A novel multi-time scale fine-grained parallel simulation method and system for distribution network

Abstract

The present invention discloses a novel multi-time-scale fine-grained parallel simulation method and system for distribution network, which belongs to the field of intelligent distribution network. Firstly, a computing resource sharing method based on segmentation cost is proposed, and a resource sharing segmentation cost mechanism for multi-time-scale simulation is constructed. The priority of the business, the complexity of the business simulation, and the resource sharing value are further considered to transfer computing resources. Secondly, a multi-time-scale fine-grained parallel simulation resource sharing segmentation cost adjustment method based on the combination of soft and hard is proposed. Considering the situation that resource-abundant containers cannot meet their own business due to excessive resource sharing, hard adjustment means are used for containers of high-priority business, and soft adjustment means are used for containers of low-priority business to ensure the normal operation of the container's own business.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent distribution networks, and in particular relates to a novel multi-time scale fine-grained parallel simulation method and system for distribution networks. Background Art

[0002] With the rapid development of energy Internet technology, the intelligence and information level of new distribution networks are constantly improving. At the same time, due to the large-scale access of new energy sources such as photovoltaics, wind power, and electric vehicles, the massive business data generated has also posed new challenges to the smooth development of various power businesses such as marketing, control, and dispatch. A simulation system is a system that uses computing resources to simulate, simulate, and analyze the behavior of complex systems. Parallel simulation is a technology designed for large-scale simulation analysis. It can decompose a simulation business into a computer system containing multiple containers for parallel execution, providing effective guidance for the safe operation of new distribution networks. Therefore, it is urgent to design a new distribution network parallel simulation method and system. By sharing computing resources between containers, the simulation performance and computing resource utilization of the new distribution network parallel simulation can be improved, and support can be provided for fine-grained parallel simulation of the new distribution network, so as to achieve observable simulation time scales of various power businesses.

[0003] The existing parallel simulation methods for distribution networks still have the following two shortcomings: First, the existing technology ignores the segmented cost of computing resource sharing, and does not consider the sharing of computing resources among containers when the business simulation time scale cannot meet the requirements, resulting in the difficulty in meeting the simulation requirements of different businesses at different time scales. Second, it does not consider the division of priorities for different businesses, and cannot adopt different cost adjustment methods according to the level of business priority, resulting in too low efficiency in updating the segmented cost of resource sharing, poor effect of resource sharing red line adjustment, and poor flexibility of business simulation. Therefore, it is urgent to design a new multi-time scale fine-grained parallel simulation method and system for distribution networks. Summary of the invention

[0004] In order to solve the technical problems existing in the background technology, the present invention aims to provide a new multi-time scale fine-grained parallel simulation method and system for distribution network, construct a resource sharing segmentation cost mechanism for multi-time scale simulation, and adopt a multi-time scale fine-grained parallel simulation resource sharing segmentation cost adjustment method combining software and hardware to meet the simulation requirements of different businesses at different time scales, and effectively improve the multi-time scale fine-grained parallel simulation efficiency and computing resource utilization of distribution network.

[0005] In order to solve the technical problem, the technical solution of the present invention is:

[0006] A novel multi-time scale fine-grained parallel simulation method for distribution network, the method comprising:

[0007] S1: Divide the business priorities, complexity and time scale of business simulation, consider the different containers of each computer in the computer cluster, and record the computing resources;

[0008] S2: Set the segmentation threshold and resource sharing red line, and divide the computing resources that can be shared by the container into three segments according to the set values, and obtain the unit cost and total cost function of each segment of computing resources;

[0009] S3: Classify and sort the containers according to available computing resources, business priorities, and required additional computing resources to obtain a set of containers with abundant resources and containers with scarce resources. Take out containers from the set of containers with scarce resources, and at the same time, take out containers with abundant resources in turn to match them, and calculate the computing resource transfer amount that meets the conditions.

[0010] S4: Based on the matching results and the amount of resource transfer, determine whether the containers with sufficient resources can meet their own business needs after all container resources are transferred. Based on the judgment results, if all containers with sufficient resources can meet their own needs, the matching results and the amount of resource transfer will be transferred. Otherwise, cost adjustment will be performed. Hard adjustment measures will be used for containers with high-priority services, and soft adjustment measures will be used for containers with low-priority services. Re-matching will be performed after adjustment.

[0011] Furthermore, in step S1, the total number of distribution network computer cluster containers I and the computing resources available for different containers are counted. Business information division module prioritizes business Complexity of business simulation and time scale The business simulation time scale and computing resources are mapped and inversely mapped according to the historical database, and the computing resources R required to achieve the business simulation of the desired time scale are calculated. i ′.

[0012] Further, the step S1 specifically includes:

[0013] Determine the total number of computer cluster containers to be I, and the i-th container has a total computing resource Each container carries the corresponding business. Containers with sufficient resources can share computing resources with other containers. The shared resources of container i do not exceed the resource sharing red line V. i Red , according to the importance of the business, the container main business is divided into L levels according to priority, and the priority level of container i business is obtained The larger the value, the higher the service priority; considering the simulation complexity of container i service The larger the value, the greater the complexity of the simulation; considering the simulation time scale requirements of container i business Covering the full time scale physical process from steady-state power flow to nanosecond electromagnetic transient state, The larger the value, the larger the time scale of the simulation;

[0014] Define mappings based on historical database Input is the simulation time scale of container i service and simulation complexity Output is the computing resources required for container i business simulation The smaller the simulation time scale and the higher the simulation complexity, the more computing resources are required; at the same time, the mapping is defined The input is the computing resources R and simulation complexity that can be provided for the business at present. The output is the business simulation time scale. The more computing resources available and the lower the simulation complexity, the smaller the achievable simulation time scale.

[0015] Further, the step S2 specifically includes:

[0016] Segment threshold V i TS1 、V i TS2 And the resource sharing red line V i Red The initial settings are as follows:

[0017]

[0018] in, Calculate the total amount of resources for container i; is the priority of the service of container i; L is the maximum service priority; Simulate the complexity of container i business, is its minimum and maximum value; is the time scale requirement for container i business simulation, is its minimum and maximum value;

[0019] According to the segmented threshold and resource sharing red line, container i shares resources based on three threshold ranges; for the first range 0<x≤V i TS1 , that is, less than the threshold V i TS1 The shared resources in this section have almost no impact on the simulation time scale of the business itself, so a fixed unit cost is used; for the second section V i TS1 <x≤V i TS2 That is, it exceeds the threshold V i TS1 But not greater than V i TS2Part of the shared resources, the computing resource sharing in this section has an impact on the simulation time scale of its own business, so a linearly increasing unit cost is adopted, and the threshold value V is exceeded. i TS1 The more, the higher the unit cost; for the third segment V i TS2 <x≤V i Red That is, it exceeds the threshold V i TS2 But not greater than the resource sharing red line V i Red The shared resources of this part have a great impact on the simulation time scale of its own business, so the unit cost of exponential growth is adopted, and the cost exceeds the threshold V i TS2 The more, the higher the unit cost; finally, we get the segmented cost function C i (x) represents the unit cost of container i sharing each segment of computing resources with other containers, which is calculated as follows:

[0020]

[0021] Among them, μ i V represents the basic unit cost of processing services by container i; i Red V represents the resource sharing red line; i TS1 、V i TS2 is the segmentation threshold; the unit cost of the second segment is (μ i ,2μ i ]; the unit cost of the third segment is between (μ i e,μ i e 2 ]; according to the resource sharing segmentation cost function C i (x) and the resource R that container i actually shares with container j i,j , we can get the total cost function of container i actually sharing computing resources with container j The calculation is as follows:

[0022]

[0023] The abundant container i shares computing resources with container j to meet the time scale requirement of container j business simulation, bringing higher business value. Considering the value improvement brought by the time scale improvement, the participation of abundant container i in container j simulation can bring benefits E. i,j , calculated as follows:

[0024]

[0025] in, is the total computing resources of container j, For the simulation complexity of container j business, bring in the mapping Get the simulation time scale that container j can achieve; R i,j is the computing resources shared by container i to container j, The simulation time scale of container j business is considered when container i shares computing resources with container j; is the time scale revenue function of container j’s business, They represent the simulation time scales that can be achieved before and after resource sharing, respectively. The time scales before and after sharing are input into the time scale benefit function, and the value brought by the improvement of the simulation time scale of business j is output;

[0026] Containers sharing computing resources must generate additional value, that is, the value generated by container resource sharing must exceed the cost of resource sharing. Consider how much value can be generated for other containers when 1 CPU cycle of computing resources is transferred from a container with abundant resources to other containers. That is, the value of computing resource sharing is greater than or equal to the cost α i times before resources can be transferred; resource sharing value coefficient α i It is related to the service priority and simulation scale of the resource-rich container. The higher the priority, the smaller the simulation scale. i The larger the value, the calculation is as follows:

[0027]

[0028] in, Indicates rounding up; is the simulation scale; is the maximum value of the simulation scale; is the business priority. Further, the step S3 specifically includes:

[0029] S31: Classify I containers; consider That is, all containers with sufficient computing resources are allocated according to the available computing resources. Sort in descending order. When available computing resources are equal, sort by business priority p i PL Sort in ascending order and obtain a set U containing N containers with sufficient resources = {u1,u2,…,u N},u n The container index representing the nth element in the collection, where the container with abundant computing resources and low business priority has a small index; consider That is, all containers whose original container computing resources cannot meet the time scale simulation requirements are allocated according to business priorities. Sort in descending order, with the same priority level based on the additional computing resources required Sort in descending order and get a set W containing M resource-scarce containers = {w1,w2,…,w M},w m The container index of the mth element in the collection, where the container with high business priority and more additional computing resources required has a small index; That is, containers whose computing resources of the original container can just meet their own time scale simulation requirements do not participate in the computing resource transfer;

[0030] S32: Initialization, set m=n=1;

[0031] S33: Take out containers w from the set W in sequence m , for container w m , take out container u from set U in sequence n Match so that container u n Satisfy container w as much as possible m business needs, proceed to step S34;

[0032] S34: At this time, according to the container u n The resource sharing segment total cost function And the container w m Required Resources Calculate the total cost of resource transfer Calculate the income at the same time Resource sharing value coefficient In satisfying Under the premise of , as much computing resources as possible are transferred, the containers are matched and the resource transfer amount is Let m=m+1, n=n+1 and return to step S33 to continue matching the remaining containers; if m>M or n>N, it means that all containers are matched, and proceed to step S35;

[0033] S35: After considering the transfer of all container resources, determine whether the container with sufficient resources can meet its own business needs. If not, adjust the cost.

[0034] Furthermore, in step S4, the high priority service cost adjustment method based on hard adjustment specifically includes:

[0035] If the service carried by container i with sufficient resources is a high-priority service and the time scale requirement of its own service simulation cannot be met, hard adjustment is used to adjust the resource sharing segment cost, that is, to reduce the resource sharing red line V i Red The higher the priority and the greater the gap between the current simulation time scale and the scale requirement, the greater the reduction of the resource sharing red line, which can better guarantee the computing resources of its own container and ensure the normal operation of high-priority services. The adjustment calculation of the resource sharing red line is as follows:

[0036]

[0037] in, Indicates the service priority; L is the maximum priority; Simulation time scale requirement for container i; Indicates that container resource i shares computing resource R with container j i,j The remaining computing resources, is the simulation complexity of the container i service, and is brought into the mapping G i (R,p SC ) obtain the simulation time scale that can be achieved by the container j's own business after sharing computing resources; is the scale gap function, which represents the simulation time scale that can be satisfied by the current resources and simulation time scale requirements The larger the gap, the larger the value, and the time scale cannot meet the requirements. therefore

[0038] Furthermore, in step S4, the low priority service cost adjustment method based on soft adjustment specifically includes:

[0039] If the service carried by container i with sufficient resources is a low-priority service and the time scale requirement of its own service simulation cannot be met, the resource sharing segmentation cost is adjusted by soft adjustment, that is, the segmentation threshold V is reduced. i TS1 、V i TS2 , increase the basic unit cost μ i The higher the container priority and the greater the gap between the current simulation time scale and the scale requirement, the greater the increase in the basic unit cost and the greater the decrease in the segmentation threshold, making resource sharing more difficult. It is more likely to ensure the normal operation of its own business or sacrifice the simulation time scale of its own business to obtain higher global benefits. The basic unit cost and segmentation threshold are adjusted and calculated as follows:

[0040]

[0041] Among them, μ i represents the basic unit cost; Indicates the service priority; L is the maximum priority; The meaning is the same as (7), indicating the gap between the simulation time scale that the current resources can meet and the simulation time scale requirement. The larger the gap, the larger the value.

[0042]

[0043] A novel multi-time-scale fine-grained parallel simulation system for distribution network, the system is applied to any of the above methods, the system comprises: a business information division module, a business simulation required resource calculation module, a container information recording module, a segmented cost mechanism construction module, a container classification module, a matching module, a business demand judgment module, a cost adjustment module and a resource transfer module;

[0044] The business information division module: divides the business priority, business simulation complexity and time scale;

[0045] The business simulation required resource calculation module: constructs mapping and inverse mapping according to the historical data set, and calculates the computing resources required to achieve the business simulation of the desired time scale;

[0046] The container information recording module records the total number of distribution network computer cluster containers and the computing resources available for different containers, and uploads them to the segmented cost mechanism construction module;

[0047] The segmentation cost mechanism construction module: sets segmentation thresholds and resource sharing red lines, and divides the computing resources that can be shared by the container into three segments according to the set values, calculates the unit cost of each segment of computing resources, that is, the segmentation cost function, and uploads it to the container classification module;

[0048] The container classification module: classifies and sorts the containers according to the available computing resources, business priorities and required additional computing resources, obtains a collection of resource-rich containers and resource-scarce containers, and uploads them to the matching module;

[0049] The matching module: takes out containers from the resource-scarce container set, and at the same time, sequentially takes out containers with abundant resources to match them, and calculates the qualified computing resource transfer amount, so that the transferred computing resources can meet the needs of the resource-scarce container as much as possible, and uploads the matching results and resource transfer amount to the business demand judgment module;

[0050] The business demand judgment module: based on the matching results and the resource transfer amount, considers whether the resource-rich container can meet its own business needs after all container resources are transferred, and uploads the demand judgment results, matching results and resource transfer amount to the cost adjustment module;

[0051] The cost adjustment module: according to the judgment result of the business demand judgment module, if the containers with sufficient resources can meet their own needs, the matching result and resource transfer amount are uploaded to the resource transfer module; otherwise, the cost adjustment is performed. The cost adjustment module adopts hard adjustment means for containers of high-priority services; and soft adjustment means for containers of low-priority services, and returns them to the matching module after adjustment;

[0052] The resource transfer module is used to transfer computing resources according to the matching results and the resource transfer amount.

[0053] Compared with the prior art, the advantages of the present invention are:

[0054] (1) The present invention proposes a computing resource sharing method based on segmentation cost. First, the priority of various services, the complexity of business simulation, and the time scale are divided. At the same time, the different containers of each computer in the computer cluster are considered and their computing resources are recorded; secondly, according to the time scale requirements, complexity, and priority of each service in the simulation system, the segmentation threshold and resource sharing red line are set to obtain the resource sharing unit cost and construct a resource sharing segmentation cost mechanism for multi-time scale simulation; finally, according to factors such as business priority, resource-abundant containers and resource-scarce containers are divided, and computing resources are transferred based on the segmentation threshold and resource sharing unit cost to meet the simulation requirements of different services at different time scales and improve the simulation performance of parallel simulation of new distribution networks.

[0055] (2) The present invention proposes a multi-time-scale fine-grained parallel simulation resource sharing segmentation cost adjustment method based on a combination of soft and hard. First, based on the priority of each business, the simulation time scale of each business is analyzed, and high-priority and low-priority businesses are divided. When the fine-grained simulation time scale demand of the high-priority business cannot be met, a hard adjustment method is adopted, that is, the resource sharing red line is adjusted to directly reduce the amount of external computing resources shared; secondly, when the simulation time scale demand of the low-priority business cannot be met, a soft adjustment method is adopted to update the resource sharing segmentation cost, that is, according to the simulation time scale gap, simulation complexity and other information, the basic unit cost is increased and the segmentation threshold is reduced, so that the resource sharing cost is increased, and the amount of external computing resources shared is indirectly reduced, which effectively improves the efficiency of the multi-time-scale fine-grained parallel simulation of the distribution network and the utilization rate of computing resources, and enhances the flexibility of the new distribution network parallel simulation system in the face of different business simulations. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 , a flow chart of a novel multi-time-scale fine-grained parallel simulation method for distribution network of the present invention;

[0057] Figure 2 , a block diagram of a novel multi-time-scale fine-grained parallel simulation system for distribution networks of the present invention. DETAILED DESCRIPTION

[0058] The specific implementation mode of the present invention is described below in conjunction with embodiments:

[0059] It should be noted that the structures, proportions, sizes, etc. shown in this specification are only used to match the contents disclosed in the specification so that people familiar with this technology can understand and read them, and are not used to limit the conditions under which the present invention can be implemented. Any structural modification, change in proportional relationship or adjustment of size should still fall within the scope of the technical content disclosed in the present invention without affecting the effects and purposes that can be achieved by the present invention.

[0060] At the same time, the terms such as "upper", "lower", "left", "right", "middle" and "one" cited in this specification are only for the convenience of description and are not used to limit the scope of implementation of the present invention. Changes or adjustments to their relative relationships should be regarded as the scope of implementation of the present invention without substantially changing the technical content.

[0061] Embodiment 1:

[0062] (1) The existing technology ignores the segmentation cost of computing resource sharing and does not consider sharing computing resources among containers when the business simulation time scale cannot meet the requirements, resulting in difficulty in meeting the simulation requirements of different businesses at different time scales.

[0063] The present invention proposes a computing resource sharing method based on segmentation cost. Firstly, the priority of various services, the complexity of business simulation, and the time scale are divided. Meanwhile, different containers of each computer in a computer cluster are considered and their computing resources are recorded. Secondly, according to the time scale requirements, complexity, and priority of each service in the simulation system, a resource sharing segmentation cost mechanism for multi-time scale simulation is constructed, including a segmentation threshold, a unit cost of resource sharing in each segment, and a resource sharing red line. Finally, based on the segmentation threshold and the unit cost of resource sharing, computing resources are transferred to meet the simulation requirements of different services at different time scales and improve the simulation performance of parallel simulation of new distribution networks.

[0064] (2) The existing technology does not consider the priority classification of different businesses and cannot adopt different cost adjustment methods according to the level of business priority, resulting in poor flexibility of the new distribution network parallel simulation system in the face of different business simulations.

[0065] The present invention proposes a multi-time-scale fine-grained parallel simulation resource sharing segmentation cost adjustment method based on a combination of soft and hard. First, the simulation time scale of each business is analyzed based on the priority of each business. When the fine-grained simulation time scale demand of a business above a certain priority cannot be met, a hard adjustment method is adopted, that is, adjusting the resource sharing red line. Secondly, when the simulation time scale demand of a low-priority business cannot be met, a soft adjustment method is adopted to update the resource sharing segmentation cost, that is, to adjust the segmentation threshold and resource sharing unit cost according to the simulation time scale gap, simulation complexity and other information, effectively improving the multi-time-scale fine-grained parallel simulation efficiency and computing resource utilization of the distribution network, and enhancing the flexibility of the new distribution network parallel simulation system in the face of different business simulations.

[0066] Embodiment 2:

[0067] The present invention proposes a novel multi-time-scale fine-grained parallel simulation method for distribution networks; first, a computing resource sharing method based on segmentation cost is proposed, and a resource sharing segmentation cost mechanism for multi-time-scale simulation is constructed, and the priority of the business, the complexity of the business simulation, and the resource sharing value are further considered to transfer computing resources. Secondly, a multi-time-scale fine-grained parallel simulation resource sharing segmentation cost adjustment method based on a combination of soft and hard is proposed. Considering the situation that resource-abundant containers cannot meet their own business due to excessive resource sharing, hard adjustment means are used for containers of high-priority business, and soft adjustment means are used for containers of low-priority business to ensure the normal operation of the container's own business.

[0068] The method steps of the present invention are as follows Figure 1 shown.

[0069] Step 1: Count the total number of distribution network computer cluster containers I and the computing resources available for different containers Business information division module prioritizes business Complexity of business simulation and time scale to divide;

[0070] Step 2: Define the mapping and inverse mapping between business simulation time scale and computing resources based on the historical database, and calculate the computing resources R required to achieve the business simulation of the desired time scale i ′;

[0071] Step 3: Construct a resource sharing segmentation cost mechanism for multi-time scale simulation and set the segmentation threshold V i TS1 、V i TS2 And the resource sharing red line V i Red, segment the computing resources that can be shared, and obtain the unit cost C of sharing each segment of computing resources i (x) and the total cost function for container i to actually share computing resources with container j

[0072] Step 4: Classify the containers and sort them according to available computing resources, business priorities, and required additional computing resources to obtain a set of containers with sufficient resources U = {u1,u2,…,u N} and the set of resource-scarce containers W = {w1,w2,…,w M}.

[0073] Step 5: Initialize, set m=n=1.

[0074] Step 6: Take out containers w from the set W one by one m , for container w m , take out container u from set U in sequence n Make a match;

[0075] Step 7: Now according to the container u n The resource sharing segment total cost function And the container w m Required Resources Calculate the total cost of resource transfer Calculate the income at the same time Resource sharing value coefficient In satisfying Under the premise of , as much computing resources as possible are transferred, the containers are matched and the resource transfer amount is Set m=m+1, n=n+1 and return to step 6 to continue matching the remaining containers. If m>M or n>N, it means all containers match, and go to step 8.

[0076] Step 8: After considering all container resource transfers, determine whether the resource-rich container can meet its own business needs. If not, adjust the cost to use container u n For example: If the resource is sufficient, container u n Business Priorities Perform hard adjustment; otherwise, perform soft adjustment. After cost adjustment, return to step 5 and re-match after initialization. If all containers with sufficient resources can meet their own business needs, transfer computing resources according to the matching results.

[0077] Specifically, the present invention provides a novel multi-time scale fine-grained parallel simulation method for distribution network, comprising:

[0078] S1: Computing resource sharing method based on segmentation cost

[0079] The present invention proposes a computing resource sharing method based on segmentation cost. First, the total number of distribution network computer cluster containers and the total amount of computing resources of different containers are counted, the priority of container services, the complexity of business simulation and the time scale are divided, and the historical database is called to obtain the computing resources required to implement the business simulation of the required time scale; secondly, a resource sharing segmentation cost mechanism for multi-time scale simulation is constructed according to the characteristics of each business simulation, including segmentation thresholds and resource sharing unit costs and resource sharing red lines in each segment; finally, computing resources are transferred based on segmentation thresholds and resource sharing unit costs. The specific introduction is as follows:

[0080] S1.1: The computing process of each business in the new distribution network can be broken down into multiple micro-applications, which flow in parallel between different containers. The time scale represents the time required to complete the simulation task, which may be in the order of nanoseconds, microseconds, or milliseconds. A smaller simulation time scale usually requires more computing resources to simulate, because a smaller time scale means that the model needs to update the state and calculation results more frequently, which places higher demands on the computer's processing power and memory capacity. Carry out fine-grained parallel simulation of the new distribution network and observe whether each business can meet the simulation time scale requirements. When the requirements cannot be met, it is necessary to share computing resources between containers, that is, containers whose simulation time scale requirements cannot be met can request computing resource sharing from containers with abundant resources.

[0081] Assume that the total number of computer cluster containers is I, and the i-th container has a total computing resource of Each container carries the corresponding business. Containers with sufficient resources can share computing resources with other containers. However, to ensure their own business needs, container i’s shared resources cannot exceed the resource sharing red line V. i Red According to the importance of the business, the main business of the container is divided into L levels according to the priority, and the priority level of the container i business is obtained The larger the value, the higher the service priority; considering the simulation complexity of container i service The larger the value, the greater the complexity of the simulation; considering the simulation time scale requirements of container i business Covering the full time scale physical process from steady-state power flow to nanosecond electromagnetic transient state, The larger the value, the larger the simulation time scale.

[0082] Define mappings based on historical database Input is the simulation time scale of container i service and simulation complexity Output is the computing resources required for container i business simulation The smaller the simulation time scale and the higher the simulation complexity, the more computing resources are required; at the same time, the mapping is defined The input is the computing resources R and simulation complexity that can be provided for the business at present. The output is the business simulation time scale. The more computing resources available and the lower the simulation complexity, the smaller the achievable simulation time scale.

[0083] S1.2: Constructing a resource sharing segmentation cost mechanism for multi-timescale simulation

[0084] The present invention considers the segmentation threshold, the unit cost of resource sharing in each segment, and the resource sharing red line to construct a resource sharing segmentation cost mechanism for multi-time scale simulation. The unit cost refers to the cost of transferring the computing resources of one CPU cycle from a container to another container. The computing resources available for sharing in the i-th container are divided into three segments, and the resource sharing red line is set according to the segmentation threshold V. i TS1 、V i TS2 And the resource sharing red line V i Red The higher the simulation complexity of the container, the smaller the simulation time scale, and the higher the service priority, the more computing resources the container needs and the more it must ensure the normal operation of its own service. Therefore, the fewer computing resources it can provide to other containers, the smaller the resource sharing red line. At the same time, the higher the priority of the service, the lower its segmentation threshold, which makes the cost of sharing the same resources higher, and it can better ensure the operation of its own service and improve the overall revenue. Segmentation threshold V i TS1 、V i TS2 And the resource sharing red line V i Red The initial settings are as follows:

[0085]

[0086] in, Calculate the total amount of resources for container i; is the priority of the service of container i; L is the maximum service priority; Simulate the complexity of container i business, is its minimum and maximum value; is the time scale requirement for container i business simulation, is its minimum and maximum value.

[0087] According to the segmented threshold and resource sharing red line, container i shares resources based on the three-segment threshold range. i TS1 , which is less than the threshold V i TS1The shared resources in this section have almost no impact on the simulation time scale of the business itself, so a fixed unit cost is used; for the second section V i TS1 <x≤V i TS2 That is, it exceeds the threshold V i TS1 But not greater than V i TS2 Part of the shared resources, the computing resource sharing in this section has some impact on the simulation time scale of its own business, so a linearly increasing unit cost is adopted, and the threshold value V is exceeded. i TS1 The more, the higher the unit cost; for the third segment V i TS2 <x≤V i Red That is, it exceeds the threshold V i TS2 But not greater than the resource sharing red line V i Red The shared resources of this part have a great impact on the simulation time scale of its own business, so the unit cost of exponential growth is adopted, and the cost exceeds the threshold V i TS2 The more, the higher the unit cost. Finally, we get the segmented cost function C i (x) represents the unit cost of container i sharing each segment of computing resources with other containers, which is calculated as follows:

[0088]

[0089] Among them, μ i V represents the basic unit cost of processing services by container i; i Red V represents the resource sharing red line; i TS1 、V i TS2 is the segmentation threshold; the unit cost of the second segment is (μ i ,2μ i ]; the unit cost of the third segment is between (μ i e,μ i e 2 ] According to the resource sharing segmentation cost function C i (x) and the resource R that container i actually shares with container j i,j , we can get the total cost function of container i actually sharing computing resources with container j The calculation is as follows:

[0090]

[0091] The abundant container i shares computing resources with container j to meet the time scale requirements of container j business simulation, bringing higher business value. Considering the value improvement brought by the time scale improvement, the participation of abundant container i in container j simulation can bring benefits E i,j , calculated as follows:

[0092]

[0093] in, is the total computing resources of container j, For the simulation complexity of container j business, bring in the mapping Get the simulation time scale that container j can achieve; R i,j is the computing resources shared by container i to container j, The simulation time scale of container j business is considered when container i shares computing resources with container j; is the time scale revenue function of container j’s business, They represent the simulation time scales that can be achieved before and after resource sharing respectively. The time scales before and after sharing are input into the time scale benefit function, and the value brought by the improvement of the simulation time scale of business j is output.

[0094] Containers sharing computing resources must generate additional value, that is, the value generated by container resource sharing must exceed the cost of resource sharing. Consider how much value can be generated for other containers when 1 CPU cycle of computing resources is transferred from a container with abundant resources to other containers. That is, the value of computing resource sharing is greater than or equal to the cost α i Resource transfer can only be carried out when the resource sharing value coefficient α i It is related to the service priority and simulation scale of the resource-rich container. The higher the priority, the smaller the simulation scale. i The larger the value, the calculation is as follows:

[0095]

[0096] in, Indicates rounding up; is the simulation scale; is the maximum value of the simulation scale; For business priority.

[0097] S1.3: Computational resource transfer based on segmentation cost mechanism

[0098] The present invention transfers computing resources based on a segmented cost mechanism. Containers are divided into resource-rich containers and resource-scarce containers, and computing resources are transferred from resource-rich containers to resource-scarce containers. The specific steps of computing resource transfer based on the segmented cost mechanism are as follows:

[0099] (1) Classify I containers. Consider That is, all containers with sufficient computing resources are allocated according to the available computing resources. Sort in descending order, and by business priority when available computing resources are equal Sort in ascending order and obtain a set U containing N containers with sufficient resources = {u1,u2,…,u N},u n The container index representing the nth element in the collection, where the container with abundant computing resources and low business priority has a small index; consider That is, all containers whose original container computing resources cannot meet the time scale simulation requirements are allocated according to business priorities. Sort in descending order, with the same priority level based on the additional computing resources required Sort in descending order and get a set W containing M resource-scarce containers = {w1,w2,…,w M},w m Indicates the container index of the mth element in the collection, where the container with high business priority and more additional computing resources required has a smaller index. That is, the computing resources of the original container can just meet the simulation requirements of its own time scale, and the container does not participate in the transfer of computing resources.

[0100] (2) Initialization, set m=n=1.

[0101] (3) Take out containers w from the set W one by one m , for container w m , take out container u from set U in sequence n Match so that container u n Satisfy container w as much as possible m According to the business needs, proceed to step (4).

[0102] (4) At this time, according to the container u n The resource sharing segment total cost function And the container w m Required Resources Calculate the total cost of resource transfer Calculate the income at the same time Resource sharing value coefficient In satisfying Under the premise of , as much computing resources as possible are transferred, the containers are matched and the resource transfer amount is Let m=m+1, n=n+1 and return to step (3) to continue matching the remaining containers. If m>M or n>N, it means all containers are matched, and proceed to step (5).

[0103] (5) After considering the transfer of all container resources, determine whether the resource-rich container can meet its own business needs. If it cannot meet the needs, adjust the cost to ensure that the container u n For example: If the resource is sufficient, container u n Business Priorities Perform hard adjustment according to the method in S2.2; otherwise, perform soft adjustment according to the method in S2.3. After cost adjustment, return to step (2) to initialize and rematch. If all containers with sufficient resources can meet their own business needs, transfer computing resources according to the matching results.

[0104] S2: A segmented cost adjustment method for resource sharing in multi-time scale fine-grained parallel simulation based on software and hardware integration

[0105] The present invention proposes a multi-time scale fine-grained parallel simulation resource sharing segmentation cost adjustment method based on a combination of software and hardware. Considering the situation that the resource-abundant container may be unable to meet its own business during the computing resource sharing process, cost adjustment is performed according to the business priority to ensure that the business of the resource-abundant container itself is carried out normally. Among them, hard adjustment means are used for containers of high-priority business, and soft adjustment means are used for containers of low-priority business. The specific introduction is as follows:

[0106] The priority of the service of container i is obtained from S1.1: assumed The container business is a high-priority business, and the rest of the business is a low-priority business. During the computing resource transfer process, the container with abundant resources shares computing resources with the container with scarce resources. However, it may happen that the container with abundant resources cannot meet its own business. When the container cannot meet its own business, different cost adjustment methods are adopted according to the priority of the container business: hard adjustment means are adopted for the container of high-priority business to reduce the resource sharing red line of the container, directly reducing the amount of computing resources shared with the outside; soft adjustment means are adopted for the container of low-priority business to increase the basic unit cost and reduce the segmentation threshold, so that the resource sharing cost increases and indirectly reduces the amount of computing resources shared with the outside. Compared with monotonous hard adjustment or soft adjustment, the combination of hard and soft adjustment means can ensure the normal operation of the container's high-priority business, and also provide a certain flexibility for resource sharing. When the resource sharing benefit of the low-priority business container is very high, the simulation time scale of its own business can be appropriately sacrificed to obtain higher global benefits.

[0107] S2.1: High priority service cost adjustment method based on hard adjustment

[0108] If the service carried by container i with sufficient resources is a high-priority service and the time scale requirement of its own service simulation cannot be met, hard adjustment is used to adjust the resource sharing segment cost, that is, to reduce the resource sharing red line V i RedThe higher the priority and the greater the gap between the current simulation time scale and the scale requirement, the greater the reduction of the resource sharing red line, which can better guarantee the computing resources of its own container and ensure the normal operation of high-priority services. The adjustment calculation of the resource sharing red line is as follows:

[0109]

[0110] in, Indicates the service priority; L is the maximum priority; Simulation time scale requirement for container i; Indicates that container resource i shares computing resource R with container j i,j The remaining computing resources, is the simulation complexity of the container i service, and is brought into the mapping G i (R,p SC ) obtain the simulation time scale that can be achieved by the container j's own business after sharing computing resources; is the scale gap function, which represents the simulation time scale that can be satisfied by the current resources and simulation time scale requirements The larger the gap, the larger the value, and the time scale cannot meet the requirements. therefore

[0111] S2.2: Low-priority service cost adjustment method based on soft adjustment

[0112] If the service carried by container i with sufficient resources is a low-priority service and the time scale requirement of its own service simulation cannot be met, the resource sharing segmentation cost is adjusted by soft adjustment, that is, the segmentation threshold V is reduced. i TS1 、V i TS2 , increase the basic unit cost μ i The higher the container priority and the greater the gap between the current simulation time scale and the required scale, the greater the increase in the basic unit cost and the greater the decrease in the segmentation threshold, making resource sharing more difficult. It is more likely to ensure the normal operation of its own business or sacrifice the simulation time scale of its own business to obtain higher global benefits. The basic unit cost and segmentation threshold are adjusted and calculated as follows:

[0113]

[0114] Among them, μ i represents the basic unit cost; Indicates the service priority; L is the maximum priority; The meaning is the same as (7), indicating the gap between the simulation time scale that the current resources can meet and the simulation time scale requirement. The larger the gap, the larger the value.

[0115] Embodiment 3:

[0116] like Figure 2 As shown, the novel distribution network multi-time scale fine-grained parallel simulation system of this embodiment includes: a business information division module, a resource calculation module required for business simulation, a container information recording module, a segmented cost mechanism construction module, a container classification module, a matching module, a business demand judgment module, a cost adjustment module and a resource transfer module.

[0117] Business information division module: divides the business priority, complexity and time scale of business simulation;

[0118] Module for calculating resources required for business simulation: Constructs mapping and inverse mapping based on historical data sets, and calculates the computing resources required to achieve business simulation on the desired time scale;

[0119] Container information recording module: records the total number of distribution network computer cluster containers and the computing resources available for different containers, and uploads them to the segmented cost mechanism construction module;

[0120] Segment cost mechanism construction module: set segment thresholds and resource sharing red lines, and divide the computing resources that can be shared by containers into three segments according to the set values. Calculate the unit cost of each segment of computing resources, that is, the segment cost function, and upload it to the container classification module.

[0121] Container classification module: Classifies and sorts containers according to available computing resources, business priorities, and required additional computing resources, obtains a collection of containers with abundant resources and containers with scarce resources, and uploads them to the matching module;

[0122] Matching module: Take out containers from the resource-scarce container set, and at the same time, take out containers with abundant resources in turn to match them, and calculate the qualified computing resource transfer amount, so that the transferred computing resources can meet the needs of the resource-scarce container as much as possible, and upload the matching results and resource transfer amount to the business demand judgment module;

[0123] Business demand judgment module: Based on the matching results and resource transfer amounts, after considering the transfer of all container resources, whether the resource-rich container can meet its own business needs, the demand judgment results, matching results and resource transfer amounts are uploaded to the cost adjustment module.

[0124] Cost adjustment module: According to the judgment result of the business demand judgment module, if the containers with sufficient resources can meet their own needs, the matching results and resource transfer amount will be uploaded to the resource transfer module, otherwise the cost adjustment will be performed. The cost adjustment module uses hard adjustment methods for containers of high-priority businesses to lower the resource sharing red line of the containers, directly reducing the amount of computing resources shared with the outside world; it uses soft adjustment methods for containers of low-priority businesses to increase the basic unit cost and lower the segmentation threshold, so that the resource sharing cost increases, indirectly reducing the amount of computing resources shared with the outside world, and returns to the matching module after adjustment;

[0125] Resource transfer module: transfers computing resources according to matching results and resource transfer amounts.

[0126] The above system is used to carry out multi-time scale fine-grained parallel simulation of the new distribution network. The specific steps are as follows.

[0127] Step 1: Initialize each module;

[0128] Step 2: The container information recording module records the total number of distribution network computer cluster containers I and the computing resources available for different containers Business information division module prioritizes business Complexity of business simulation and time scale to divide;

[0129] Step 3: The business simulation required resource calculation module defines the mapping and inverse mapping of business simulation time scale and computing resources according to the historical database, and calculates the computing resources R′ required to achieve the business simulation of the expected time scale i ;

[0130] Step 4: Segmentation cost mechanism building module sets segmentation threshold V i TS1 、V i TS2 And the resource sharing red line V i Red , segment the computing resources that can be shared, and obtain the unit cost C of sharing each segment of computing resources i,j (x) and the total cost function for container i to actually share computing resources with container j

[0131] Step 5: The container classification module classifies and sorts the containers according to the available computing resources, business priorities, and required additional computing resources, and obtains a set of containers with sufficient resources U = {u1,u2,…,u N} and the set of resource-scarce containers W = {w1,w2,…,w M}.

[0132] Step 6: The matching module takes out containers w from the set W in sequence m , for container w m , take out container u from set U in sequence n Match; according to container u n The resource sharing segment total cost function And the container w m Required Resources Calculate the total cost of resource transfer Calculate the income at the same time Resource sharing value coefficient In satisfying Under the premise of , as much computing resources as possible are transferred, the containers are matched and the resource transfer amount is After all containers are matched, the matching results and resource transfer amounts are uploaded to the demand judgment module.

[0133] Step 7: After all container resources are transferred, the business demand judgment module determines whether the resource-sufficient container can meet its own business needs, and uploads the judgment result, matching result and resource transfer amount to the cost adjustment module.

[0134] Step 8: If the resource-rich containers can meet their own business needs, the cost adjustment module uploads the matching results and resource transfer amount to the resource transfer module. If the resource-rich containers cannot meet their own business needs, cost adjustment is performed, and hard adjustment is performed for high-priority businesses, otherwise soft adjustment is performed. After the cost adjustment is completed, the cost adjustment module returns the adjustment data to the matching module and re-matches the containers.

[0135] Step 9: The resource transfer module transfers computing resources according to the matching results and resource transfer amounts.

[0136] The preferred embodiments of the present invention are described in detail above, but the present invention is not limited to the above embodiments, and various changes can be made within the knowledge scope of ordinary technicians in this field without departing from the purpose of the present invention.

[0137] Many other changes and modifications may be made without departing from the concept and scope of the present invention.It should be understood that the present invention is not limited to the specific embodiments, and the scope of the present invention is defined by the appended claims.

Claims

1. A novel multi-time scale fine-grained parallel simulation method for distribution network, characterized in that: The method comprises: S1: Divide the business priorities, complexity and time scale of business simulation, consider the different containers of each computer in the computer cluster, and record the computing resources; The step S1 specifically includes: Determine the total number of computer cluster containers to be I, and the i-th container has a total computing resource Each container carries the corresponding business. Containers with sufficient resources can share computing resources with other containers. The shared resources of container i do not exceed the resource sharing red line V. i Red , according to the importance of the business, the container main business is divided into L levels according to priority, and the priority level of container i business is obtained The larger the value, the higher the service priority; considering the simulation complexity of container i service The larger the value, the greater the complexity of the simulation; considering the simulation time scale requirements of container i business Covering the full time scale physical process from steady-state power flow to nanosecond electromagnetic transient state, The larger the value, the larger the time scale of the simulation; Define mappings based on historical database Input is the simulation time scale of container i service and simulation complexity Output is the computing resources required for container i business simulation The smaller the simulation time scale and the higher the simulation complexity, the more computing resources are required; at the same time, the mapping is defined The input is the computing resources R and simulation complexity that can be provided for the business at present. The output is the business simulation time scale. The more computing resources available and the lower the simulation complexity, the smaller the simulation time scale that can be achieved. S2: Set the segmentation threshold and resource sharing red line, and divide the computing resources that can be shared by the container into three segments according to the set values, and obtain the unit cost and total cost function of each segment of computing resources; The step S2 specifically includes: Segment threshold V i TS1 、V i TS2 And the resource sharing red line V i Red The initial settings are as follows: in, Calculate the total amount of resources for container i; is the priority of the service of container i; L is the maximum service priority; Simulate the complexity of container i business, is its minimum and maximum value; is the time scale requirement for container i business simulation, is its minimum and maximum value; According to the segmented threshold and resource sharing red line, container i shares resources based on three threshold ranges; for the first range 0<x≤V i TS1 , that is, less than the threshold V i TS1 The shared resources in this section have almost no impact on the simulation time scale of the business itself, so a fixed unit cost is used; for the second section V i TS1 <x≤V i TS2 That is, it exceeds the threshold V i TS1 But not greater than V i TS2 Part of the shared resources, the computing resource sharing in this section has an impact on the simulation time scale of its own business, so a linearly increasing unit cost is adopted, and the threshold value V is exceeded. i TS1 The more, the higher the unit cost; for the third segment V i TS2 <x≤V i Red That is, it exceeds the threshold V i TS2 But not greater than the resource sharing red line V i Red The shared resources of this part have a great impact on the simulation time scale of its own business, so the unit cost of exponential growth is adopted, and the cost exceeds the threshold V i TS2 The more, the higher the unit cost; finally, we get the segmented cost function C i (x) represents the unit cost of container i sharing each segment of computing resources with other containers, which is calculated as follows: Among them, μ i V represents the basic unit cost of processing services by container i; i Red V represents the resource sharing red line; i TS1 、V i TS2 is the segmentation threshold; the unit cost of the second segment is (μ i ,2μ i ]; the third unit cost is between (μ i e,μ i e 2 ]; according to the resource sharing segmentation cost function C i (x) and the resource R that container i actually shares with container j i,j , we can get the total cost function of container i actually sharing computing resources with container j The calculation is as follows: The abundant container i shares computing resources with container j to meet the time scale requirement of container j business simulation, bringing higher business value. Considering the value improvement brought by the time scale improvement, the participation of abundant container i in container j simulation can bring benefits E. i,j , calculated as follows: in, is the total computing resources of container j, is the simulation complexity of container j business, brought into mapping G j (R,p j SC ) to obtain the simulation time scale that can be achieved by container j itself; R i,j is the computing resources shared by container i to container j, The simulation time scale of container j business is considered when container i shares computing resources with container j; is the time scale revenue function of container j’s business, They represent the simulation time scales that can be achieved before and after resource sharing, respectively. The time scales before and after sharing are input into the time scale benefit function, and the value brought by the improvement of the simulation time scale of business j is output; Containers sharing computing resources must generate additional value, that is, the value generated by container resource sharing must exceed the cost of resource sharing. Consider how much value can be generated for other containers when 1 CPU cycle of computing resources is transferred from a container with abundant resources to other containers. That is, resource transfer can only be performed when the resource sharing value is greater than or equal to αi times the cost; the resource sharing value coefficient αi is related to the business priority and simulation scale of the resource-abundant container. The greater the priority, the smaller the simulation scale, and the greater αi. The calculation is as follows: in, Indicates rounding up; is the simulation scale; is the maximum value of the simulation scale; For business priorities; S3: Classify and sort the containers according to the available computing resources, business priorities, and required additional computing resources to obtain a set of containers with abundant resources and containers with scarce resources. Take out containers from the set of containers with scarce resources, and at the same time, take out containers with abundant resources in turn to match them, and calculate the computing resource transfer amount that meets the conditions. S4: Based on the matching results and the amount of resource transfer, determine whether the containers with sufficient resources can meet their own business needs after all container resources are transferred. Based on the judgment results, if all containers with sufficient resources can meet their own needs, the matching results and the amount of resource transfer will be transferred. Otherwise, cost adjustment will be performed. Hard adjustment measures will be used for containers with high-priority services, and soft adjustment measures will be used for containers with low-priority services. Re-matching will be performed after adjustment.

2. According to claim 1, a novel multi-time scale fine-grained parallel simulation method for distribution network is characterized in that: In step S1, the total number of distribution network computer cluster containers I and the computing resources available for different containers are counted. Business information division module prioritizes business Complexity of business simulation and time scale The business simulation time scale and computing resources are mapped and inversely mapped according to the historical database, and the computing resources R required to achieve the business simulation of the desired time scale are calculated. i ′.

3. According to claim 1, a novel multi-time scale fine-grained parallel simulation method for distribution network is characterized in that: The step S3 specifically includes: S31: Classify I containers; consider That is, all containers with sufficient computing resources are allocated according to the available computing resources. Sort in descending order, and by business priority when available computing resources are equal Sort in ascending order and obtain a set U containing N containers with sufficient resources = {u1,u2,…,u N },u n The container index representing the nth element in the collection, where the container with abundant computing resources and low business priority has a small index; consider That is, all containers whose original container computing resources cannot meet the time scale simulation requirements are allocated according to business priorities. Sort in descending order, with the same priority level based on the additional computing resources required Sort in descending order and get a set W containing M resource-scarce containers = {w1,w2,…,w M },w m The container index of the mth element in the collection, where the container with high business priority and more additional computing resources required has a small index; That is, containers whose computing resources of the original container can just meet their own time scale simulation requirements do not participate in the computing resource transfer; S32: Initialization, set m=n=1; S33: Take out containers w from the set W in sequence m , for container w m , take out container u from set U in sequence n Match so that container u n Satisfy container w as much as possible m business needs, proceed to step S34; S34: At this time, according to the container u n The resource sharing segment total cost function And the container w m Required Resources Calculate the total cost of resource transfer Calculate the income at the same time Resource sharing value coefficient In satisfying Under the premise of , as much computing resources as possible are transferred, the containers are matched and the resource transfer amount is Let m=m+1, n=n+1 and return to step S33 to continue matching the remaining containers; if m>M or n>N, it means that all containers are matched, and proceed to step S35; S35: After considering the transfer of all container resources, determine whether the container with sufficient resources can meet its own business needs. If not, adjust the cost.

4. According to claim 1, a novel multi-time scale fine-grained parallel simulation method for distribution network is characterized in that: In step S4, the high priority service cost adjustment method based on hard adjustment specifically includes: If the service carried by container i with sufficient resources is a high-priority service and the time scale requirement of its own service simulation cannot be met, hard adjustment is used to adjust the resource sharing segment cost, that is, to reduce the resource sharing red line V i Red The higher the priority and the greater the gap between the current simulation time scale and the scale requirement, the greater the reduction of the resource sharing red line, which can better guarantee the computing resources of its own container and ensure the normal operation of high-priority services. The adjustment calculation of the resource sharing red line is as follows: in, Indicates the service priority; L is the maximum priority; Simulation time scale requirement for container i; Indicates that container resource i shares computing resource R with container j i,j The remaining computing resources, is the simulation complexity of the container i service, and is brought into the mapping G i (R,p SC ) obtain the simulation time scale that can be achieved by the container j's own business after sharing computing resources; is the scale gap function, which represents the simulation time scale that can be satisfied by the current resources and simulation time scale requirements The larger the gap, the larger the value, and the time scale cannot meet the requirements. therefore 5. According to claim 1, a novel multi-time scale fine-grained parallel simulation method for distribution network is characterized in that: In step S4, the low priority service cost adjustment method based on soft adjustment specifically includes: If the service carried by container i with sufficient resources is a low-priority service and the time scale requirement of its own service simulation cannot be met, the resource sharing segmentation cost is adjusted by soft adjustment, that is, the segmentation threshold V is reduced. i TS1 、V i TS2 , increase the basic unit cost μ i The higher the container priority and the greater the gap between the current simulation time scale and the scale requirement, the greater the increase in the basic unit cost and the greater the decrease in the segmentation threshold, making resource sharing more difficult. It is more likely to ensure the normal operation of its own business or sacrifice the simulation time scale of its own business to obtain higher global benefits. The basic unit cost and segmentation threshold are adjusted and calculated as follows: Among them, μ i represents the basic unit cost; Indicates the service priority; L is the maximum priority; The meaning is the same as (7), indicating the gap between the simulation time scale that the current resources can meet and the simulation time scale requirement. The larger the gap, the larger the value.

6. A new type of multi-time scale fine-grained parallel simulation system for distribution network, characterized by: The system is applied to the method described in any one of claims 1 to 5, and the system comprises: a business information division module, a business simulation required resource calculation module, a container information recording module, a segmented cost mechanism construction module, a container classification module, a matching module, a business demand judgment module, a cost adjustment module and a resource transfer module; The business information division module: divides the business priority, business simulation complexity and time scale; The business simulation required resource calculation module: constructs mapping and inverse mapping according to the historical data set, and calculates the computing resources required to achieve the business simulation of the desired time scale; The container information recording module records the total number of distribution network computer cluster containers and the computing resources available for different containers, and uploads them to the segmented cost mechanism construction module; The segmentation cost mechanism construction module: sets segmentation thresholds and resource sharing red lines, and divides the computing resources that can be shared by the container into three segments according to the set values, calculates the unit cost of each segment of computing resources, that is, the segmentation cost function, and uploads it to the container classification module; The container classification module: classifies and sorts the containers according to the available computing resources, business priorities and required additional computing resources, obtains a collection of resource-rich containers and resource-scarce containers, and uploads them to the matching module; The matching module: takes out containers from the resource-scarce container set, and at the same time, sequentially takes out containers with abundant resources to match them, and calculates the qualified computing resource transfer amount, so that the transferred computing resources can meet the needs of the resource-scarce container as much as possible, and uploads the matching results and resource transfer amount to the business demand judgment module; The business demand judgment module: based on the matching results and the resource transfer amount, considers whether the resource-rich container can meet its own business needs after all container resources are transferred, and uploads the demand judgment results, matching results and resource transfer amount to the cost adjustment module; The cost adjustment module: according to the judgment result of the business demand judgment module, if the containers with sufficient resources can meet their own needs, the matching result and resource transfer amount are uploaded to the resource transfer module; otherwise, the cost adjustment is performed. The cost adjustment module adopts hard adjustment means for containers of high-priority services; and soft adjustment means for containers of low-priority services, and returns them to the matching module after adjustment; The resource transfer module is used to transfer computing resources according to the matching results and the resource transfer amount.

Images