Calculation method of calculation network arrangement scheduling strategy based on multi-factor algorithm

Abstract

The invention relates to the technical field of big data cloud computing, in particular to a computing method of a computing network arrangement scheduling strategy based on a multi-factor algorithm. By fully knowing user requirements and providing personalized resource schemes, actual requirements of users are met, and user experience and satisfaction are improved. Through an accurate resource allocation algorithm, calculation, network, storage and other resources are reasonably allocated according to actual demands, the resource utilization rate is improved, and resource waste is reduced. The automatic resource allocation process reduces the necessity of manual intervention, improves the resource allocation efficiency, and accelerates the resource supply speed. Intelligent scheduling is carried out according to user demands and a resource allocation algorithm, so that waste caused by excess or insufficient resources is avoided, and resource purchase and maintenance costs are reduced. According to the preference and preference of the user, a plurality of recommendation schemes suitable for the user are generated, more choices are provided for the user, the final scheme is determined according to secondary screening of the user, and personalized requirements are met.

Description

Technical Field

[0001] The present invention relates to the technical field of big data cloud computing, and in particular to a method for calculating a computing network orchestration scheduling strategy based on a multi-factor algorithm. Background Art

[0002] In recent years, technologies such as the Internet, big data, cloud computing, artificial intelligence, and blockchain have accelerated innovation and have been increasingly integrated into the entire process of economic and social development. Countries are competing to formulate digital economic development strategies and introduce encouraging policies. The speed of digital economic development, the scope of radiation, and the depth of impact are unprecedented. It is becoming a key force in restructuring global factor resources, reshaping the global economic structure, and changing the global competitive landscape.

[0003] Computing power is the productivity in the digital economy era and a solid foundation for supporting the development of the digital economy. The key resources in the digital economy era are data, computing power and algorithms. Data is the new means of production, computing power is the new productivity, and algorithms are the new production relations, which constitute the most basic production cornerstone in the digital economy era. In recent years, my country has attached great importance to the development of computing power and proposed the strategic pattern of "East Data West Computing", namely the East Data West Computing Project, which refers to the orderly guidance of the computing power demand in the east to the west through the construction of a new computing power network system integrating data centers, cloud computing and big data, optimizing the layout of data center construction, and promoting the coordinated linkage between the east and the west. In February 2022, the construction of national computing power hub nodes was started in 8 places, including Beijing-Tianjin-Hebei, Yangtze River Delta, Guangdong-Hong Kong-Macao Greater Bay Area, Chengdu-Chongqing, Inner Mongolia, Guizhou, Gansu and Ningxia, and 10 national data center clusters were planned. At this point, the overall layout design of the national integrated big data center system has been completed, and the East Data West Computing Project has been officially launched in full swing. Simply put, it is to allow the computing power resources in the west to more fully support the calculation of data in the east and better empower digital development. The "number" in "East Data West Computing" refers to data, and "computing" refers to computing power, that is, the ability to process data. "East Data West Computing" is to guide the computing power demand in the east to the west in an orderly manner by building a new computing power network system integrating data centers, cloud computing, and big data, optimize the layout of data center construction, and promote coordinated linkage between the east and the west. This shows the importance of computing power.

[0004] With the rapid development of Internet technology, the application of technologies such as big data, artificial intelligence, cloud computing, and blockchain is becoming more and more extensive, the domestic computing power development environment is becoming more and more perfect, the domestic and international market competition for computing power resources is increasing, and international cooperation is expanding. The computing power is becoming more and more powerful, and the demand for computing resources is also increasing. However, the existing computing power resource scheduling strategies often have problems such as low resource utilization and uneven distribution. In order to solve these problems, the present invention proposes a calculation method for a computing network orchestration scheduling strategy based on a multi-factor algorithm. Summary of the invention

[0005] From the rise of the Internet to its rapid development today, the allocation of computer resources has always been a highly concerned issue. There are innumerable types of computer resources. For the convenience of research, we classify them into the following ten categories, namely computing, network, storage, resource type, application scenario, user preference, business location, billing method, usage time and purchase quantity; computing can be divided into VCPU, memory and GPU; network can be divided into bandwidth and latency; storage can be divided into disk type and capacity; resource type can be divided into government network resource area and Internet resource area; application scenarios can be divided into data processing tasks, enterprise back-end applications, website front-end, web service network, distributed analysis, high-performance database, data mining and analysis, in-memory database, high-performance scientific computing, GPU computing, GPU graphics acceleration and NPU computing; user preferences can be divided into comprehensive needs, cost priority, distance priority, performance priority and green electricity priority; billing methods can be divided into volume-based calculation and annual and monthly packages.

[0006] Based on the above classification, the resource allocation of the present invention is divided into the following steps: first, at the user level, users apply for relevant resources according to their own needs; second, system allocation, the system allocates according to the data training results, according to the user's needs for computing, network, storage, resource type, application scenario, user preference, business location, billing method, usage time and purchase quantity; third, recommendation plan, based on the allocation results and custom weights, recommend multiple appropriate plans for users to choose; fourth, user selection, users can make a secondary choice based on the system recommended plan and their own preferences; fifth, determine the plan, according to the user's choice, determine the plan. Sixth, resource sale, according to the plan requirements, calculate the selling price and feedback the price to the user; seventh, put into use, officially put into use according to the plan.

[0007] In order to achieve the above steps, the present invention provides a calculation method for a computing network orchestration scheduling strategy based on a multi-factor algorithm to improve the utilization and allocation efficiency of computer computing resources.

[0008] To achieve the above object, the present invention adopts the following technical solutions:

[0009] A method for calculating a computing network scheduling strategy based on a multi-factor algorithm includes the following steps:

[0010] Computing network perception: Through the computing power scheduling system, the computing power gateway is used as the access point of computing power resources in the network. Through the BGP protocol extension technology, the computing power gateways in the computing power network are connected to obtain all cloud resource pools and computing power routing information within the management scope. The computing power gateway includes data collection and traffic data analysis.

[0011] Data collection is the basic data source required for route traffic monitoring and analysis. The system needs to be able to provide equipment information collection, route information collection, and traffic flow data collection to provide data support for traffic monitoring warning and traffic flow analysis.

[0012] Traffic data analysis supports traffic analysis and statistics based on router / interface dimensions, traffic analysis and statistics based on TCP / UDP / ICMP protocols and ports, traffic direction analysis and statistics based on networks / subnets (provincial networks / metropolitan area networks / IDCs), traffic direction analysis and statistics based on services (such as home customers / collective customers / 4G / Internet, etc.), metropolitan area network traffic matrix, metropolitan area network to other provincial networks, and metropolitan area network traffic matrix functions.

[0013] Computing network orchestration and scheduling: Computing resources are allocated to users based on their computing needs and resource availability. In the Internet Protocol (IP) bearer network domain, through refined dynamic perception, the network controller or network node can create a computing routing table based on the computing resources and service status in the multi-cloud pool, and perform the orchestration and scheduling of computing resources and services accordingly.

[0014] The orchestration components generally include two parts: the cloud network orchestrator and the slice network orchestrator. The cloud network network orchestrator focuses on automatically connecting the cloud private network segments and opening up various provincial extension segments of cloud private lines; the slice network orchestrator connects to the wireless network, transmission network and core network sub-domains to realize the opening of slices.

[0015] As the only channel for real-time control and collection, cloud network acquisition and control supports all disciplines to complete data collection and control through a common software framework and tool capabilities. Data collection is stored in the cloud network basic data sharing platform in real time, and network port control capabilities are provided in accordance with openAPI standards. Based on the dynamic data perceived by the computing network, computing power deconstruction scheduling, network intelligent routing, resource automatic scheduling and opening, and real-time end-to-end transparent visualization of the process are realized, matching computing power consumers with the computing network resources that best meet their needs.

[0016] Computing power trading: The computing resources allocated to users are sold and the corresponding fees are charged. Computing power trading consists of two functional layers: management and control and trading. The management and control layer is responsible for the collection, allocation and network scheduling of underlying resource information. The trading layer is responsible for the resource demand analysis of the business, the generation of transaction views, the selection of resources and the generation of transaction bills. The computing power network management and trading platform distributes and collects underlying resource information, integrates the user's resource needs to generate a resource transaction view, and allows users to make independent or intelligent resource selections, realizes the dynamic and flexible allocation of operators' computing power resources, and provides users with comprehensive service capabilities of ICT collaborative integration.

[0017] The beneficial effects of the present invention are:

[0018] Improve user satisfaction: By fully understanding user needs and providing personalized resource solutions, we can meet users' actual needs and improve user experience and satisfaction. Improve resource utilization: Through accurate resource allocation algorithms, we can reasonably allocate computing, network, storage and other resources according to actual needs, improve resource utilization and reduce resource waste. Improve system efficiency:

[0019] The automated resource allocation process reduces the need for manual intervention, improves the efficiency of resource allocation, and speeds up resource supply. Reduce costs: Intelligent scheduling based on user needs and resource allocation algorithms avoids waste caused by excess or insufficient resources and reduces resource procurement and maintenance costs. Provide personalized choices: Generate multiple recommended solutions suitable for users based on user preferences, provide users with more choices, and determine the final solution based on user secondary screening to meet personalized needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is a schematic diagram of the implementation process of the present invention;

[0021] Figure 2 The figure is a flow chart of the calculation method of the present invention. DETAILED DESCRIPTION

[0022] A method for calculating a computing network scheduling strategy based on a multi-factor algorithm includes the following steps:

[0023] Computing network perception: Through the computing power scheduling system, the computing power gateway is used as the access point of computing power resources in the network. Through the BGP protocol extension technology, the computing power gateways in the computing power network are connected to obtain all cloud resource pools and computing power routing information within the management scope. The computing power gateway includes data collection and traffic data analysis.

[0024] Data collection is the basic data source required for route traffic monitoring and analysis. The system needs to be able to provide equipment information collection, route information collection, and traffic flow data collection to provide data support for traffic monitoring warning and traffic flow analysis.

[0025] Traffic data analysis supports traffic analysis and statistics based on router / interface dimensions, traffic analysis and statistics based on TCP / UDP / ICMP protocols and ports, traffic direction analysis and statistics based on networks / subnets (provincial networks / metropolitan area networks / IDCs), traffic direction analysis and statistics based on services (such as home customers / collective customers / 4G / Internet, etc.), metropolitan area network traffic matrix, metropolitan area network to other provincial networks, and metropolitan area network traffic matrix functions.

[0026] Computing network orchestration and scheduling: Computing resources are allocated to users based on their computing needs and resource availability. In the Internet Protocol (IP) bearer network domain, through refined dynamic perception, the network controller or network node can create a computing routing table based on the computing resources and service status in the multi-cloud pool, and perform the orchestration and scheduling of computing resources and services accordingly.

[0027] The orchestration components generally include two parts: the cloud network orchestrator and the slice network orchestrator. The cloud network network orchestrator focuses on automatically connecting the cloud private network segments and opening up various provincial extension segments of cloud private lines; the slice network orchestrator connects to the wireless network, transmission network and core network sub-domains to realize the opening of slices.

[0028] As the only channel for real-time control and collection, cloud network acquisition and control supports all disciplines to complete data collection and control through a common software framework and tool capabilities. Data collection is stored in the cloud network basic data sharing platform in real time, and network port control capabilities are provided in accordance with openAPI standards. Based on the dynamic data perceived by the computing network, computing power deconstruction scheduling, network intelligent routing, resource automatic scheduling and opening, and real-time end-to-end transparent visualization of the process are realized, matching computing power consumers with the computing network resources that best meet their needs.

[0029] Computing power trading: The computing resources allocated to users are sold and the corresponding fees are charged. Computing power trading consists of two functional layers: management and control and trading. The management and control layer is responsible for the collection, allocation and network scheduling of underlying resource information. The trading layer is responsible for the resource demand analysis of the business, the generation of transaction views, the selection of resources and the generation of transaction bills. The computing power network management and trading platform distributes and collects underlying resource information, integrates the user's resource needs to generate a resource transaction view, and allows users to make independent or intelligent resource selections, realizes the dynamic and flexible allocation of operators' computing power resources, and provides users with comprehensive service capabilities of ICT collaborative integration.

[0030] 1. Computing network perception

[0031] Computing network perception is to predict the overall usage of computing power and network resources in the domain through software and hardware facilities. By analyzing and judging historical data, business traffic, business nature and other factors that affect computing network resources, the usage of computing power and network resources is predicted, and resource planning and allocation are automatically made, so that computing power resources can be automatically elastically scaled to meet the business needs of different scenarios. Secondly, from the perspective of the platform, computing power perception technology can help governments and data center operators understand the trend of computing power demand and guide computing power planning by analyzing the computing network resources of the overall platform.

[0032] 1.1 Receive user demand information;

[0033] The computing system receives information, pre-processes the information, converts the data into recognizable system data, and transfers it to the next-level operating system.

[0034] 1.2 Retrieve available computing resources based on user demand information;

[0035] After receiving the relevant information, the computing network perception system predicts the overall usage of computing power and network resources in the domain through hardware and software facilities. It analyzes and evaluates historical data, business traffic, business nature and other factors that affect computing network resources, predicts the usage of computing power and network resources, and automatically makes resource planning and allocation that is consistent with the received information, and transfers the relevant information to the next level.

[0036] 1.3 Return the retrieved available computing resources and select the system based on user needs.

[0037] Receive the resources pre-matched for user information, plan, organize and sort them, and send the user information and matching resources to the computing network orchestration and scheduling system.

[0038] 2. Computing network orchestration and scheduling

[0039] Before preparing the best resource allocation plan for users, we still need to do some preparation and pre-process the system resources.

[0040] 1) Variable assumptions:

[0041]

[0042] 2) Preparation

[0043] Entropy is a measure used to measure the certainty of a random variable. The larger the entropy, the greater the uncertainty of the random variable. First, we divide the existing data into D as the training set, A as the attribute value, X as a finite random variable, and x i is the i-th training variable, so the probability distribution of X is:

[0044] P(X=x i )=pi,i=1,2,…,n.

[0045] From this we can get the entropy of the random variable

[0046]

[0047] By generalizing it, we can get H(D).

[0048] Conditional entropy H(Y|X) represents the uncertainty of random variable Y under the condition that random variable X is given. The conditional entropy of Y under the condition that X is given is:

[0049]

[0050] By generalizing it, we can get the conditional entropy between data sets:

[0051]

[0052] 3) Information Gain

[0053] We use information gain to measure the contribution of a feature value to the classification result.

[0054] g(D,A)=H(D)-H(D|A).

[0055] 4) Select the best features

[0056] Information gain is a commonly used feature selection method. It calculates the impact of different features on the classification of sample data, selects the most helpful feature as the optimal feature according to the information gain, and splits it. The size of the information gain is closely related to the impact of the feature on the classification. That is, the larger the information gain, the greater the impact of the feature on the classification. Therefore, we use the feature with the largest information gain to achieve the highest classification accuracy.

[0057] 5) Generate a decision tree

[0058] The optimal feature is selected by the value of information gain, and then the data is divided into several subsets according to the value of the optimal feature. The two steps of selecting the optimal feature and dividing the data set are performed for each subset until the conditions are met. Finally, the subtrees are merged to merge all the subtrees into a complete decision tree.

[0059] 2.1 Determine the optimal resource allocation plan based on the user's computing needs and resource availability;

[0060] As user information is input and received, appropriate solutions are screened out through our trained and improved decision tree. These solutions are scored, sorted and summarized according to the system's existing weight standards, and wait for the next instruction.

[0061] 2.2 Send the determined optimal resource allocation plan to the user;

[0062] It is detected that the decision tree has selected the corresponding solution, and the summary results are displayed to the user through the recommendation system, waiting for the user to confirm or make a second selection.

[0063] 2.3 Receive the user's confirmation information on the optimal resource allocation plan;

[0064] After the recommendation system recommends the results to the user, the user makes a choice. If the user confirms the choice, he / she will directly enter the plan allocation stage. If the user makes a second choice, that is, the user is more inclined to one aspect, he / she will be able to customize the weight and make a second choice among the existing results until the user finally decides on the plan.

[0065] 2.4 Allocate computing resources to users based on their confirmation of the optimal resource allocation plan.

[0066] After the user determines the plan, he / she enters the system allocation phase. The system allocates computing resources to the user based on the plan determined by the user, and then enters the trading system.

[0067] 3. Computing power trading

[0068] Provide various computing resources of the computing power provider to the computing power consumers on demand, including but not limited to the resource access of the computing power provider, analysis of the resource requirements of the computing power consumers and various business / application scenario requirements, etc., to match the best resources for the computing power users.

[0069] 3.1 Pricing the computing resources allocated to users;

[0070] 3.2 Release the priced computing resources to the market;

[0071] 3.3 Receive the user's purchase request for the priced computing resources;

[0072] 3.4 Based on the user's purchase request, the computing resources will be sold to the user and the corresponding fees will be charged.

[0073] Beneficial effects of the invention: The invention provides a commercial model for resource retrieval, allocation and sale based on a computing power scheduling system. Through this model, the utilization rate and allocation efficiency of computing resources can be effectively improved, providing users with more convenient and efficient computing services.

[0074] The specific technical effects of the present invention are as follows:

[0075] 1. Technical points

[0076] 1.1 User demand analysis: Collect user demand information, including computing, network, storage, resource type, application scenario, user preference, business location, billing method, usage duration and purchase quantity.

[0077] 1.2 Data training and model building: Use big data analysis and machine learning algorithms to train and model user needs and historical data to generate accurate resource allocation models.

[0078] 1.3 Resource allocation algorithm: Design and develop a resource allocation algorithm that comprehensively considers the weights of various indicators based on user needs, training models and other related information to achieve automated resource allocation decisions.

[0079] Recommendation plan generation: Based on the resource allocation results and user-defined weights, multiple recommendation plans suitable for user needs are generated to provide personalized choices.

[0080] 1.4 User selection and secondary screening: Allow users to make choices based on recommended options and personal preferences, and conduct secondary screening to ensure that the final selected option meets the user's actual needs.

[0081] 1.5 Plan determination and confirmation: Based on the user's choice, the resource allocation plan is finally determined and confirmed to ensure that resources are allocated accurately and timely.

[0082] 2. Effect

[0083] 2.1 Improve user satisfaction: By fully understanding user needs and providing personalized resource solutions, we can meet users' actual needs and improve user experience and satisfaction.

[0084] 2.2 Improve resource utilization: Through precise resource allocation algorithms, computing, network, storage and other resources are reasonably allocated according to actual needs to improve resource utilization and reduce resource waste.

[0085] 2.3 Improve system efficiency: The automated resource allocation process reduces the need for manual intervention, improves resource allocation efficiency, and speeds up resource supply.

[0086] 2.4 Reduce costs: Intelligent scheduling based on user needs and resource allocation algorithms avoids waste caused by excess or insufficient resources and reduces resource procurement and maintenance costs.

[0087] 2.5 Provide personalized choices: Generate multiple recommended solutions suitable for users based on their likes and preferences, provide users with more choices, and determine the final solution based on the user's secondary screening to meet personalized needs.

[0088] Glossary

[0089] Multi-factor algorithm: Multi-factor algorithm can be applied to evaluate and schedule computing tasks to optimize resource utilization and performance. The algorithm can comprehensively consider multiple factors, such as the computing requirements of the task, resource availability, network bandwidth, task priority, etc., to determine the best task scheduling solution.

[0090] Computing network orchestration: Network orchestration refers to the process of managing and scheduling the computing network. Through computing network orchestration, computing resources can be uniformly managed and scheduled to achieve optimal configuration and efficient utilization of computing resources. Multi-factor algorithms can be applied to computing network orchestration to improve the efficiency and accuracy of computing scheduling.

Claims

1. A calculation method for network scheduling strategy based on a multi-factor algorithm, characterized in that The following steps are involved: Computing network perception: Through the computing power scheduling system, the computing power gateway is used as the access point of computing power resources in the network. The computing power gateways in the computing power network are connected through the BGP protocol extension technology, so as to obtain all cloud resource pools and computing power routing information within the management scope. The computing power gateway includes data collection and traffic data analysis; Data collection is the basic data source required for route traffic monitoring and analysis. The system needs to be able to provide equipment information collection, route information collection, and traffic flow data collection to provide data support for traffic monitoring warning and traffic flow analysis. Traffic data analysis supports traffic analysis and statistics based on router / interface dimensions, TCP / UDP / ICMP protocols and ports, network / subnet-based traffic flow analysis and statistics, business-based traffic flow analysis and statistics, metropolitan area network traffic matrix, metropolitan area network to other provincial networks, and metropolitan area network traffic matrix functions; Computing network orchestration and scheduling: Computing resources are allocated to users based on their computing needs and resource availability. In the Internet IP protocol bearer network domain, through refined dynamic perception, the network controller or network node creates a computing routing table based on the computing resources and service status in the multi-cloud pool, and orchestrates and schedules computing resources and services accordingly. The orchestration components generally include two parts: the cloud network orchestrator and the slice network orchestrator. The cloud network orchestrator focuses on automatically connecting the cloud private network segments and opening up the provincial extension segments of various cloud private lines. The slice network orchestrator connects to the wireless network, transmission network and core network subdomains to realize the opening of slices; As the only channel for real-time control and collection, cloud network acquisition and control supports various disciplines to complete data collection and control through a common software framework and tool capabilities. Data collection is stored in the cloud network basic data sharing platform in real time, and network port control capabilities are provided in accordance with the openAPI standard. Based on the dynamic data perceived by the computing network, computing power deconstruction scheduling, network intelligent routing, automatic resource scheduling and activation, and real-time end-to-end transparent visibility of the process are realized, matching computing power consumers with the computing network resources that best meet their needs. Computing power trading: selling computing resources allocated to users and charging corresponding fees; Computing power trading consists of two functional layers: management and control, and trading. The management and control layer is responsible for the collection, allocation, and network scheduling of underlying resource information; the trading layer is responsible for business resource demand analysis, generation of transaction views, resource selection, and generation of transaction bills. The computing power network management and trading platform distributes and collects underlying resource information, generates resource transaction views based on user resource needs, and allows users to make independent or intelligent resource selections, thereby achieving dynamic and flexible allocation of operator computing power resources and providing users with comprehensive service capabilities for ICT collaborative integration.

Images