Cloud-based computer network resource management system

Abstract

The invention relates to the technical field of network optimization, in particular to a cloud-based computer network resource management system, which comprises a network bandwidth demand analysis module, a cloud platform computer network task management module, a cloud platform computer network task management module, a cloud platform computer network task management module, a cloud platform computer network task management module, a cloud platform computer network task management module and a cloud platform computer network task management module, and according to the peak value change and the distribution condition of the task bandwidth demand, marking the bandwidth demand area exceeding a preset frequency threshold value, and generating network bandwidth distribution characteristic data. According to the invention, a resource management process taking data driving, characteristic analysis and real-time prediction as cores is formed through dynamic distribution and optimization management of network resources. In the bandwidth demand analysis, the over-limit region is marked through the peak change of the time sequence data, the accuracy of resource distribution is optimized, and the problem of bandwidth waste or resource overload is effectively avoided.

Description

Technical Field

[0001] The present invention relates to the technical field of network optimization, and particularly to a cloud-based computer network resource management system. Background Art

[0002] The technical field of network optimization includes relevant methods and systems for improving the utilization rate of network resources and service performance through technical means. The core content of this technical field includes network topology design, traffic scheduling and management, bandwidth allocation strategies, and resource allocation optimization, etc. The goal of network optimization technology is to reasonably allocate limited network resources, improve the efficiency and stability of data transmission, and ensure the communication quality between different network nodes.

[0003] Among them, a computer network resource management system refers to a system for dynamically allocating and efficiently managing various resources in a network. For the allocation and scheduling of network resources, it covers the centralized management and optimization of computing resources, storage resources, and bandwidth resources. Specifically, by constructing a unified resource allocation model, according to real-time network status information, using optimization algorithms to analyze and schedule resource requests, so as to realize the dynamic adjustment and allocation of resources.

[0004] The existing technology has limitations mainly based on static allocation in resource allocation and scheduling, and it is difficult to respond to the dynamic changes of network traffic in real time. The unified resource allocation model is difficult to reflect personalized needs in the face of complex network environments, which may lead to insufficient or redundant resource allocation. The existing technology fails to refine the fluctuation characteristics of time series data, resulting in the inability of bandwidth allocation to respond to peak changes in a timely manner. This lag is likely to cause network congestion and a decline in service quality in high-traffic scenarios. The analysis of the dependency relationship between tasks is relatively rough, ignoring the complex interaction characteristics between nodes, which is likely to cause an unreasonable task execution order and prolong the task completion time. The bandwidth occupancy monitoring data is only used as a static reference and fails to predict future bandwidth requirements through historical trend analysis, which may lead to the lag of resource allocation and further exacerbate the bottleneck of network resource usage. These deficiencies make the network resource management system have low adaptability and scheduling efficiency in the face of sudden tasks or complex task flows, and are prone to problems such as task delays and system instability. Summary of the Invention

[0005] The purpose of the present invention is to solve the disadvantages existing in the prior art, and to propose a cloud-based computer network resource management system.

[0006] In order to achieve the above purpose, the present invention adopts the following technical solutions: The cloud-based computer network resource management system includes:

[0007] The network bandwidth requirement analysis module collects the bandwidth usage records and real-time network traffic data in the cloud platform computer network tasks, analyzes the peak changes in the time series data of bandwidth usage and network traffic, marks the bandwidth requirement areas exceeding the preset frequency threshold with reference to the peak changes and the distribution of task bandwidth requirements, and generates network bandwidth distribution characteristic data;

[0008] Based on the network bandwidth distribution characteristic data, the task flow dependency scheduling module analyzes the traffic interaction characteristics and dependency relationships among computer network tasks, determines the sequential dependency order of the computer network task flow, adjusts the execution sequence of task nodes, and generates an optimized execution sequence of the task flow;

[0009] The resource isolation configuration module analyzes the computing resources, storage resources, and bandwidth resource requirements of task nodes in the computer network in the optimized execution sequence of the task flow, selects the resource boundaries of task nodes, and starting from the resource boundaries, adjusts the division of virtual resource pools in the network according to task priorities and bandwidth occupancy, and generates a computer network resource allocation list;

[0010] The bandwidth dynamic optimization module applies the computer network resource allocation list, analyzes the bandwidth occupancy change rate and latency according to the traffic usage status, divides the bandwidth occupancy change rate and latency according to time windows, predicts the bandwidth requirement trend within the time window, and generates an optimized result of computer network bandwidth resources.

[0011] As a further solution of the present invention, the step of analyzing the peak changes in the time series data of bandwidth usage is specifically as follows:

[0012] Collect the bandwidth usage records and real-time network traffic data in the cloud platform computer network tasks, classify the bandwidth usage records in chronological order and extract the corresponding values, collect and store the real-time network traffic data, and perform adjustment and outlier processing on the bandwidth usage records and real-time network traffic to obtain bandwidth and traffic time series data;

[0013] Based on the bandwidth and traffic time series data, use the formula:

[0014]

[0015] Calculate the fluctuation index P of the i-th time series i ;

[0016] where S t,i is the value at time t in the i-th time series, max(S t,i ) is the maximum value of the i-th time series, min(S t,i ) is the minimum value of the i-th time series, S t,i,j is the value of the j-th data point in the time series S t,i ; is the mean of the time series S t,i and m is the total number of data points in the time series S t,i .

[0017] As a further solution of the present invention, the step of obtaining the network bandwidth distribution characteristic data is specifically as follows:

[0018] Based on the fluctuation indices corresponding to the bandwidth usage time series and the network traffic time series, analyze the bandwidth demand distribution characteristics of network nodes, mark the time periods when the node bandwidth demand peaks exceed the limit as the over-limit regions, and generate the bandwidth demand over-limit node data;

[0019] Based on the bandwidth demand over-limit node data, integrate the over-limit frequency, time period characteristics, and peak bandwidth values, map the data uniformly to the time axis, and count the characteristics of the over-limit regions to obtain the network bandwidth distribution characteristic data.

[0020] As a further solution of the present invention, the step of determining the sequence of dependencies of computer network task flows is specifically as follows:

[0021] Based on the network bandwidth distribution characteristic data, analyze the traffic interaction characteristics and dependency relationships between computer network tasks, and use the formula:

[0022]

[0023] Calculate the traffic interaction correlation index D of tasks i' and j' i′j′ ;

[0024] where F i′,k is the traffic value of task i' within time period k, F j′,k is the traffic value of task j' within time period k, and n is the number of time periods;

[0025] According to the traffic interaction correlation index of each task, sort out the input and output traffic of task nodes, analyze the traffic distribution and interaction order, and sort the dependency relationships of tasks according to the traffic distribution and interaction order to obtain the sequence of dependencies of the task flow.

[0026] As a further solution of the present invention, the step of obtaining the optimized execution sequence of the task flow is specifically as follows:

[0027] Based on the sequence of dependencies of the task flow, mark the dependency link relationships between task nodes one by one, sort out the traffic interaction intensities of all task nodes and compare the correlations, mark the direct dependency links and indirect dependency links, and generate a task node relationship diagram;

[0028] Based on the dependency link relationship in the task node relationship graph and the network resource requirements of the task nodes, sort the bandwidth requirements and traffic interaction intensities of the task nodes, adjust the execution order of the task nodes according to the resource allocation priority, and generate an optimized execution sequence of the task flow.

[0029] As a further solution of the present invention, the step of determining the resource boundary of the selected task node is specifically as follows:

[0030] Based on the optimized execution sequence of the task flow, extract the resource requirements, storage resource requirements, and bandwidth resource requirements from the operation monitoring data of the nodes, classify and judge the resource requirement levels of each node according to a preset range, and organize the resource requirements to generate the resource requirement characteristic data of the task nodes;

[0031] Based on the resource requirement characteristic data of the task nodes, use the formula;

[0032]

[0033] Calculate the resource weight value V of task node a a , quantify the overall level of node resource requirements, and determine the resource boundary of the task node according to the overall level of node resource requirements;

[0034] Among them, C a is the computing resource requirement of task node a, H a is the storage resource requirement of task node a, B a is the bandwidth resource requirement of task node a, w c , w h , w b respectively represent the weight parameters of computing resources, storage resources, and bandwidth resources.

[0035] As a further solution of the present invention, the step of obtaining the computer network resource allocation list is specifically as follows:

[0036] Based on the resource boundary of the task node, divide the task nodes into multiple groups according to the priority, obtain the peak value and average value of the bandwidth occupied by each node, and allocate the bandwidth resources according to the peak value and average value of the bandwidth occupied according to the node priority to generate the resource allocation result of the task nodes in the virtual resource pool;

[0037] Based on the resource allocation result of the task nodes in the virtual resource pool, record the allocation situations of computing resources, storage resources, and bandwidth resources according to the task node priority, and organize the virtual resource pool identifiers to which each node belongs, review and adjust the resource allocation result, and generate a computer network resource allocation list.

[0038] As a further solution of the present invention, the steps for obtaining the optimization result of the computer network bandwidth resources are specifically as follows:

[0039] Based on the computer network resource allocation list, the traffic usage status of the task nodes is monitored in real time. The traffic values and delay data of each time period are recorded by the bandwidth monitoring device. The bandwidth occupancy change rate and the delay situation are divided according to the time window, and the data is sorted to generate the time series data of the traffic usage of the task nodes.

[0040] Based on the time series data of the traffic usage of the task nodes, the formula

[0041]

[0042] is used to calculate the predicted bandwidth demand value B of the time window t′ + 1 t′+1 , and the bandwidth demand prediction result is obtained;

[0043] where B t′ is the bandwidth demand value within the time window t′, R t′ is the bandwidth occupancy change rate of the time window t′, indicating the change amplitude of the bandwidth demand within the current time window, is the average value of the bandwidth occupancy change rate of the time window, and γ is the adjustment coefficient;

[0044] Based on the bandwidth demand prediction result, the bandwidth allocation ratio of the task nodes is dynamically adjusted, the delay and occupancy of the traffic path are analyzed, the path planning is optimized, the traffic allocation rules of the task nodes are adjusted, and the data is recorded to generate the optimization result of the computer network bandwidth resources.

[0045] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0046] In the present invention, through the dynamic allocation and optimized management of network resources, a resource management process centered on data-driven, characteristic analysis, and real-time prediction is formed. In the bandwidth demand analysis, the over-limit regions are marked by the peak changes of time series data, optimizing the accuracy of resource distribution and effectively avoiding the problems of bandwidth waste or resource overload. The analysis of traffic interaction characteristics and dependency relationships, integrating the execution sequences and resource requirements of task nodes, helps to reasonably adjust the priorities among task nodes, reduce resource conflicts, and improve the task processing ability of the network. The combined method of resource demand classification, weight calculation, and group priority allocation significantly improves the allocation accuracy of bandwidth, computing resources, and storage resources, avoiding the randomness and imbalance of resource allocation. In the dynamic bandwidth optimization, through the real-time monitoring and analysis of the traffic change rate and latency data, the bandwidth demand trend within the time window is incorporated into the resource allocation process, thereby realizing the dynamic adjustment of resources and the optimal planning of paths. This not only enhances the system's adaptability to bursty traffic but also achieves the efficient utilization of resources, improving the operational stability and resource utilization efficiency of the entire network. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 is the system flow chart of the present invention;

[0048] Figure 2 is the flow chart for analyzing the peak changes of time series data of bandwidth usage in the present invention;

[0049] Figure 3 is the flow chart for obtaining the data of network bandwidth distribution characteristics in the present invention;

[0050] Figure 4 is the flow chart for determining the sequential dependency order of computer network task flows in the present invention;

[0051] Figure 5 is the flow chart for obtaining the optimized execution sequence of task flows in the present invention;

[0052] Figure 6 is the flow chart for selecting the resource boundaries of task nodes in the present invention;

[0053] Figure 7 is the flow chart for obtaining the computer network resource allocation list in the present invention;

[0054] Figure 8 is the flow chart for obtaining the optimized result of computer network bandwidth resources in the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0055] In order to make the objectives, technical solutions, and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0056] In the description of the present invention, it should be understood that the orientation or positional relationship indicated by the terms "length", "width", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. is based on the orientation or positional relationship shown in the drawings, and is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus should not be construed as a limitation to the present invention. In addition, in the description of the present invention, the meaning of "a plurality of" is two or more, unless otherwise specifically defined.

[0057] Please refer to Figure 1 , the cloud-based computer network resource management system includes:

[0058] The network bandwidth demand analysis module collects the bandwidth usage records and real-time network traffic data in the cloud platform computer network tasks, converts the bandwidth usage records and real-time network traffic data into time series, analyzes the peak changes of the time series data of bandwidth usage and network traffic, sorts out the distribution of task bandwidth demands in the computer network nodes, marks the bandwidth demand areas exceeding the preset frequency threshold with reference to the peak changes and the distribution of task bandwidth demands, integrates the marked bandwidth demand areas, and generates network bandwidth distribution characteristic data;

[0059] The task flow dependency scheduling module analyzes the traffic interaction characteristics and dependency relationships among the computer network tasks based on the network bandwidth distribution characteristic data, sorts out the input and output traffic of the task nodes according to the traffic interaction characteristics and dependency relationships, determines the sequential dependency order of the computer network task flow, adjusts the execution sequence of the task nodes, marks the dependency link relationships among the task nodes according to the sequential dependency order, and at the same time combines the network resource requirements of the task nodes to generate an optimized execution sequence of the task flow;

[0060] The resource isolation configuration module analyzes the computing resources, storage resources, and bandwidth resource requirements of the task nodes in the computer network in the optimized execution sequence of the task flow, selects the resource boundaries of the task nodes according to the computing resources, storage resources, and bandwidth resource requirements of the task nodes, starting from the resource boundaries, adjusts the division of the virtual resource pool in the network according to the task priorities and bandwidth occupations, plans the resource usage ranges of each task node in the virtual resource pool, and generates a computer network resource allocation list;

[0061] The bandwidth dynamic optimization module applies the computer network resource allocation list, monitors the traffic usage status of task nodes in the computer network in real time, analyzes the bandwidth occupancy change rate and latency according to the traffic usage status, divides the bandwidth occupancy change rate and latency according to time windows, predicts the bandwidth demand trend within the time window, dynamically adjusts the bandwidth allocation ratio of task nodes in the network according to the predicted bandwidth demand, optimizes the traffic path planning and resource allocation, marks the network bandwidth allocation status, and generates the computer network bandwidth resource optimization result;

[0062] The network bandwidth distribution characteristic data includes the traffic peak change range, the bandwidth demand frequency distribution, and the bandwidth demand area exceeding the preset frequency threshold; the task flow optimization execution sequence specifically refers to the input-output relationship of task nodes, the dependency link annotation of task nodes, and the optimized execution order of task nodes; the computer network resource allocation list includes the computing resource range, storage resource range, and bandwidth resource usage range of task nodes; the computer network bandwidth resource optimization result is specifically the bandwidth allocation ratio planning, traffic path optimization rules, and network bandwidth allocation status annotation.

[0063] Please refer to Figure 2 , and the steps for analyzing the peak change of the time series data of bandwidth usage are specifically as follows:

[0064] Collect the bandwidth usage records and real-time network traffic data in the cloud platform computer network tasks, classify the bandwidth usage records in chronological order and extract relevant values, collect and store the real-time network traffic data, adjust and handle outliers for the bandwidth usage records and real-time network traffic, and obtain the bandwidth and traffic time series data;

[0065] First, collect the usage records of the task bandwidth. Organize the bandwidth usage into daily and hourly traffic data according to the time series. For example, the bandwidth usage record of a certain network node shows that the traffic peak on the first day is 100 Mbps, and the traffic peak on the second day is 120 Mbps, with a traffic growth rate of 20%. Extract the time series changes by calculating the daily traffic peak change rate; record the traffic changes of each node in real time. For example, during the peak hours of a day (such as from 10:00 to 12:00), the bandwidth usage increases from 60 Mbps to 110 Mbps, and then drops back to 50 Mbps during the afternoon low peak hours (such as from 15:00 to 17:00). Eliminate the traffic records with breakpoints and outliers, and use data cleaning methods to segment and classify the daily data. Store the cleaned time series data as a distribution table of daily peak traffic, traffic growth rate, and total traffic. Through the analysis of the cleaned data, extract the time series change information of the bandwidth usage in each time period, and divide the task bandwidth demand trend of each cycle according to the time interval. For example, the bandwidth demand of a certain task is stable between 100 Mbps and 120 Mbps during the traffic peak hours, while the demand fluctuation range during the low peak hours is 30 Mbps to 50 Mbps. Integrate the time series data of the traffic of each node to generate the bandwidth usage time series data.

[0066] Based on the bandwidth and traffic time series data, use the formula:

[0067]

[0068] Calculate the fluctuation index P of the i-th time series (such as the bandwidth usage time series or network traffic time series) i ;

[0069] where S t,i is the value at time t in the i-th time series (in units of Mbps, such as bandwidth usage or network traffic), collected by a network traffic monitoring device (such as the NetFlow tool) at time points, max(S t,i ) is the maximum value of the i-th time series, representing the peak in the series, directly extracted from the series data using a data analysis tool (such as the max() function of Pandas), min(S t,i ) is the minimum value of the i-th time series, representing the lowest value in the series, directly extracted from the series data using a data analysis tool (such as the min() function of Pandas), S t,i,j is the value of the j-th data point in the time series S t,i , obtained from the original data recorded by the traffic monitoring device, is the mean of the time series S t,i , representing the average level of the series data, and m is the time series S t,iThe total number of data points in the data. Determine the number of sampling points through a data recording tool. For example, record once per second. It is for the time series S t,i Calculate the sum of the squared deviations from the mean for each data point in the time series S.

[0070] If the bandwidth usage time series is: S t,1 = [20, 35, 50, 65, 80, 90] (unit: Mbps), and the network traffic time series: S t,2 = [10, 25, 40, 55, 70, 85] (unit: Mbps). For the calculation of the time series S t,1 Calculate the maximum and minimum values: max(S t,1 ) = 90, min(S t,1 ) = 20.

[0071] Calculate the peak difference: max(S t,1 ) - min(S t,1 ) = 90 - 20 = 70;

[0072] Calculate the mean:

[0073] Calculate the dispersion factor:

[0074] Substitute into the formula to calculate P 1 :

[0075] For the calculation of the time series S t,2 Calculate the maximum and minimum values: max(S t,2 ) = 85, min(S t,2 ) = 10. Calculate the peak difference: max(S t,2 ) - min(S t,2 ) = 85 - 10 = 75

[0076] Calculate the mean:

[0077] Calculate the dispersion factor:

[0078] Substitute into the formula to calculate P 2 :

[0079] The results show that the fluctuation index of the bandwidth usage time series is P 1 ≈1.17, and the fluctuation index of the network traffic time series is P 2 ≈1.20. The fluctuation range of network traffic is slightly larger than that of bandwidth usage.

[0080] Please refer to Figure 3 , the steps for obtaining the network bandwidth distribution characteristic data are specifically as follows:

[0081] Based on the fluctuation indices corresponding to the bandwidth usage time series and the network traffic time series, analyze the bandwidth demand distribution characteristics of network nodes, label the time periods when the peak bandwidth demand of nodes exceeds the limit as over-limit regions, and generate data on nodes with over-limit bandwidth demands;

[0082] Sort out the distribution of task bandwidth demands in computer network nodes. According to the calculation results that the fluctuation index of the bandwidth usage time series is approximately 1.17 and the fluctuation index of the network traffic time series is approximately 1.20, analyze the bandwidth demand distribution characteristics of each node. Divide the data of all network nodes into several time periods according to the time series. Record the bandwidth usage and traffic peaks within each time period. By comparing the above-mentioned fluctuation indices with the historical reference values (for example, the historical fluctuation index is 1.10), it is found that the current fluctuation range of bandwidth usage is slightly higher than the reference value, while the fluctuation range of network traffic is much higher than the reference value, indicating that there may be a rapid increase in bandwidth demands for some nodes in the network. Combining these data, analyze the nodes that exceed the historical reference value, count the changes in the peak bandwidth demands of each node within a specific time period, and record the proportion of time periods when the bandwidth demand exceeds the threshold. Label the nodes with a frequency exceeding 30% as nodes with high-frequency over-limit bandwidth demands. Further sort out the over-limit situations of each node, including the over-limit proportion of time periods, peak bandwidth values, and distribution characteristics. By comparing with other nodes in the same task, it is found that the nodes with higher fluctuation indices have a more significant over-limit frequency during peak periods. For example, the bandwidth demand of a node exceeds the historical reference value in 4 out of 10 time periods, and the over-limit frequency is 40%, which is higher than the threshold, so it is labeled as a high-frequency over-limit node. The final distribution characteristic data is integrated into a multi-dimensional array, including the identifier, over-limit frequency, peak bandwidth value, and time period characteristics of each node.

[0083] Based on the data of nodes with over-limit bandwidth demands, integrate the over-limit frequency, time period characteristics, and peak bandwidth values, map the data onto the time axis uniformly, and count the characteristics of the over-limit regions to obtain the network bandwidth distribution characteristic data;

[0084] Integrate the labeled bandwidth demand regions. According to the labeled data of nodes with over-limit bandwidth demands, sort out the over-limit frequency, time period characteristics, and peak bandwidth values uniformly according to the time series, map the bandwidth demand data of different nodes onto the same time axis to form a unified time series distribution, count the number of over-limit nodes and the bandwidth peaks within each time period, and extract the main characteristics of the over-limit regions, including the number distribution of over-limit nodes, the peak bandwidth demand within the time period, and the over-limit duration, and finally generate the network bandwidth distribution characteristic data.

[0085] Please refer toFigure 4 , the steps to determine the sequential dependency order of computer network task flows are specifically as follows:

[0086] Based on the network bandwidth distribution characteristic data, analyze the traffic interaction characteristics and dependency relationships between computer network tasks, and use the formula:

[0087]

[0088] Calculate the traffic interaction correlation index D between task i' and task j' i′j′ , indicating the strength of the dependency relationship between the two tasks in the time series;

[0089] Among them, F i′,k is the traffic value of task i' within the time period k, with the unit of Mbps, collected by the network traffic monitoring device, and the data is recorded by task identifier and time segment. F j′,k is the traffic value of task j' within the time period k, with the unit of Mbps, also collected by the traffic monitoring device, similar to F i′,k , but belonging to different tasks. n is the number of time periods, representing the total number of sampling points during the monitoring period, calculated from the sampling frequency of the task flow monitoring device and the total monitoring duration. represents the summation of the traffic interaction values within all time periods, associated with the time period index represented by k. and respectively represent the square root of the sum of the squares of the traffic values of task i' and task j' within all time periods, reflecting the overall distribution characteristics of the task traffic intensity.

[0090] If the total number of time periods n = 6. Traffic values: F i′,k = [10, 15, 20, 25, 30, 35] Mbps, F j′,k =

[0091] [5, 10, 15, 20, 25, 30] Mbps

[0092] Calculate the numerator part:

[0093] Calculate the denominator part:

[0094] The sum of the squares of the traffic of task j':

[0095] Calculate the denominator:

[0096] Calculate the correlation index:

[0097] The results show that the traffic interaction correlation index between task i′ and task j′ is 0.996.

[0098] Based on the traffic interaction correlation index of each task, sort the input and output traffic of the task nodes, analyze the traffic distribution and interaction order, and sort the dependency relationships of the tasks according to the traffic distribution and interaction order to obtain the sequential dependency sequence of the task flow;

[0099] Since the traffic interaction correlation index between task i′ and j′ is 0.996, which is close to 1, it indicates that there is a very high dependency relationship between the two tasks. The determination range of the correlation index is usually divided into weak correlation (0.0–0.3), medium correlation (0.3–0.7), strong correlation (0.7–0.9), and extremely strong correlation (0.9–1.0). Therefore, the correlation index between task i′ and j′ belongs to the extremely strong correlation range, clearly indicating that the matching of the dependency relationship needs to be considered first between the two tasks. According to the above traffic interaction characteristics and dependency relationships, sort the input and output traffic of the task nodes. First, extract the input and output traffic data of task i′ and j′ in all time periods, analyze the output traffic of task i′ as the input traffic of task j′, record the distribution and change trend of its traffic value, compare the traffic interaction of other task nodes, screen out the connections with relatively low interaction intensity between task nodes for optimization, and reassign the connections with relatively small input traffic to tasks with higher correlation. By determining the dependency order, arrange task i′ at the upstream node of the task flow and task j′ at its downstream node, and combine the traffic distribution characteristics of other task nodes to plan the execution order of the overall task flow in sequence, and finally construct the optimized dependency sequence of the task flow.

[0100] Please refer to Figure 5 , and the specific steps for obtaining the optimized execution sequence of the task flow are as follows:

[0101] Based on the sequential dependency order of the sequential dependency sequence of the task flow, label the dependency link relationships between task nodes one by one, sort out the traffic interaction intensity of all task nodes and compare the correlations, label the direct dependency links and indirect dependency links, and generate a task node relationship diagram;

[0102] Based on the precedence dependency order of the optimized dependency sequence, the dependency link relationships between task nodes are marked according to the precedence of the tasks. Based on the extremely strong correlation between task i' and j', the output traffic of task i' is clearly marked as the input traffic of task j'. According to the task priorities in the optimized dependency sequence, the dependency relationships between all task nodes are sorted one by one, the traffic interaction intensities of each node are compared, and the task pairs with a correlation index higher than 0.9 are preferentially marked as direct dependency link relationships, forming a task node relationship diagram containing dependency link markings. At the same time, the node pairs that do not reach the direct dependency strength are secondarily screened and marked as indirect dependency relationship links, and finally, the marking of all dependency link relationships between task nodes is completed.

[0103] Based on the dependency link relationships in the task node relationship diagram and the network resource requirements of the task nodes, the bandwidth requirements and traffic interaction intensities of the task nodes are sorted, and the execution order of the task nodes is adjusted according to the resource allocation priorities to generate an optimized execution sequence of the task flow;

[0104] Sort out the network bandwidth requirements, traffic interaction intensities, and dependency relationship links of each task node, sort the task nodes according to the priorities of network resource requirements, preferentially allocate task nodes with high traffic and high bandwidth requirements to earlier positions in the execution sequence, and at the same time adjust the node order in the dependency link relationships to ensure that the resource requirements of downstream nodes are met. Generate an optimized execution sequence of the task flow according to the execution priorities of the task nodes and the network resource allocation strategy, and record the execution order, resource requirements, and traffic distribution of each task node in tabular form, and finally complete the planning of the optimized execution sequence of the task flow.

[0105] Please refer to Figure 6 , and the steps for determining the resource boundaries of the selected task nodes are specifically as follows:

[0106] Based on the optimized execution sequence of the task flow, extract the resource requirements, storage resource requirements, and bandwidth resource requirements from the operation monitoring data of the nodes, classify and judge the resource requirement levels of each node according to the preset range, and sort out the resource requirements to generate the resource requirement characteristic data of the task nodes;

[0107] In the analysis of the optimized execution sequence of the task flow, the computing resources, storage resources, and bandwidth resource requirements of task nodes in the computer network are determined. According to the specific content of the optimized execution sequence of the task flow, the resource requirement characteristics of each task node are extracted one by one. Through the node operation monitoring data, the computing resource requirements, storage resource requirements, and bandwidth resource requirement data of the task nodes are obtained respectively. Suppose the resource requirement data of node A and node B are as follows: Node A: The computing resource requirement is 8 cores, the storage resource requirement is 64 GB, and the bandwidth resource requirement is 120 Mbps. Node B: The computing resource requirement is 16 cores, the storage resource requirement is 32 GB, and the bandwidth resource requirement is 60 Mbps. Judgment of computing resource requirements: According to the preset range, the computing resource requirements are divided into low demand (0–8 cores), medium demand (8–16 cores), and high demand (16 cores and above). The computing resource requirement of node A is 8 cores, which belongs to the upper limit of low demand, while the computing resource requirement of node B is 16 cores, which belongs to the upper limit of medium demand. Judgment of storage resource requirements: Referring to the division of the storage resource requirement range, the low demand is 0–32 GB, the medium demand is 32–64 GB, and the high demand is 64 GB and above. The storage resource requirement of node A is 64 GB, which belongs to the upper limit of medium demand, while the storage resource requirement of node B is 32 GB, which belongs to the upper limit of low demand. Judgment of bandwidth resource requirements: According to the bandwidth resource requirement division rule, the low demand is 0–50 Mbps, the medium demand is 50–100 Mbps, and the high demand is 100 Mbps and above. The bandwidth resource requirement of node A is 120 Mbps, which belongs to high demand, while the bandwidth resource requirement of node B is 60 Mbps, which belongs to medium demand.

[0108] Based on the resource requirement characteristic data of the task node, the formula is adopted;

[0109]

[0110] Calculate the comprehensive resource weight value V of task node a a , quantify the overall level of node resource requirements, and select the resource boundary of the task node according to the overall level of node resource requirements;

[0111] Among them, C a is the computing resource requirement of task node a, with the unit of CPU cores, indicating the number of processors required for the node to execute the computing task, which is obtained by means of node operation load monitoring. For example, record the peak computing requirement during the load, and extract the maximum resource requirement as the parameter value. H a is the storage resource requirement of task node a, with the unit of GB, indicating the storage capacity required for the node to complete the task, which is obtained by analyzing the node storage occupancy. For example, record the maximum storage space usage value during the task process, and organize it in combination with historical task data to obtain, B ais the bandwidth resource requirement of task node a, with the unit of Mbps, representing the network bandwidth required during the task data interaction process, which is obtained through a real-time traffic monitoring tool. For example, the maximum bandwidth requirement value is extracted during the peak period of task operation, w c and w h and w b represent the weight parameters of computing resources, storage resources, and bandwidth resources respectively, reflecting the degree of emphasis of the task on the three resources, which are set according to the task type or scenario. For example, a computationally intensive task can be assigned a higher w c , while for a data-intensive task, the weight of w h is increased. The computing resource weight is positively correlated with the computational intensity of the task node. The higher the computational intensity of the task, the greater the required CPU usage rate, and the higher the weight w c . The setting range is: w c ∈[0.4, 0.6]. The upper limit is taken for computationally intensive tasks, and a lower value is taken for other tasks. The storage resource weight is positively correlated with the storage demand of the task node. The larger the amount of data that the task needs to store, the higher the weight w h . For tasks such as data storage, video processing, and large-scale log processing, the setting range is: w h ∈[0.2, 0.4]. The upper limit is taken for data-intensive tasks, and a lower value is taken for compute-priority tasks. The bandwidth resource weight is positively correlated with the frequency of network data transmission of the task node. The more bandwidth resources the task requires, the higher the weight w b . The setting range is: w b ∈[0.2, 0.3]. The upper limit is taken for tasks with high real-time requirements, and a lower value is taken for storage-priority tasks. w c + w h + w b is the sum of the resource weight parameters, ensuring the correct normalization of the relative proportions of each weight.

[0112] If the current computing resource requirement C a = 8 cores; the storage resource requirement H a = 64 GB; the bandwidth resource requirement B a = 120 Mbps; the weight parameter w c = 0.5, w h = 0.3, w b = 0.2.

[0113] Calculate the comprehensive resource weight value:

[0114] This result shows that the comprehensive resource weight value V of task node a a= 47.2, indicating the comprehensive resource demand of this node. By comparing with the preset resource demand division threshold, the resource boundary of the task node is determined. First, in combination with the historical baseline value, the weight values are divided into different resource demand levels, such as the low demand range, the medium demand range, and the high demand range. For nodes with the comprehensive resource weight value falling within the low demand range, the minimum resource boundary is allocated to only meet the basic task execution requirements; for nodes with the weight value in the medium demand range, a moderate resource boundary is allocated to ensure support for resource usage fluctuations during task execution; for nodes with the weight value reaching or exceeding the high demand range, a larger resource boundary is allocated to prioritize ensuring their computing, storage, and bandwidth requirements. At the same time, the allocation is adjusted according to the priority and dependency of the actual tasks executed by the nodes, and finally the division of the virtual resource pool is determined by combining the resource boundaries of each node.

[0115] Please refer to Figure 7 , the steps for obtaining the computer network resource allocation list are specifically as follows:

[0116] Based on the resource boundaries of the task nodes, the task nodes are divided into multiple groups according to the priority. By obtaining the peak and average bandwidth occupancy of each node, according to the peak and average bandwidth occupancy, the bandwidth resources are allocated according to the node priority to generate the resource allocation result of the task nodes in the virtual resource pool;

[0117] According to the comprehensive resource weight value V of the task node a and the resource boundary, the preset range is: nodes with a weight value lower than 30 are classified into the low priority group, nodes with a weight value between 30 and 50 are classified into the medium priority group, and nodes with a weight value higher than 50 are classified into the high priority group. For the task node with a weight value of 47.2, it is in the medium priority range, and the basic requirements for computing resources, storage resources, and bandwidth resources need to be met in the allocated virtual resource pool. By calculating the actual peak bandwidth usage of this node as 300 Mbps and the average value as 250 Mbps, 350 Mbps of bandwidth is reserved in combination with the virtual resource pool capacity as the maximum allocation value for this node. In the high priority group, for example, the task node with a weight value of 52 has a computing resource requirement of 16 cores and a bandwidth peak of 500 Mbps, exceeding the bandwidth range of the medium priority. Therefore, more bandwidth resources are preferentially allocated and the bandwidth allocation value of the low priority task nodes is reduced to ensure that the high priority tasks complete resource configuration. For the task nodes in the low priority group, according to the remaining resource range, the peak bandwidth that can be allocated to each node is calculated. For example, the node with a weight value of 28 can be allocated a maximum of 120 Mbps of bandwidth, and the storage resource limit is 64 GB. Through grouping and calculation, the resource usage range planning of each task node in the virtual resource pool is completed.

[0118] Based on the resource allocation results of task nodes in the virtual resource pool, record the allocation of computing resources, storage resources, and bandwidth resources according to the task node priorities, organize the virtual resource pool identifiers to which each node belongs, review and adjust the resource allocation results, and generate a computer network resource allocation list;

[0119] Record the resource allocation of each node in order from high to low according to the task node priorities. The preset format of the resource allocation list includes: node identifier, computing resources (number of CPU cores), storage resources (GB), bandwidth resources (Mbps), and the identifier of the corresponding virtual resource pool. Taking a high-priority node as an example, the allocation result of a node with a weight value of 52 is 16-core CPU, 256GB storage, and 500Mbps bandwidth, and mark the virtual resource pool it belongs to as A; for a medium-priority node, the allocation result of a node with a weight value of 47.2 is 10-core CPU, 128GB storage, and 350Mbps bandwidth, and mark the virtual resource pool it belongs to as B; for a low-priority node such as a node with a weight value of 28, the allocation result is 4-core CPU, 64GB storage, and 120Mbps bandwidth, and mark the virtual resource pool it belongs to as C. Review the resource allocation of each node in the list to determine whether there is a situation exceeding the resource pool capacity. For example, if the total capacity of resource pool A is 1000Mbps bandwidth, if the total bandwidth of all allocated nodes exceeds 1000Mbps, re-adjust the bandwidth allocation order and ratio between nodes, and finally complete a conflict-free computer network resource allocation list.

[0120] Please refer to Figure 8 , and the specific steps for obtaining the optimization result of computer network bandwidth resources are as follows:

[0121] Based on the computer network resource allocation list, monitor the traffic usage status of task nodes in real time, record the traffic values and latency data in each time period through a bandwidth monitoring device, divide the bandwidth occupancy change rate and latency situation according to time windows, and organize the data to generate time series data of task node traffic usage;

[0122] Collect the traffic usage data of task nodes in different time periods, record the traffic values per second through a bandwidth monitoring device, including upload and download rates, and extract the maximum traffic value, minimum traffic value, and average traffic value of task nodes within a fixed time window (for example, every 5 minutes) to obtain the specific value of the bandwidth occupancy change range. At the same time, measure the latency data between nodes multiple times within the time window through a network latency measurement tool, and statistically calculate the average value and standard deviation of the latency. Organize the collected bandwidth and latency data into time series data in units of time windows, and use a visualization tool to generate a change trend graph for each time window, and finally complete the time window division and data organization of the bandwidth occupancy change rate and latency situation.

[0123] Based on the time - series data of task - node traffic usage, use the formula

[0124]

[0125] to calculate the predicted bandwidth demand value B of the time window t′ + 1 t′+1 (unit: Mbps), which represents the bandwidth demand of the next time window, and obtain the bandwidth demand prediction result;

[0126] where B t′ is the bandwidth demand value within the time window t′ (unit: Mbps), representing the total bandwidth demand in the current time window. Through the bandwidth monitoring device, the traffic usage within the time window is statistically analyzed to obtain the maximum traffic value, minimum traffic value, and average traffic value to quantify the bandwidth demand. Through the bandwidth monitoring data of a 5 - minute time window, the maximum traffic is recorded as 520 Mbps, the minimum traffic is 480 Mbps, and after calculating the average traffic, B t′ = 500 Mbps, R t′ is the bandwidth occupancy change rate of the time window t′, representing the change amplitude of the bandwidth demand within the current time window, and is obtained through the formula where B max 、B min 、B avg are the maximum traffic value, minimum traffic value, and average traffic value within the time window respectively, is the average value of the bandwidth occupancy change rate of the time window, used as a reference value to evaluate the deviation of the current bandwidth change rate. Through the statistical analysis of historical data of multiple time windows, the average value of the bandwidth occupancy change rate is calculated. By calculating the average value of the bandwidth change rate data of the past 10 time windows (such as 0.06, 0.08, 0.07,...), γ is an adjustment coefficient used to adjust the sensitivity of the bandwidth demand prediction to the change rate deviation, and is set according to the fluctuation characteristics of historical network traffic. For example, a lower value (such as 0.1 - 0.2) is selected when the network fluctuation is small, and a higher value (such as 0.3 - 0.5) is selected when the fluctuation is large.

[0127] If the bandwidth monitoring device records that the maximum traffic B max = 520 Mbps, the minimum traffic B min = 480 Mbps, the average traffic B avg = 500 Mbps, and the average value of the historical time - window bandwidth change rate The adjustment coefficient γ = 0.3, which is determined by the historical traffic fluctuation characteristics.

[0128] Calculate the bandwidth occupancy change rate:

[0129] Calculate the predicted value of the bandwidth requirement for the next time window:

[0130] This result indicates that in the time window t′ + 1, the predicted bandwidth requirement is 500.003 Mbps.

[0131] Based on the bandwidth requirement prediction result, dynamically adjust the bandwidth allocation ratio of task nodes, analyze the delay and occupancy of the traffic path, optimize the path planning, adjust the traffic allocation rules of task nodes, and record the data to generate the optimization result of computer network bandwidth resources;

[0132] Dynamically adjust the bandwidth allocation ratio of task nodes in the network according to the predicted bandwidth requirement. First, based on the predicted bandwidth requirement value B of the time window t′ + 1 calculated in Paragraph 2 t′+1 = 500.003 Mbps, compare it with the current bandwidth allocation situation to determine the change range of the bandwidth requirement. For task nodes with an increasing bandwidth requirement, for example, the currently allocated bandwidth is 480 Mbps and the predicted bandwidth requirement is higher than the current value. Calculate the amount of bandwidth that needs to be increased as 500.003 - 480 = 20.003 Mbps. When reallocating, first adjust the bandwidth occupancy of other low-priority task nodes, such as reducing the bandwidth allocation of nodes with lower bandwidth requirements from 120 Mbps to 100 Mbps, and allocate the released bandwidth to high-demand nodes. Next, optimize the traffic path planning. By analyzing the delay and occupancy of the current traffic path, select a low-latency path to adjust the traffic routing of high-demand nodes. For example, switch the current path with a delay of 50 ms to an alternative path with a delay of 20 ms to ensure the traffic transmission efficiency of high-demand nodes. Finally, label the adjusted network bandwidth allocation status, record the bandwidth allocation value of each task node, the change amount before and after adjustment, and the corresponding traffic path, and integrate the allocation data of all nodes into the optimization result of network bandwidth resources for subsequent monitoring and adjustment.

[0133] The above is only a preferred embodiment of the present invention, and it does not limit the present invention in other forms. Any person skilled in the art may use the disclosed technical content to make changes or modifications into equivalent embodiments with equivalent changes and apply them to other fields. However, as long as it does not depart from the technical solution content of the present invention, any simple modification, equivalent change, and modification made to the above embodiments based on the technical essence of the present invention still fall within the protection scope of the technical solution of the present invention.

Claims

1. A cloud-based computer network resource management system, characterized in that: The system comprises: The network bandwidth demand analysis module collects bandwidth usage records and real-time network traffic data in the cloud platform computer network tasks, analyzes the peak changes in the time series data of bandwidth usage and network traffic, marks the bandwidth demand areas that exceed the preset frequency threshold with reference to the peak changes and the distribution of task bandwidth demand, and generates network bandwidth distribution characteristic data; The task flow dependency scheduling module analyzes the traffic interaction characteristics and dependency relationships between computer network tasks based on the network bandwidth distribution characteristic data, determines the priority dependency sequence of computer network task flows, adjusts the execution sequence of task nodes, and generates a task flow optimization execution sequence; The resource isolation configuration module analyzes the computing resources, storage resources and bandwidth resource requirements of the task nodes in the computer network in the task flow optimization execution sequence, selects the resource boundary of the task node, and adjusts the virtual resource pool division in the network according to the task priority and bandwidth occupancy based on the resource boundary, and generates a computer network resource allocation list; The bandwidth dynamic optimization module applies the computer network resource allocation list, analyzes the bandwidth occupancy change rate and delay according to the traffic usage status, divides the bandwidth occupancy change rate and delay according to the time window, predicts the bandwidth demand trend within the time window, and generates the computer network bandwidth resource optimization result.

2. The cloud-based computer network resource management system according to claim 1, characterized in that: The steps of analyzing the peak value change of the time series data of bandwidth usage are specifically as follows: Collect bandwidth usage records and real-time network traffic data in cloud platform computer network tasks, classify bandwidth usage records in chronological order and extract corresponding values, collect and store real-time network traffic data, adjust bandwidth usage records and real-time network traffic and process outliers to obtain bandwidth and traffic time series data; Based on the bandwidth and traffic time series data, the formula is used: Calculate the volatility index P of the i-th time series i ; Among them, S t,i is the value of the ith time series at time t, max(S t,i ) is the maximum value of the i-th time series, min(S t,i ) is the minimum value of the i-th time series, S t,i,j is the time series S t,i The value of the jth data point in is the time series S t,i The mean of the time series S t,i The total number of data points in .

3. The cloud-based computer network resource management system according to claim 2, characterized in that: The steps for acquiring the network bandwidth distribution characteristic data are specifically as follows: Based on the fluctuation index corresponding to the bandwidth usage time series and the network traffic time series, the bandwidth demand distribution characteristics of the network nodes are analyzed, the time period when the node bandwidth demand peak exceeds the limit is marked as an over-limit area, and bandwidth demand over-limit node data is generated; Based on the bandwidth demand exceeding node data, the exceeding frequency, time period characteristics and peak bandwidth value are integrated, the data is uniformly mapped to the time axis and the characteristics of the exceeding area are counted to obtain the network bandwidth distribution characteristic data.

4. The cloud-based computer network resource management system according to claim 3, characterized in that: The step of determining the order of dependency of the computer network task flow is specifically as follows: Based on the network bandwidth distribution characteristic data, the traffic interaction characteristics and dependencies between computer network tasks are analyzed, and the formula is adopted: Calculate the traffic interaction correlation index D between task i′ and task j′ i′j′ ; Among them, F i′,k is the flow value of task i′ in time period k, F j′,k is the flow value of task j′ in time period k, and n is the number of time periods; According to the traffic interaction correlation index of each task, the input and output traffic of the task node is sorted, the traffic distribution and interaction sequence are analyzed, and the dependencies of the tasks are sorted according to the traffic distribution and interaction sequence to obtain the sequential dependency sequence of the task flow.

5. The cloud-based computer network resource management system according to claim 4, characterized in that: The steps for obtaining the task flow optimization execution sequence are specifically as follows: Based on the sequential dependency order of the sequential dependency sequence of the task flow, the dependency link relationships between the task nodes are marked one by one, the traffic interaction intensity of all task nodes is sorted out and the correlation is compared, the direct dependency links and the indirect dependency links are marked, and a task node relationship diagram is generated; Based on the dependent link relationships in the task node relationship graph and the network resource requirements of the task nodes, the bandwidth requirements and traffic interaction intensity of the task nodes are sorted, the execution order of the task nodes is adjusted according to the resource allocation priority, and a task flow optimization execution sequence is generated.

6. The cloud-based computer network resource management system according to claim 5, characterized in that: The steps of selecting the resource boundary of the task node are specifically as follows: Based on the task flow optimization execution sequence, resource requirements, storage resource requirements and bandwidth resource requirements are extracted from the operation monitoring data of the node, the resource requirement level of each node is classified and judged according to a preset range, and the resource requirements are sorted to generate resource requirement characteristic data of the task node; Based on the resource demand characteristic data of the task node, a formula is adopted; Calculate the resource weight value V of task node a a , quantify the overall level of node resource demand, and select the resource boundary of the task node according to the overall level of node resource demand; Among them, C a is the computing resource requirement of task node a, H a is the storage resource requirement of task node a, B a is the bandwidth resource requirement of task node a, w c 、w h 、w b They represent the weight parameters of computing resources, storage resources, and bandwidth resources respectively.

7. The cloud-based computer network resource management system according to claim 6, characterized in that: The steps of obtaining the computer network resource allocation list are specifically as follows: Based on the resource boundaries of the task nodes, the task nodes are divided into multiple groups according to priority, and bandwidth resources are allocated according to the node priority by obtaining the bandwidth occupancy peak value and average value of each node, and the resource allocation results of the task nodes in the virtual resource pool are generated according to the bandwidth occupancy peak value and average value; Based on the resource allocation results of the task nodes in the virtual resource pool, the allocation of computing resources, storage resources and bandwidth resources is recorded according to the task node priority, and the virtual resource pool identifier to which each node belongs is sorted, the resource allocation results are reviewed and adjusted, and a computer network resource allocation list is generated.

8. The cloud-based computer network resource management system according to claim 7, characterized in that: The steps for obtaining the computer network bandwidth resource optimization result are specifically as follows: Based on the computer network resource allocation list, the flow usage status of the task node is monitored in real time, the flow value and delay data of each time period are recorded through the bandwidth monitoring device, the bandwidth occupancy change rate and delay situation are divided according to the time window, and the data is sorted to generate the time series data of the flow usage of the task node; Based on the time series data of the task node traffic usage, the formula is used Calculate the predicted bandwidth demand value B for time window t′+1 t′+1 , get the bandwidth demand prediction result; Among them, B t′ is the bandwidth requirement value in the time window t′, R t′ is the bandwidth occupancy change rate of time window t′, indicating the change amplitude of bandwidth demand in the current time window. is the average value of the bandwidth occupancy change rate in the time window, and γ is the adjustment coefficient; Based on the bandwidth demand prediction results, the bandwidth allocation ratio of the task node is dynamically adjusted, the delay and occupancy of the traffic path are analyzed, the path planning is optimized, the traffic allocation rules of the task node are adjusted, and data is recorded to generate computer network bandwidth resource optimization results.

Images