents the set of virtualized devices, n represents the total number of performance indicators, and u j Represents the jth virtualized device, and virtualized device u j The virtual resource sequence in the relevant resource pool is u j ={s j1 ,s j2 ,…,s jk}, k represents the total number of virtual resources generated under a single virtualization device, where s j1 Indicates the first virtual resource in the virtual resource sequence, s j2 Indicates the second virtual resource in the virtual resource sequence, s jk Represents the kth virtual resource in the virtual resource sequence.
[0290] The above resource sequence is S, where the virtualized device u j Virtual resources generated by client registration jk Through the security device node c i After access is approved and services are provided externally, it can be represented as s jki Therefore, the resource service of the ith virtualized device can be expressed as c i ={s 11i ,s 12i ,...,s mni}, where s 11i Indicates the virtual resource s generated by the registration of the virtualized device u1 11 Through the security device node c i After access is approved, the services provided to the outside world are 12i Indicates the virtual resource s generated by the registration of the virtualized device u1 12 Through the security device node c i After access is approved, the services provided to the outside world are mni Indicates virtualized device u m Virtual resources generated by client registration mn Through the security device node c i Services provided to the outside world after access is approved.
[0291] S200: Constructing a resource request scheduling model for a virtualized device according to the resource pool node sequence and resource sequence.
[0292] Specifically, according to the resource request of the device, the execution time and execution cost required to complete the resource scheduling service are analyzed. Since the execution time and cost of each virtualized device will vary depending on the underlying security device, and the underlying devices are independent of each other, the resource scheduling and allocation problem can be decomposed into many scheduling sub-problems. (Purpose: To predict the load demand of the underlying security device at each stage, and provide a basis for determining the reservable resources in the next step)
[0293] S210: Considering the resource reservation and demand forecast of each underlying security device, a weighted combination forecasting model is proposed to effectively avoid resource shortage and resource waste, which mainly includes basic forecasting and incentive backup forecasting.
[0294] Based on the time series of virtualized equipment load demand, the basic prediction model uses an improved double exponential smoothing method to predict the weight coefficient of the next stage based on the principle of minimum error.
[0295]
[0296] In the formula, y t+T represents the forecast value for period t, represents the exponential smoothing value of the tth period, It represents the quadratic exponential smoothing value of the tth period, μ represents the actual weight coefficient value, which can be set according to the actual data, t represents the sequence number of the forecast period, and T represents the resource adjustment stage.
[0297] Considering the fitting of virtualized device load demand, the incentive backup prediction aims to predict the problem of insufficient resource allocation when virtualized devices have sudden charge load problems, and mainly uses the time series prediction method EMA of weighted moving average to adjust.
[0298]
[0299] Where P represents the number of abstractable security device models. represents the predicted value based on the double exponential smoothing method, Represents the predicted value based on the EMA method.
[0300] S300: Considering the flexibility and scalability of virtual resource scheduling of virtual devices after virtualization of security devices, a multi-stage resource optimization method based on dynamic security and resource elasticity configuration is adopted. It can not only protect system security and reduce the possibility of being attacked, but also adjust the amount of resources according to business requests and reduce the waste of virtual resources. (Purpose: To determine the reasonable type and number of virtualized devices)
[0301] S310: Virtualization device reservation stage
[0302] Considering the possible attacks faced by virtualized devices and the dynamic demand changes based on business adjustments, a random integer programming model is designed to solve the problem of reserving virtualized device resources based on the principle of minimizing device virtualization cost and maintenance cost. Firstly, the resource reservation of the same type of virtualized devices is calculated based on the random integer programming model; secondly, the appropriate virtualized resource allocation scheme is solved for various allocation methods of virtualized resources.
[0303] Specifically, the objective function in the random integer programming model is to minimize the virtualization cost and the maintenance cost caused by the attack.
[0304]
[0305] In the formula, z represents the multi-objective optimization minimum value of virtualization cost and maintenance cost caused by attack, Indicates the number of reserved virtualized devices. represents the 0-1 decision variable related to virtualized devices in the reservation phase, ξ(x,ω) represents the virtualization cost of virtualized devices under the condition of uncertain service layer demand, Represents the maintenance cost faced by virtualized devices when the attack type is unknown.
[0306] Constraint 1: The virtualization device reserved resources that the enterprise can set up in the initial stage.
[0307]
[0308] Constraint 2: The virtualized device variable is within the upper limit N0.
[0309]
[0310] Constraint 3: The used resources of the virtualized device must be within the maximum load capacity of the system limit within the range.
[0311]
[0312] in, represents the decision variables assigned to the underlying security devices, represents the resource decision variable in the system demand plan, T represents the resource adjustment stage, P represents the virtualizable device supply in the underlying security device, Indicates the maximum acceptable resource capacity range of the system.
[0313] The above model has obtained a feasible solution for the reserved virtualization device resources. Here, the random integer programming model is further used to solve the optimal solution within the feasible solution range. Due to the variety of virtualization device types, in addition to reducing the cost of abstracting the underlying security devices into virtualization devices, it is also necessary to determine the virtualization device types and the actual applicable number of the system.
[0314] Constraints: The number of underlying security devices reserved above must be greater than the business demand.
[0315]
[0316] in, represents a 0-1 decision variable in the reservation phase, Indicates the number of reserved virtualized devices, and demand indicates the current resource request and the resource demand to be met by the business.
[0317] S320: Automatic resource adjustment phase
[0318] Dynamically adjust the number of virtualized devices based on the predicted demand. If the predicted demand is greater than the actual demand, consider adding virtualized devices with the lowest cost. The optimization objectives are as follows:
[0319]
[0320] Where c(v) represents the energy consumption cost of a single underlying device abstracted as a virtualized device. represents the performance utilization of the u-th indicator of the j-th virtualized device, v j Indicates whether the current j-th virtualization device is required by the system. It is a 0-1 variable. m represents the total number of virtualization devices belonging to the current resource pool management node. J represents the virtualization device set.
[0321] If the predicted amount is less than the actual amount, consider reallocating virtual resources and shutting down the virtualization device with the largest consumption. The optimization goals are as follows:
[0322]
[0323] S330: Resource integration stage
[0324] In order to reduce the energy consumption of equipment and address the issue of equipment placement in power grid virtualization, the energy consumption factor k1 is introduced to model the server consumption and further standardize the solution.
[0325] P=k1×P max +(1-k1)×P max ×UT CPU
[0326] In the formula, P represents the number of abstractable security device models, k1 represents the energy consumption factor, and P max Indicates the total number of safety device models, UT CPU Indicates CPU utilization.
[0327] Specifically, in the actual resource integration process, the optimization target is designed considering SLA and server energy consumption overhead.
[0328] Objective function: maxαF SLA +βF power
[0329] Constraints: CPU utilization and memory utilization are both greater than 0, indicating that the current virtualization device is available.
[0330] maxαF SLA +βF power
[0331] stUT CPU ≥0,UT neicun ≥0
[0332] .
[0333] .
[0334] In the formula, α represents the proportion of SLA energy consumption in the total energy consumption, F SLA represents the energy consumption cost of SLA, β represents the proportion of the server energy consumption cost in the total energy consumption cost, and F power Represents the energy consumption of the server, UT neicun Indicates memory utilization, UT mem Indicates storage utilization.
[0335] S400: Based on the types and number of virtualized devices determined in S300, a dynamic load balancing strategy based on sequence operation is proposed, and the Harris Eagle optimization algorithm is improved to optimize the resource loss rate and performance of different levels to determine the hotspot hosts and virtualized devices to be migrated (purpose: to accurately locate the location of virtualized devices)
[0336] S410: Determine the hotspot host.
[0337] Step 1: Obtain the indicator information of each type of virtualized device at different times. Specifically, collect the key indicator information of the virtualized device at each interval t, construct a decision matrix of [m,n], and then perform normalization to obtain the standardized decision matrix K.
[0338] Step 2: Obtain the sovereignty value corresponding to each indicator at different times, and sort the importance according to the information and weight value of each indicator.
[0339] Step 3: Perform convolution operations on the importance sequence of each type of indicator to obtain multiple key indicator sequences
[0340] Step 4: Determine the comprehensive evaluation value of the host based on multiple key indicator sequences, so as to determine the hotspot host.
[0341] S420: Determine the sequence of virtualized devices to be migrated
[0342] Using the CPU and memory utilization indicators, select virtualized devices with low idle utilization as migration targets, and migrate the physical security devices mapped to them. Define the idle virtual machine utilization as:
[0343]
[0344] in, represents the performance utilization of the u-th indicator of the j-th virtualized device, v j Indicates whether the current j-th virtualized device is needed by the system. It is a 0-1 variable. m represents the total number of virtualized devices belonging to the current resource pool management node. J represents the virtualized device set.
[0345] S430: Determine the location of the underlying security device
[0346] The virtualized devices in the list to be migrated are divided, matched to the actual underlying security devices, and redistributed to the reasonable resource pool management nodes to achieve load balancing. The improved Harris Eagle optimization algorithm is mainly used to implement it, and the fitness function is set through the Pareto optimal idea to achieve the mapping from virtualized devices to underlying security devices and to resource pool management nodes.
[0347] Specifically, the fitness function is set to optimize resource utilization and reduce energy consumption, and to find Pareto optimal solutions and weak Pareto optimal solutions based on the Pareto principle. The Pareto optimal solution and weak Pareto optimal solution are determined from two aspects: full load and zero load.
[0348] Among them, full load indicates the ideal state in which the current resource pool management node has fully and rationally allocated resources.
[0349]
[0350] The zero-load state indicates that the resource pool management node does not occupy additional resources, and a weak Pareto optimal solution can be found.
[0351]
[0352] Based on the above two solutions, the final fitness objective function is:
[0353]
[0354] The specific Harris Hawk improved algorithm is as follows:
[0355] Step 1: Parameter initialization setting. Set the population size N, the maximum number of iterations T m , spatial dimension M, population search range [ub, lb], escape energy factor E and jump intensity J.
[0356] Step 2: Population initialization settings. Initialize the population position, calculate the individual fitness and use the individual fitness function as the objective function, and store the current optimal solution and suboptimal solution.
[0357] Step 3: Determine the termination condition. If the maximum number of iterations T is reached m , then terminate and output the optimal position, otherwise return and re-execute the next step to continue the search.
[0358] Step 4: Adjust the new search strategy to determine the optimal prey location. Consider both the optimal solution and the suboptimal solution during the search process, and construct an elite neighborhood by combining the current individual, the optimal individual, and the suboptimal individual in pairs. Calculate the fitness values of the three types of individuals in the elite neighborhood. If the maximum fitness value of the current individual is still less than the fitness value of the set prey, it means that the optimization result in the current elite neighborhood is more ideal, so the individual with the minimum fitness among the three types of individuals is selected as the current optimal population prey.
[0359] Step 5: Adjust the current escape energy factor and update the escape ability factor and escape step size according to the current number of iterations.
[0360]
[0361] In the formula, E0 represents the initial escape energy, t0 represents the initial value of energy in the range of (-1,1), and T m represents the maximum number of iterations, and st represents the number of times the prey recovers its physical strength during the escape process.
[0362] Step 6: Determine which search strategy to adopt based on the change trend of the escape energy factor.
[0363] If E>1, enter the search phase.
[0364] Specifically, analyze the changing speed of the current escape energy factor. If it decreases rapidly, it means that the prey consumes a lot of energy during the escape process, and the original position update formula is used to update the individual position. If it changes slowly, it means that the prey is in the energy recovery stage, and it can be forced to enter the search stage. The position of the search agent is updated according to the following formula to dynamically adjust the search space.
[0365]
[0366] Where X(t) represents the location of the new search agent, X max and X min are the individuals farthest and closest to the average position of the population respectively; Rand is a random number located in (0, 1); ub and lb are the upper and lower limits respectively.
[0367] If E < 1, it enters the local arrest stage, and the prey's update position is determined by the comprehensive escape energy factor and the individual fitness convergence trend. It can be divided into the following four situations:
[0368] Strategy 1: If E>=0.5, R>=0.5, enter soft encirclement, that is, the prey has physical strength but is eventually captured.
[0369] Strategy 2: If E<0.5, R>=0.5, enter hard encirclement, that is, the prey does not have sufficient energy and is captured directly.
[0370] Strategy 3: If E>=0.5, R<0.5, and E decreases rapidly, enter a rapid dive soft encirclement, that is, the prey has the strength to escape, use the following formula to flexibly update the position and implement the encirclement strategy.
[0371] Strategy 4: If E<0.5, R<0.5, enter a fast dive hard encirclement, that is, the prey does not have sufficient energy, and the prey can be directly captured by continuously narrowing the encirclement range.
[0372] The existing resource scheduling model has a wide range of applications. Its goal is usually to maximize resource utilization or to achieve the optimization of a certain performance indicator under specific constraints, and to combine dynamic programming models and heuristic algorithms to achieve model solving. In a virtual resource environment, the scheduling model generally implements the dynamic allocation and sharing of different physical resources based on virtualization technology. The introduction of an additional virtualization layer in the scheduling process will bring about certain performance losses and have high requirements for the real-time performance of the scene. In order to solve the problem of unbalanced load of scheduling schemes in scheduling models, a virtual resource scheduling method for protection components based on policy selection is proposed in the present invention to determine the reasonable type and number of virtualized devices, which can protect system security and reduce the possibility of attacks, while considering the business request to adjust the amount of resources and reduce the waste of virtual resources. Specifically, the resource reservation amount of the same type of virtualized devices is obtained based on the random integer programming model; secondly, for a variety of allocation methods of virtualized resources, the attacks that virtualized devices may face and the dynamic demand changes based on business adjustments are considered, and the appropriate virtualized resource allocation scheme is solved based on the principle of minimizing the equipment virtualization cost and maintenance cost.
[0373] In addition, the present invention realizes the all-round management of virtual resources from multiple aspects such as virtual resource scheduling, resource integration, and underlying resource equipment positioning, thereby improving the overall resource utilization and effectively reducing the energy consumption of underlying resources. Specifically, a random integer programming model is considered to solve the reserved protection component resources, and the energy consumption is reduced by adding additional device nodes or shutting down redundant device nodes. The CPU and memory utilization are used as the main measurement factors. If the predicted amount is greater than the actual reserved resource amount, a random integer programming model is constructed with the minimum cost as the goal to determine the minimum cost security device model and the reservable virtual resources.
[0374] Embodiment 3:
[0375] The present invention based on the same inventive concept also provides a protection component virtual resource scheduling system based on policy selection, including:
[0376] A judgment module is used to determine whether to trigger a resource scheduling strategy based on the demand changes of the upper application layer and the performance status of the resource pool of the virtualization layer;
[0377] The resource scheduling module is used to schedule virtual resources from the perspective of the overall survivability of resource pool performance, the rationality of underlying device settings, or resource load balancing when the resource scheduling policy is triggered.
[0378] Optionally, the resource scheduling module includes:
[0379] The first resource scheduling submodule is used to determine the increase or decrease of security device nodes and the allocation of virtual resources from the perspective of the overall survivability of the resource pool performance, based on the changes in the demand for virtual resources by the upper application layer and the target constraint model of the consumption cost of the resource pool management node built in advance;
[0380] The second resource scheduling submodule is used to determine the number of reservable virtual resources and abstract underlying device models from the rationality of the underlying devices, based on the historical data of virtualized device load demand, the amount of reserved resources, the pre-built random integer programming model, and the pre-built weighted combination prediction model, and allocate virtual resources;
[0381] The third resource scheduling submodule is used to migrate virtualized resources from the perspective of resource load balancing by adopting a load balancing strategy of an improved Harris Hawk algorithm.
[0382] Optionally, the first resource scheduling submodule includes:
[0383] The demand increase adjustment unit is used to determine the security device node with the lowest consumption cost through the pre-built management node minimum consumption model when the demand of the upper application layer increases, and to increase the security device node with the lowest consumption cost to expand the virtual standby resources;
[0384] The demand reduction adjustment unit is used to determine the security device node with the highest consumption cost through the pre-built maximum consumption model of the management node when the demand of the upper application layer decreases, and shut down the security device node with the highest consumption cost under the resource pool management node;
[0385] Wherein, the target constraint model of the resource pool management node consumption cost includes: a management node minimum consumption model and a management node maximum consumption model;
[0386] The minimum consumption model of the management node and the maximum consumption model of the management node are both determined based on the energy consumption cost of a single security device node and the indicator utilization of the currently selected security device node.
[0387] Optionally, a fine-grained resource adjustment unit is also included for:
[0388] Use the entropy method to design the critical value of resource pool performance;
[0389] Divide resource pool performance levels based on business security and business importance;
[0390] When the resource pool performance level is lower than the lower performance threshold of the resource pool performance threshold, the current resource pool service is in a good state;
[0391] When the resource pool performance level is between the lower performance threshold and the upper performance threshold of the resource pool performance threshold, the current resource pool service is in normal status;
[0392] When the resource pool performance level is greater than the upper performance threshold of the resource pool performance threshold, the current resource pool service is in a very poor state, and a fine-grained resource increase strategy is adopted for the security device node.
[0393] Optionally, the second resource scheduling submodule is specifically used to:
[0394] The demand forecast of virtualized equipment is obtained based on the historical data of virtualized equipment load demand and the pre-built weighted combination forecasting model;
[0395] When the demand forecast is greater than the reserved resource amount, determining the number of underlying device models based on the reserved resource amount combined with a pre-built random integer programming model;
[0396] The random integer programming model is constructed with the goal of minimizing the cost of the underlying device model and the constraints of the upper application layer service adjustment request.
[0397] Optionally, the random integer programming model is shown in the following formula:
[0398]
[0399] in, Indicates the number of reserved safety device models. represents the 0-1 decision variable in the resource reservation phase, P represents the number of abstractable security device models, p is the sequence number of the abstractable security device model, T represents the resource adjustment phase, t represents the sequence number of the time, represents the number of decision variables assigned to the upper application layer, represents the decision variable assigned to the upper application layer, represents the number of decision variables available for virtual resource demand planning, represents the decision variables available for virtual resource demand planning, Indicates the maximum resource capacity range that the set resource pool can accept, N0 indicates the total number of decision variables, represents the decision variable number, r represents the total number of decision variables allocated to the upper application layer, and R represents the total number of decision variables in the reserved stage.
[0400] Optionally, the number of abstracted device models is calculated as follows:
[0401]
[0402] in, represents a 0-1 decision variable in the reservation phase, Indicates the number of reserved virtualized device models, and demand indicates the current virtual resource requirements.
[0403] Optionally, the specific implementation steps of obtaining the demand forecast of the virtualized device based on the historical data of the virtualized device load demand combined with the pre-built weighted combination forecasting model in the second scheduling submodule include:
[0404] Based on the time series of historical data of virtualized equipment load demand, the improved double exponential smoothing method is used to make basic predictions of virtualized equipment and obtain basic prediction quantities.
[0405] When the virtualization device burst resource allocation is insufficient, based on the fit of the historical data of the virtualization device load demand, the weighted moving average time series prediction method is used to perform incentive backup prediction to obtain the predicted resource demand;
[0406] The basic forecast amount and the incentive backup forecast amount are weightedly combined to obtain the demand forecast amount of the virtualized device.
[0407] Optionally, the third resource scheduling submodule includes:
[0408] A matching unit, configured to decode the coding information of the communication port between the virtualized resource pool and the underlying device, and match and search for a current resource pool management node corresponding to the underlying device;
[0409] A sequence determination unit, used to determine a sequence of virtualized devices to be migrated based on CPU and memory utilization;
[0410] The position determination unit is used to use the improved Harris Hawk optimization algorithm to reallocate the virtualized devices in the sequence of virtualized devices to be migrated to the resource pool management node and determine the position of the virtualized devices.
[0411] Optionally, the position determination unit is specifically used to:
[0412] Step 1: Use the virtualized devices to be migrated in the sequence table of virtualized devices to be migrated as the Harris Hawk population and set the maximum number of iterations T m ;
[0413] Step 2: Calculate the fitness of each individual based on the fitness objective function, and sort the Harris hawk population according to the fitness;
[0414] Step 3: Select the individual with the best fitness as prey and update the escape energy factor E;
[0415] Step 4: Determine the search strategy to be adopted according to the change trend of the escape energy factor E, and update the Harris hawk population based on the search strategy, increase the number of iterations by 1, and return to step 3;
[0416] Step 5: Determine whether the number of iterations reaches the maximum number of iterations T m , if yes, go to step 6, otherwise, return to step 4;
[0417] Step 6: Determine the virtual device location by sorting the virtual devices in the latest Harris Hawk population;
[0418] Wherein, the fitness function is determined by the idea of Pareto optimality.
[0419] Optionally, the specific implementation steps of the search strategy adopted according to the change trend of the escape energy factor E in the position determination unit include:
[0420] If the escape energy factor E>1, the normal search phase is entered to analyze the change speed of the current escape energy factor E;
[0421] If E < 1, it enters the local arrest stage, and the comprehensive escape energy factor E and the convergence trend of individual fitness determine the prey's update position.
[0422] Optionally, if E<1 in the position determination unit, the local arrest phase is entered, and the prey update position is determined by combining the escape energy factor E and the individual fitness convergence trend. The specific implementation steps include:
[0423] Strategy 1: If E>=0.5, R>=0.5, enter soft encirclement, the prey has physical strength but is eventually captured;
[0424] Strategy 2: If E < 0.5, R > = 0.5, it enters a hard encirclement, the prey does not have enough energy and is captured directly;
[0425] Strategy 3: If E>=0.5, R<0.5, E is decreasing rapidly, enter a fast dive soft encirclement, and the prey has the energy to escape. Use the escape energy factor formula to flexibly update the position and implement the encirclement strategy;
[0426] Strategy 4: If E<0.5, R<0.5, enter a fast dive hard encirclement. The prey does not have enough energy, and the prey is captured directly by continuously narrowing the encirclement.
[0427] Optionally, the escape energy factor formula is expressed as:
[0428]
[0429] In the formula, E0 represents the initial escape energy, t0 represents the initial value of energy in the range of (-1,1), and T m represents the maximum number of iterations, and st represents the number of times the prey recovers its physical strength during the escape process.
[0430] Optionally, if the escape energy factor E>1 in the position determination unit, the normal search phase is entered to analyze the change speed of the current escape energy factor E, including:
[0431] If the change rate of the current escape energy factor E is greater than the set threshold, the original position update formula is used to update the individual position; otherwise, the forced search phase is entered, and the search space is dynamically adjusted according to the existing position update formula and the position of the search agent is updated.
[0432] A protection component virtual resource scheduling system based on policy selection provided by the present invention further includes: a request monitoring module;
[0433] The request module is used to monitor resource or service adjustment requests proposed by the upper application layer.
[0434] The virtual resource scheduling system involves monitoring components, management components, global scheduling components and user management components. In the global scheduling module, dynamic scheduling is adopted to monitor the entire life cycle of resources in the virtual environment, give priority to the scheduling of important resources, and do not consider the changes in resource supply and demand during the scheduling cycle in depth. The process of managing virtual resources is relatively complicated. The present invention realizes the all-round management of virtual resources from multiple modules such as virtual resource scheduling, resource integration, and underlying resource equipment positioning, improves the overall resource utilization rate, and effectively reduces the energy consumption of underlying resources.
[0435] Embodiment 4:
[0436] Based on the same inventive concept, the present invention also provides a computer device, which includes a processor and a memory, wherein the memory is used to store a computer program, the computer program includes program instructions, and the processor is used to execute the program instructions stored in the computer storage medium. The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. It is the computing core and control core of the terminal, which is suitable for implementing one or more instructions, and is specifically suitable for loading and executing one or more instructions in the computer storage medium to implement the corresponding method flow or corresponding function, so as to implement the steps of a virtual resource scheduling method of a protection component based on policy selection in the above embodiment.
[0437] Embodiment 5:
[0438] Based on the same inventive concept, the present invention also provides a storage medium, specifically a computer-readable storage medium (Memory), which is a memory device in a computer device for storing programs and data. It can be understood that the computer-readable storage medium here can include both built-in storage media in a computer device and, of course, extended storage media supported by the computer device. The computer-readable storage medium provides a storage space, which stores the operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by a processor are also stored in the storage space, and these instructions can be one or more computer programs (including program codes). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory or a non-volatile memory, such as at least one disk memory. The processor can load and execute one or more instructions stored in the computer-readable storage medium to implement the steps of a virtual resource scheduling method for a protection component based on policy selection in the above embodiment.
[0439] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0440] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0441] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0442] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.
[0443] The above are merely embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention are included in the scope of the claims of the present invention to be approved.
Claims
1. A virtual resource scheduling method for protection components based on policy selection, characterized in that: include: Determine whether to trigger the resource scheduling policy based on the demand changes of the upper application layer and the current resource pool performance status of the virtualization layer; If the resource scheduling policy is triggered, virtual resources are scheduled based on the overall survivability of the resource pool performance, the rationality of the underlying device settings, or the resource load balancing. Otherwise, virtual resources are not scheduled.
2. The method according to claim 1, characterized in that The scheduling of virtual resources from the perspective of overall survivability of resource pool performance or rationality of underlying device settings or resource load balancing includes: From the perspective of the overall survivability of the resource pool performance, the allocation of virtual resources is determined by increasing or decreasing security device nodes based on the changes in the upper application layer's demand for virtual resources and the pre-built target constraint model of the resource pool management node consumption cost; From the perspective of the rationality of the underlying device settings, based on the historical data of virtualized device load demand, the amount of reserved resources, the pre-built random integer programming model, and the pre-built weighted combination prediction model, the number of reservable virtual resources and abstracted underlying device models is determined to allocate virtual resources; In terms of resource load balancing, the load balancing strategy of the improved Harris Eagle algorithm is used to migrate virtualized resources; Among them, the random integer programming model is constructed with the objective function and constraints constructed by minimizing the virtualization cost and the maintenance cost suffered due to the attack; The weighted combination forecasting model is constructed by using the improved double exponential smoothing method for basic forecasting and the incentive backup forecasting based on the time series forecasting method based on weighted moving average.
3. The method according to claim 2, characterized in that From the aspect of the overall survivability of the resource pool performance, the increase or decrease of security device nodes is determined based on the change in the demand for virtual resources by the upper application layer combined with the pre-built target constraint model of the resource pool management node consumption cost, and the allocation of virtual resources includes: When the demand for the upper application layer increases, the security device node with the lowest consumption cost is determined through the pre-built management node minimum consumption model, and the security device node with the lowest consumption cost is added to expand the virtual resources; When the demand of the upper application layer decreases, the security device node with the highest consumption cost is determined through the pre-built maximum consumption model of the management node, and the security device node with the highest consumption cost under the resource pool management node is shut down; Wherein, the target constraint model of the resource pool management node consumption cost includes: a management node minimum consumption model and a management node maximum consumption model; The management node minimum consumption model and the management node maximum consumption model are both determined based on the energy consumption cost of a single security device node and the indicator utilization of the currently selected security device node.
4. The method according to claim 2, characterized in that From the aspect of the rationality of the underlying device settings, based on the historical data of the virtualized device load demand, the reserved resource amount, the pre-built random integer programming model, and the pre-built weighted combination prediction model, the number of reservable virtual resources and the abstracted underlying device model is determined, and the virtual resources are allocated, including: The demand forecast of virtualized equipment is obtained based on the historical data of virtualized equipment load demand and the pre-built weighted combination forecasting model; When the demand forecast is greater than the reserved resource amount, determining the number of underlying device models based on the reserved resource amount combined with a pre-built random integer programming model; The random integer programming model is constructed with the goal of minimizing the cost of the underlying device model and the constraints of the upper application layer service adjustment request.
5. The method according to claim 4, characterized in that The random integer programming model is shown in the following formula: in, Indicates the number of reserved safety device models. represents the 0-1 decision variable in the resource reservation phase, P represents the number of abstractable security device models, p is the sequence number of the abstractable security device model, T represents the resource adjustment phase, t represents the sequence number of the time, represents the number of decision variables assigned to the upper application layer, represents the decision variables assigned to the upper application layer, represents the number of decision variables available for virtual resource demand planning, represents the decision variables available for virtual resource demand planning, Indicates the maximum resource capacity range that the set resource pool can accept, N0 indicates the total number of decision variables, represents the decision variable number, r represents the total number of decision variables allocated to the upper application layer, and R represents the total number of decision variables in the reserved stage.
6. The method according to claim 5, characterized in that The number of underlying device models is calculated as follows: in, represents a 0-1 decision variable in the reservation phase, Indicates the number of reserved virtualized device models, and demand indicates the current virtual resource demand.
7. The method according to claim 4, characterized in that The demand forecast of the virtualized device is obtained by combining the historical data of the virtualized device load demand with the pre-built weighted combination forecasting model, including: Based on the time series of historical data of virtualized equipment load demand, the improved double exponential smoothing method is used to make basic predictions of virtualized equipment and obtain basic prediction quantities. When the burst resource allocation of virtualized devices is insufficient, based on the fit of the historical data of the virtualized device load demand, the time series prediction method of weighted moving average is used to perform incentive backup prediction and obtain the incentive backup prediction amount; The basic forecast amount and the incentive backup forecast amount are weightedly combined to obtain the demand forecast amount of the virtualized device.
8. The method according to claim 2, characterized in that From the aspect of resource load balancing, the load balancing strategy of the improved Harris Eagle algorithm is adopted to migrate virtualized resources, including: Decoding is performed based on the coded information of the communication port between the virtualized resource pool and the underlying device, and matching is performed to find the current resource pool management node corresponding to the underlying device; The sequence of virtualized devices to be migrated is determined based on CPU and memory utilization; the improved Harris Eagle optimization algorithm is used to reallocate the virtualized devices in the sequence of virtualized devices to be migrated to the resource pool management node to determine the location of the virtualized devices.
9. The method according to claim 8, characterized in that The improved Harris Hawk optimization algorithm is used to reallocate the virtualized devices in the sequence of virtualized devices to be migrated to the resource pool management node to determine the location of the virtualized devices, including: Step 1: Use the virtualized devices to be migrated in the sequence table of virtualized devices to be migrated as the Harris Hawk population and set the maximum number of iterations T m ; Step 2: Calculate the fitness of each individual based on the fitness objective function, and sort the Harris hawk population according to the fitness; Step 3: Select the individual with the best fitness as prey and update the escape energy factor E; Step 4: Determine the search strategy to be adopted according to the change trend of the escape energy factor E, and update the Harris hawk population based on the search strategy, increase the number of iterations by 1, and return to step 3; Step 5: Determine whether the number of iterations reaches the maximum number of iterations T m , if yes, go to step 6, otherwise, return to step 4; Step 6: Determine the virtual device location by sorting the virtual devices in the latest Harris Hawk population; Wherein, the fitness function is determined by the idea of Pareto optimality.
10. The method according to claim 9, characterized in that The search strategy adopted according to the change trend of the escape energy factor E includes: If the escape energy factor E>1, the normal search phase is entered to analyze the change speed of the current escape energy factor E; If E < 1, it enters the local arrest stage, and the comprehensive escape energy factor E and the convergence trend of individual fitness determine the prey's update position.
11. The method according to claim 10, characterized in that If E < 1, the local arrest phase is entered, and the prey update position is determined by the comprehensive escape energy factor E and the individual fitness convergence trend, including: Strategy 1: If E>=0.5, R>=0.5, enter soft encirclement, the prey has physical strength but is eventually captured; Strategy 2: If E < 0.5, R > = 0.5, it enters a hard encirclement, the prey does not have enough energy and is captured directly; Strategy 3: If E>=0.5, R<0.5, E is decreasing rapidly, enter a fast dive soft encirclement, and the prey has the energy to escape. Use the escape energy factor formula to flexibly update the position and implement the encirclement strategy; Strategy 4: If E<0.5, R<0.5, enter a fast dive hard encirclement. The prey does not have enough energy, and the prey is captured directly by continuously narrowing the encirclement.
12. The method according to claim 10, characterized in that If the escape energy factor E>1, the normal search phase is entered to analyze the change speed of the current escape energy factor E, including: If the change rate of the current escape energy factor E is greater than the set threshold, the original position update formula is used to update the individual position; otherwise, the forced search phase is entered, and the search space is dynamically adjusted according to the existing position update formula and the position of the search agent is updated.
13. The method according to claim 11, characterized in that The escape energy factor formula is expressed as: In the formula, E0 represents the initial escape energy, t0 represents the initial value of energy in the range of (-1,1), and T m represents the maximum number of iterations, and st represents the number of times the prey recovers its physical strength during the escape process.
14. A virtual resource scheduling system for protection components based on policy selection, characterized in that: include: A judgment module is used to determine whether to trigger a resource scheduling strategy based on the demand changes of the upper application layer and the performance status of the resource pool of the virtualization layer; The resource scheduling module is used to schedule virtual resources from the perspective of overall survivability of resource pool performance, rationality of underlying device settings, or resource load balancing if the resource scheduling policy is triggered. Otherwise, virtual resources will not be scheduled.
15. The system of claim 14, wherein: The resource scheduling module includes: The first resource scheduling submodule is used to determine the increase or decrease of security device nodes and the allocation of virtual resources from the perspective of the overall survivability of the resource pool performance, based on the changes in the demand for virtual resources by the upper application layer and the target constraint model of the consumption cost of the resource pool management node built in advance; The second resource scheduling submodule is used to determine the number of reservable virtual resources and abstract underlying device models from the perspective of rationality of underlying devices, based on historical data of virtualized device load demand, reserved resource quantity, pre-built random integer programming model, and pre-built weighted combination prediction model, and allocate virtual resources; The third resource scheduling submodule is used to migrate virtualized resources from the perspective of resource load balancing by adopting a load balancing strategy of an improved Harris Hawk algorithm.
16. The system of claim 15, wherein: The first resource scheduling submodule includes: The demand increase adjustment unit is used to determine the security device node with the lowest consumption cost through the pre-built management node minimum consumption model when the demand of the upper application layer increases, and to increase the security device node with the lowest consumption cost to expand the virtual resources; The demand reduction adjustment unit is used to determine the security device node with the highest consumption cost through the pre-built maximum consumption model of the management node when the demand of the upper application layer decreases, and shut down the security device node with the highest consumption cost under the resource pool management node; Wherein, the target constraint model of the resource pool management node consumption cost includes: a management node minimum consumption model and a management node maximum consumption model; The management node minimum consumption model and the management node maximum consumption model are both determined based on the energy consumption cost of a single security device node and the indicator utilization of the currently selected security device node.
17. The system of claim 15, wherein: The second resource scheduling submodule is specifically used for: The demand forecast of virtualized equipment is obtained based on the historical data of virtualized equipment load demand and the pre-built weighted combination forecasting model; When the demand forecast is greater than the reserved resource amount, determining the number of underlying device models based on the reserved resource amount combined with a pre-built random integer programming model; The random integer programming model is constructed with the goal of minimizing the cost of the underlying device model and the constraints of the upper application layer service adjustment request.
18. A computer device, characterized in that: include: at least one processor and memory; The memory is used to store one or more programs; When the one or more programs are executed by the at least one processor, a protection component virtual resource scheduling method based on policy selection as described in any one of claims 1 to 13 is implemented.
19. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed, a protection component virtual resource scheduling method based on policy selection as described in any one of claims 1 to 13 is implemented.
Citation Information
Cited By
Hybrid key task scheduling method, system and equipment based on virtualization technology
CN120560861A
