A multi-task processing method and system in an algorithm network fusion environment

Abstract

The present application relates to the technical field of task processing, in particular to a multi-task processing method and system in an algorithm network fusion environment, comprising the following steps: obtaining CPU peak value and memory retention time in a three-period task, calculating resource concentration, generating a load compact record, judging whether the concentration sequence is increasing and the threshold value task is in the queue, combining the node peak interval fluctuation to determine the schedulable window, calculating the resource adaptation ratio to generate the matching mapping, eliminating the high latency link, and completing the task scheduling.In the present application, the CPU occupancy rate peak value and memory duration of the task in the continuous three resource allocation periods are quantitatively analyzed, the high access task with the resource demand tending to increase is accurately identified, the adaptation degree between the task demand and the node resource is verified in real time by combining the node resource adaptation ratio analysis, the unstable link is timely eliminated by constructing the round-trip delay variation ratio sequence and calculating the continuous period difference, and more reliable and efficient task data transmission is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of task processing, in particular to a multi-task processing method and system in an algorithm-network fusion environment. BACKGROUND

[0002] The technical field of task processing mainly involves the scheduling, allocation, execution and resource coordination of multiple processing tasks in a computing system or network environment, with the core goal of improving task execution efficiency, shortening response time, improving system throughput and resource utilization. This field covers task modeling, task division, priority allocation, scheduling strategy, task migration mechanism, and fault tolerance and load balancing solutions. Typical applications include distributed systems, cloud computing platforms, high-performance computing systems, and edge computing environments. Task processing also needs to consider the dynamic and heterogeneous nature of the system, and then develop matching scheduling and resource allocation strategies to achieve multi-task parallel processing and optimal resource allocation.

[0003] Among them, the multi-task processing method in the algorithm-network fusion environment refers to a scheme for efficient allocation and execution of multiple processing tasks in a computing and network resource deep fusion architecture. The purpose of this topic is to coordinate computing power nodes and network resources, achieve optimal scheduling and execution path selection of tasks in a distributed, multi-level computing power system, and improve the efficiency and reliability of overall business processing. The method is suitable for large-scale task processing scenarios that require algorithm-network collaborative support, such as multi-access edge computing, intelligent manufacturing, and smart cities.

[0004] Traditional processing methods only develop scheduling and resource allocation strategies based on static task information or isolated period resource usage, without effectively identifying task resource demand trends and node resource fluctuation period characteristics, resulting in frequent occurrences of task resource demand growth trends and node resource allocation strategies being out of sync, causing resource idling or task overload, lack of fine-grained resource adaptation analysis in task and node resource matching, and existence of task assignment to unsuitable nodes, exacerbating resource waste and low execution efficiency. Static detection and evaluation of link transmission delay lack periodic dynamic change monitoring mechanism, causing link delay fluctuations to exceed expectations without timely detection, seriously affecting task data transmission efficiency and reliability, such as frequent occurrence of task data delay or transmission interruption in intelligent manufacturing or edge computing, reducing overall processing performance. SUMMARY

[0005] The purpose of the present application is to solve the shortcomings in the prior art and to propose a multi-task processing method and system in an algorithm-network fusion environment.

[0006] In order to achieve the above-mentioned purpose, the present application adopts the following technical scheme: a multi-task processing method in an algorithm-network fusion environment, comprising the following steps:

[0007] S1: Obtain CPU usage rate, memory occupation rate and task scheduling log of the registered task in the algorithm network fusion architecture in three continuous resource allocation periods, calculate resource usage concentration of the task in each period, and generate a period load compact record set;

[0008] S2: Based on the period load compact record set, according to the resource usage concentration sequence of the task, the length of the monotonically increasing segment of the sequence is calculated, and is compared with the set resource density threshold value, if the increasing in three continuous periods and exceeding the threshold value are met, the task is listed in the high access tendency, and a priority in list task list is obtained;

[0009] S3: Based on the priority in list task list, the CPU peak period in each period is extracted, the interval between adjacent two peaks of the node is calculated, the fluctuation degree of the sequence is calculated, whether the fluctuation is lower than the set stability threshold value is judged, and a node scheduling adaptive segment list is generated;

[0010] S4: Based on the node scheduling adaptive segment list, according to the task request parameter, the resource adaptation ratio is calculated, if the resource adaptation ratio is greater than the set resource adaptation threshold value, it is marked as matching success, and a multi-task matching node mapping table is obtained.

[0011] As a further scheme of the application, the period load compact record set comprises a task period resource scheduling continuity parameter, a unit period peak resource pressure occupation factor and a period average resource load balancing index, the priority in list task list specifically comprises a task access demand trend label, a period priority identification state value and a density trend stability flag item, the node scheduling adaptive segment list comprises a period fluctuation mean factor, a period standard deviation stability mark and an adjustable segment boundary index, and the multi-task matching node mapping table specifically refers to a node resource remaining capacity index, a task resource demand mapping relationship and a matching task identification collation number.

[0012] As a further scheme of the application, the acquisition step of the period load compact record set is specifically:

[0013] S101: Obtain CPU usage rate, memory occupation rate and task scheduling log of the registered task in the algorithm network fusion architecture in three continuous resource allocation periods, according to the task trigger time, end time and resource usage paragraph identification in the task scheduling log, extract the CPU usage rate sequence and the memory occupation rate sequence corresponding to the task in each unit time period, and establish a task corresponding period resource usage basic data set;

[0014] S102: Based on the periodic resource usage basis data set, the maximum value of the CPU usage rate sequence of each task in a unit time period is extracted as a CPU occupancy rate peak value, the length of a time period that is continuously not lower than a set threshold value in the memory occupancy rate sequence is called, the resource usage interval concentration degree of the task in the period is calculated, and a resource concentration expression value list is generated;

[0015] S103: Based on the resource concentration expression value list, the resource interval concentration degree values of the task in three continuous periods are weighted and averaged according to the same task, the periodic load convergence strength of the task is obtained in combination with the task existence identifier and the execution section coverage in the period, and a periodic load compact record set is established.

[0016] As a further scheme of the present application, the obtaining step of the priority in list is specifically:

[0017] S201: Based on the periodic load compact record set, the resource usage concentration degree values of each task in three continuous scheduling periods are extracted, the values are arranged in a period sequence in time sequence, the adjacent items in the period sequence are judged for a greater than relationship, and the number of continuously established paragraphs is counted to obtain the resource usage growth trend length of the task in the corresponding period section, and a resource concentration increasing section length value is generated.

[0018] S202: Based on the resource concentration increasing section length value, a set resource density threshold value is obtained according to the growth section length of each task, the resource usage concentration degree sequence of the task in three periods is called, a greater than relationship judgment is performed on each value, the number of items meeting the threshold value condition is counted and compared with the growth section length, the task identifier meeting the three-period increasing and higher than the threshold value is obtained, and a resource trend compliant task number set is generated.

[0019] S203: The resource trend compliant task number set is called, the original task list is filtered according to the task identifier information, the task items that do not meet the continuous growth and high resource density conditions are excluded, the task set meeting the access tendency judgment requirement is obtained, and a priority in list is established.

[0020] As a further scheme of the present application, the obtaining step of the node scheduling adaptation section list is specifically:

[0021] S301: Based on the priority in list, the CPU instruction cycle completion number and the memory usage percentage of the node in the last three scheduling periods are collected according to the assigned node corresponding to each task, the time period corresponding to the maximum value of the CPU usage rate sequence in each period is extracted and recorded as a period CPU peak time period, and a period peak time index set is generated.

[0022] S302: Based on the periodic peak time index set, calculate the interval on the time axis according to two adjacent CPU peak time periods and construct the time interval sequence in sequence. Call the node memory usage percentage sequence within the corresponding period to calculate the adjacent interval sequence and obtain the scheduling period stability scale value.

[0023] S303: Based on the scheduling cycle stability scale value, determine whether it is below the set stability threshold range, obtain the cycle index segment that meets the scheduling conditions, mark it as a schedulable window, extract the cycle time period range that is consistent with the task mapping relationship, and establish a node scheduling adaptation segment list.

[0024] As a further aspect of the present invention, the formula for obtaining the scheduling cycle stability scale value is specifically as follows:

[0025] ;

[0026] in, Represents the stability scale value of the scheduling cycle. Representing the Normalized value of each CPU peak time point Representing the Normalized value of each CPU peak time point Representing the Percentage of memory usage of a node within a scheduling cycle. This represents the mean of the normalized time interval divided by the corresponding memory metric. This represents the total number of scheduling cycles.

[0027] As a further aspect of the present invention, the step of obtaining the multi-task matching node mapping table specifically includes:

[0028] S401: Based on the node scheduling adaptation segment list, extract the executable state interval of each node within the task request cycle, collect the current remaining number of CPU cores and unallocated memory capacity of each node, record the node resource status information, and establish a set of adjustable resource parameters for the node.

[0029] S402: Based on the set of adjustable resource parameters of the node, obtain the number of CPU core requests and the memory capacity request value of each task, calculate the ratio with the remaining number of CPU cores and the remaining memory capacity of the node respectively, and combine the node task response latency value and the duration of occupation for joint adjustment, calculate the resource adaptation strength value between the task and the node, compare it with the set resource adaptation threshold, filter the task node pairs that exceed the threshold, and obtain the resource adaptation matching result;

[0030] The formula for calculating the resource compatibility strength value between tasks and nodes is as follows:

[0031] ;

[0032] wherein, represents the resource adaptation intensity value of the first task, represents the CPU core number requested by the first task, represents the currently allocable CPU core number of the first node, represents the memory capacity requested by the first task, represents the currently remaining allocable memory capacity of the first node, represents the occupation duration number of the first task, represents the occupation duration number of the first task, represents the response delay normalized value of the first task under the assigned node.

[0033] S403: calling the resource adaptation matching result, according to the task node pair identifier, summarizing and arranging the number, parameters and matching state between the task and the node, eliminating the matching combination that does not meet the condition, and establishing a multi-task matching node mapping table.

[0034] As a further scheme of the present application, the method further comprises the following steps:

[0035] S5: based on the multi-task matching node mapping table, according to the transmission link of the node, constructing a round-trip delay variation ratio sequence, and calculating the difference value with the corresponding position of the previous period sequence, if the difference value is greater than the round-trip delay variation threshold, eliminating the path, using the remaining link to implement task data transmission processing, and obtaining task scheduling execution information;

[0036] The task scheduling execution information includes a link stability state marker, a path connectivity level classification item and a node path screening index set.

[0037] As a further scheme of the present application, the task scheduling execution information acquisition step is specifically:

[0038] S501: based on the multi-task matching node mapping table, according to the established task and node pairing information, extracting the transmission link of each task dependent node, collecting the minimum round-trip delay and the maximum round-trip delay of the link within the last three monitoring periods, respectively constructing the round-trip delay ratio sequence in each period, and obtaining a round-trip delay variation ratio sequence set;

[0039] S502: calling the round-trip time variation ratio sequence set, selecting the corresponding positions of the current period and the previous period, calculating the difference of the corresponding ratios of the two periods, calling the set round-trip time variation threshold, comparing the difference, marking the link greater than the threshold as an unstable path, and generating link fluctuation state marking information;

[0040] S503: calling the link fluctuation state marking information, eliminating the corresponding link task pair from the multi-task matching node mapping table, extracting the corresponding link of the remaining matching task as a data transmission path and completing task data forwarding, and establishing task scheduling execution information.

[0041] A multi-task processing system in an algorithm-network fusion environment, the multi-task processing system in the algorithm-network fusion environment is used for realizing the multi-task processing method in the algorithm-network fusion environment, and the system comprises:

[0042] The period load analysis module obtains the CPU usage rate, memory occupation rate and task scheduling log of the registered task in the algorithm-network fusion architecture in three continuous resource allocation periods, calculates the resource use concentration degree of the task in each period, and generates a period load compact record set;

[0043] The task priority evaluation module calculates the length of the monotonically increasing segment of the sequence according to the resource use concentration degree sequence of the task based on the period load compact record set, and compares the length with the set resource density threshold value, if the length meets the condition of increasing in three continuous periods and exceeds the threshold value, the task is listed as a high access tendency, and a priority in list task list is obtained;

[0044] The node scheduling analysis module extracts the CPU peak period in each period based on the priority in list task list, calculates the interval between adjacent two peaks of the node, calculates the fluctuation degree of the sequence, judges whether the fluctuation is lower than the set stability threshold value, and generates a node scheduling adaptation section list;

[0045] The task matching evaluation module calculates the resource adaptation ratio according to the task request parameter based on the node scheduling adaptation section list, if the resource adaptation ratio is greater than the set resource adaptation threshold value, the task matching evaluation module marks the task matching evaluation module as matching success, and obtains a multi-task matching node mapping table;

[0046] The scheduling execution module constructs a round-trip time variation ratio sequence according to the transmission link of the node based on the multi-task matching node mapping table, calculates the difference of the corresponding positions of the sequence of the previous period, if the difference is greater than the round-trip time variation threshold value, the path is eliminated, the task data transmission processing is implemented by using the remaining link, and task scheduling execution information is obtained.

[0047] Compared with the prior art, the advantages and positive effects of the application are that:

[0048] In the application, by quantitatively analyzing the CPU occupancy peak value and memory duration of the task in the continuous three resource allocation periods, the resource use concentration is obtained, the high access task with increasing resource demand is accurately identified, the task priority is sorted, and the scheduling accuracy is improved, the stable schedulable window of the node resource is determined based on the sequence fluctuation degree calculation of the node CPU peak period interval, the mismatch between the node resource scheduling and the task demand is effectively avoided, the adaptation degree between the task demand and the node resource is verified in real time combined with the node resource adaptation ratio analysis, the efficient matching of the task and the node resource is ensured, and the resource waste and task delay are avoided, in the task transmission link selection aspect, by constructing the round-trip delay variation ratio sequence and calculating the continuous period difference, the unstable link is timely eliminated, more reliable and efficient task data transmission is realized, and the stability, accuracy and resource utilization efficiency of the task processing in the whole algorithm network fusion environment are improved. BRIEF DESCRIPTION OF DRAWINGS

[0049] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0050] Figure 1 The workflow diagram of the present application;

[0051] Figure 2 The S1 refinement flowchart of the present application;

[0052] Figure 3 The S2 refinement flowchart of the present application;

[0053] Figure 4 The S3 refinement flowchart of the present application;

[0054] Figure 5 The S4 refinement flowchart of the present application;

[0055] Figure 6 The S5 refinement flowchart of the present application;

[0056] Figure 7 The system flowchart of the present application. DETAILED DESCRIPTION

[0057] The technical solutions in the present application will be described below in combination with the drawings.

[0058] In the embodiments of the present application, the words such as "exemplary", "for example", etc. are used to represent an example, illustration or description. Any embodiment or design scheme described as "exemplary" in the present application should not be interpreted as more preferred or more advantageous than other embodiments or design schemes. Rather, the word "exemplary" is intended to present the concept in a specific manner. In addition, in the embodiments of the present application, the meaning expressed by "and / or" can be both, or can be one of the two.

[0059] In the embodiments of the present application, "image" and "picture" can be used interchangeably at times, and it should be pointed out that the meanings expressed are consistent when the distinction is not emphasized. "Of", "corresponding" and "corresponding" can be used interchangeably at times, and it should be pointed out that the meanings expressed are consistent when the distinction is not emphasized.

[0060] In the embodiments of the present application, sometimes the subscript such as W1 can be written in the form of non-subscript such as W1, and the meanings expressed are consistent when the distinction is not emphasized.

[0061] In order to make the technical problems, technical schemes and advantages to be solved by the present application more clear, the following will be described in detail in conjunction with the drawings and specific embodiments.

[0062] Please refer to Figure 1 The present application provides a technical scheme: a multi-task processing method in an algorithm network fusion environment, comprising the following steps:

[0063] S1: Obtain the CPU usage rate, memory occupation rate and task scheduling log of the registered tasks in the algorithm network fusion architecture within three continuous resource allocation periods, calculate the resource usage concentration of the tasks in each period based on the CPU occupation rate peak value and memory retention time of each task in a unit time period, and generate a period load compact record set;

[0064] S2: Based on the period load compact record set, calculate the length of the monotonically increasing segment of the sequence according to the resource usage concentration sequence of the tasks in three scheduling periods, and compare it with the set resource density threshold value. If the increase in three periods and the threshold value are met, the task is listed as high access tendency, and a priority entry task list is obtained;

[0065] S3: based on the priority task list, according to the task assignment node, the CPU instruction cycle completion number and the memory usage percentage in the last three cycles of the node, the CPU peak period in each cycle is extracted, the interval between adjacent two peaks of the node is calculated, the peak interval sequence is constructed, and the fluctuation degree of the sequence is calculated, whether the fluctuation is lower than the set stability threshold is judged, if it is satisfied, the period section is identified as a schedulable window, and a node scheduling adaptive section list is generated;

[0066] S4: based on the node scheduling adaptive section list, according to the executable state interval of the node, the remaining schedulable CPU core number and the allocatable memory capacity in the task request period are extracted, the resource adaptation ratio is calculated according to the task request parameters, if the resource adaptation ratio is greater than the set resource adaptation threshold, it is marked as matching success, and a multi-task matching node mapping table is obtained;

[0067] S5: based on the multi-task matching node mapping table, according to the transmission link of the node, the minimum round trip time and the maximum round trip time of the link in the last three monitoring periods are extracted, the round trip time variation ratio sequence is constructed, and the difference value is calculated with the corresponding position of the previous period sequence, if the difference value is greater than the round trip time variation threshold, the link is marked as unstable state, the path is removed, the task data transmission processing is implemented by using the remaining link, and task scheduling execution information is obtained;

[0068] The cycle load compact record set includes task cycle resource scheduling continuity parameters, unit cycle peak resource pressure occupation factor and cycle average resource load balancing index, the priority task list specifically includes task access demand trend label, cycle priority identification state value and density trend stability mark item, the node scheduling adaptive section list includes cycle fluctuation mean factor, cycle standard deviation stability mark and adjustable section boundary index, the multi-task matching node mapping table specifically refers to node resource remaining capacity index, task resource demand mapping relationship and matching task identification contrast number, and the task scheduling execution information includes link stability state mark, path connectivity level classification item and node path screening index set.

[0069] Please refer to Figure 2 The acquisition step of the cycle load compact record set is specifically:

[0070] S101: the CPU usage rate, memory occupation rate and task scheduling log of the registered task in the continuous three resource allocation periods in the algorithm network fusion architecture are acquired, the CPU usage rate sequence and the memory occupation rate sequence corresponding to the task in each cycle unit time period are extracted according to the task trigger time, end time and resource use paragraph identification in the task scheduling log, and the task corresponding cycle resource use basic data set is established;

[0071] The CPU usage rate, memory occupation rate and task scheduling log of the registered task in the continuous three resource allocation periods in the acquisition algorithm network fusion architecture are obtained. First, the task scheduling log is subjected to field extraction operation, each task trigger time and end time is parsed item by item, and the scheduling start and end time of the task in each resource allocation period is extracted. The task number field in the log is matched with the period time axis index, the task event is mapped to the time period in the fixed period, and then the resource data extraction operation is performed on the time period in each mapping period. The CPU usage rate data in the unit time period (such as 1 second) is extracted from the structured CPU usage record to form a time series, and the memory usage record in the same period is subjected to time alignment operation. The memory occupation rate sequence value at the corresponding time point is extracted. If there is a task segmentation execution identifier in the log, the CPU and memory data need to be partitioned according to the scheduling time period, the scheduling interruption time period data is removed, and only the resource usage record in the continuous scheduling time is reserved. Further, the CPU and memory usage rate are subjected to task number aggregation operation to ensure that each task forms an independent resource usage sequence in different periods. In the example, if a task taskA runs from 0 seconds to 5 seconds in the first period, the CPU usage rate is 42%, 51%, 47%, 53%, 49% and 50% respectively, and the memory occupation rate is 62%, 63%, 63%, 64%, 65% and 64% respectively, then the resource sequence of taskA in the first period is recorded as the above data. All tasks are subjected to batch processing, and the task ID, period number, CPU usage rate sequence and memory occupation rate sequence are subjected to field structure combination. Finally, the period resource usage basic data set corresponding to the task is established;

[0072] S102: Based on the period resource usage basic data set, the maximum value of the CPU usage rate sequence of each task in the unit time period is extracted as the CPU occupation rate peak value, the length of the time period in which the memory occupation rate sequence is continuously not less than the set threshold value is called, and the resource concentration degree of the task in the period is calculated to generate a list of resource concentration expression values;

[0073] Based on the period resource usage basic data set, first, the maximum value identification operation is performed on the CPU usage rate sequence of each task in each period. The peak CPU usage rate is selected from the time sequence and the corresponding value and time point are recorded. The memory occupation rate sequence in the period is traversed, and the memory value in each time slice is compared with the preset memory threshold value. If the continuous value is greater than the memory threshold value, it is counted into the resource maintenance section, and the continuous length is counted as the memory occupation duration length. The memory threshold value is set to , which is set according to the average memory occupation rate of all tasks in the current system scheduling period plus the maximum tolerance fluctuation value of the system scheduling , that is: , wherein the average memory occupancy of all tasks in the period, a fixed system scheduling safety margin (such as 5%), for example when , then , if the memory sequence of a task in 5 seconds is 62%, 63%, 63%, 64%, 65%, 64%, the continuous segment length is 6 seconds, on this basis, two parameters of each task in each period are constructed: CPU peak value and memory occupancy duration, which are used as the resource set description indicators of the task in the period, and further combined according to the task dimension to form the task resource expression structure, in the example, if the CPU peak value of taskA in period 1 is 53% and the memory maintenance time period is 6 seconds, the resource set description parameter of taskA is (53, 6), each task is processed in sequence to generate the resource expression sequence of all tasks in three periods, and finally the resource set expression value list is generated;

[0074] S103: Based on the resource set expression value list, the resource interval concentration degree values of the same task in each period are weighted and averaged according to the resource interval concentration degree values of the task in the continuous three periods, and the task period load convergence strength is obtained by combining the task existence identifier and the execution segment coverage in the period, and the period load compact record set is established.

[0075] According to the resource interval concentration degree values of each task in the resource set expression value list in the continuous three periods, first, the resource expression values of each task in the three periods are weighted and integrated, and the CPU peak value and the memory duration are weighted and averaged. The weight used in the weighting operation is , which is determined by the execution continuity of the task in the period, if the effective running time of a task in the period is not less than 80% of the average scheduling time of the task , then is set, otherwise is set, this rule aims to distinguish the weight difference between complete scheduling and interrupted tasks, in the example, if the continuous running time of taskA in period 1 and period 2 is greater than , and the running is interrupted in period 3, then the weighted average value of the CPU peak value of taskA is:

[0076] ;

[0077] wherein, is the CPU peak value of the period, For the corresponding period weight value, and its memory maintains the length of similar weighting operation, while judging the task in the period of execution segment coverage ratio is more than 50%, if less than this coverage ratio, its period weight is automatically set to 0, prevent the task is not completed but the resource occupation abnormal situation influence overall judgment, finally to the weighted average results of structured arrangement, form the task dimension of the period resource centralized description structure, the structure will be as one of the basic reflection index of resource behavior stability, output task number and weighted CPU centralized value, memory centralized value, finally establish the period load compact record set.

[0078] Please refer to Figure 3 , the priority list of tasks is obtained by:

[0079] S201: Based on the period load compact record set, extract the resource use concentration value of each task in three consecutive scheduling periods, arrange them in time sequence as period sequence, judge the greater relationship of adjacent items in period sequence and count the number of continuous established paragraphs, obtain the resource use growth trend length of the task in the corresponding period segment, and generate resource concentration incremental segment length value;

[0080] Based on the period load compact record set, extract the resource use concentration value of each task in three consecutive scheduling periods, first analyze the scheduling period number corresponding to the task, and extract its resource concentration value in period 1, 2 and 3 from the record set with task number as the primary key. The extracted values are arranged in time sequence to form a one-dimensional ordered sequence, and then the greater relationship of adjacent two items in the sequence is executed, that is, whether the value of the first period is greater than the value of the second period, if the judgment is established, it is counted as an incremental, if not, the current incremental paragraph is interrupted and re-counted, in the process, the variable records the length of the continuous paragraph, in the example, if the resource concentration sequence of a task is 42.3, 48.9 and 52.1, both comparisons are established, and the final is obtained, if the sequence is 45.2, 44.6 and 50.3, it is interrupted once, only the single incremental record is retained, and the final is obtained, finally, the key value pair structure is constructed according to the task number and its incremental length, and the resource concentration incremental segment length value is obtained;

[0081] S202: Based on the resource concentration incremental segment length value, according to the growth segment length of each task, obtain the set resource density threshold value, call the resource use concentration sequence of the task in three periods, execute the greater relationship judgment for each value, count the number of items meeting the threshold condition and compare with the growth segment length, obtain the task identification meeting the three period incremental and higher than the threshold value, generate resource trend compliance task number set;

[0082] Based on the growth segment length of each task within the resource concentration increment segment length value, after obtaining the resource density threshold, the resource concentration values ​​of each task in the three periods are compared and judged one by one. The resource density threshold is set as follows: The threshold value was set with reference to the average resource consumption of the task. Concentration offset coefficient allowed by the system Using formula ,in This represents the average resource concentration of all tasks within the period. To ensure a fixed setting value, it is generally set to 6. In this implementation example, it is set as follows: , Then there is Determine whether each task exceeds the threshold in each of the three cycles, and execute if the threshold is greater than [a certain value]. The judgment operation is as follows: if three consecutive conditions are met and the corresponding value of the task in the increasing segment length value of the resource set is equal to 2, then the task meets the two conditions of "continuous growth and high density". The task number that meets the conditions is recorded. In the example, if the resource value of taskX is 52.3, 54.6, and 56.2 and the increasing segment length is 2, then it is recorded as a qualified task. Finally, a numbering structure is formed to obtain the resource trend conforms to the task number set.

[0083] S203: Call the resource trend matching task number set, filter the original task list according to the task identification information, filter out task items that do not meet the conditions of continuous growth and high resource density, obtain the task set that meets the access tendency judgment requirements, and establish a priority task list.

[0084] The task identification information in the task number set is called to retrieve resource trends. The task number is then filtered on the original task list, and only task entries with matching task numbers are retained. The task number field in the original task table is compared with the number set. If a match is found, the task is retained; otherwise, it is removed. The task attribute field information is then extracted based on the retained items, and the task ID, scheduling cycle index, and CPU and memory resource requirement values ​​are structured and organized into a schedulable task structure example. If taskZ exists in the number set and its fields are CPU cores requested (4 cores), memory requested (8GB), and scheduling cycle (1-3), it is directly included in the structure output result, completing the task access judgment process and finally establishing a priority task list.

[0085] Please see Figure 4 The specific steps for obtaining the node scheduling adaptation segment list are as follows:

[0086] S301: Based on the priority task list, according to the assigned node corresponding to each task, the CPU instruction cycle completion number and the memory usage percentage of the node in the last three scheduling periods are collected, the time period corresponding to the maximum value of the CPU usage rate sequence in each period is extracted and recorded as the peak period of the CPU, and a peak time index set is generated;

[0087] The assigned node corresponding to each task in the priority task list is obtained, the mapping information between the task and the node is extracted one by one, the task number and node identifier field in the task scheduling relationship table are located, the resource statistics record of each node in the last three scheduling periods is searched, the CPU instruction cycle completion number in the period dimension and the memory usage percentage sequence corresponding to the period are extracted, the CPU execution record is aggregated by second or millisecond time slice to form a continuous usage rate sequence, the CPU usage rate sequence in each period is called, the peak usage rate point is extracted and its corresponding absolute timestamp is recorded, three time values are constructed, and then they are arranged into a group of time sequence values. If the maximum values of the CPU usage rate recorded by a node in three periods appear at the 10th second, the 24th second and the 38th second respectively, the timestamp is , , The group of time points constitutes a peak time index set in a period.

[0088] S302: Based on the peak time index set in a period, according to the adjacent two CPU peak time periods, the interval on the time axis is calculated and the time interval sequence is constructed in order, the adjacent interval sequence is called for the node memory usage percentage sequence in the corresponding period, and the scheduling period stability scale value is obtained by operation;

[0089] The formula for calculating the scheduling period stability scale value is as follows:

[0090] ;

[0091] Among them, represents the scheduling period stability scale value, represents the normalized value of the first CPU peak time point, represents the normalized value of the second CPU peak time point, represents the memory usage percentage of the node in the first scheduling period, represents the normalized time interval divided by the corresponding memory index, represents the total number of scheduling periods.

[0092] ​​​The scheduling period stability scale value is used to quantify the fluctuation degree of the scheduling load of the algorithm network fusion node in multiple scheduling periods. Its essence is a variable standard deviation index, reflecting the stability of the CPU load peak time point and the resource scheduling smoothness of the node memory usage linkage. The smaller this value is, the smaller the time interval change of the CPU peak value in the scheduling period is, and the more stable the system scheduling is. On the contrary, it means that there is a burst load peak in the system, and the scheduling interval is unstable.

[0093] The formula is essentially the square root of the variance of the weighted normalized time interval sequence, which meets the statistical standard modeling mode of the stability index. The "+1" is added to avoid the zero division exception caused by memory being 0, and to strengthen the amplification effect of the low memory high fluctuation scenario.

[0094] Based on the three CPU peak time stamps in the cycle peak time index set, first, each time value is normalized, and the maximum and minimum normalization method is used to transform the time value. The specific processing method is: assuming that the peak times of the node in the three periods are the 10th second, the 24th second and the 38th second, then 、 、 The normalization formula is:

[0095] ;

[0096] Among them, 、 , then:

[0097] ;

[0098] ;

[0099] ;

[0100] Next, the time difference value sequence is constructed, and the following is obtained:

[0101] ;

[0102] ;

[0103] Then, the memory usage percentage of the node in the three periods is obtained, which is: 、 , expressed as a proportion value 、 , and the intermediate result of the sum of each time difference and memory ratio value is calculated:

[0104] ;

[0105] ;

[0106] Further calculate the average value

[0107]

[0108] Substitute the formula:

[0109]

[0110] Finally, the scheduling period stability scale value of the node in the current three periods is , which is used to determine whether it meets the stability threshold condition set in the subsequent steps. The entire process completes the complete processing flow from the original time data and memory occupation data to normalization, ratio conversion, mean extraction and standard deviation calculation, which constitutes the basis for the scheduling fluctuation judgment.

[0111] S303: According to the scheduling period stability scale value, it is judged whether it is lower than the set stability threshold range, the period index section that meets the scheduling condition is obtained, which is marked as a schedulable window, the period time section range consistent with the task mapping relationship is extracted, and the node scheduling adaptive section list is established.

[0112] According to the scheduling period stability scale value, the set stability threshold is called for comparison. The threshold setting is jointly determined according to the node hardware properties and task type. The specific setting logic is: when the CPU type of the task carried by the node is a general task, the tolerance of the system to fluctuation stability is set to 0.15; if the task type is a graph computing task, it requires more stringent scheduling rhythm, and the fluctuation threshold is set to 0.004. In the example, when the node's , its task is a graph computing task, because does not meet the stability requirement, it is not marked as a schedulable window. If the node's , it meets the determination standard, marks the period as a schedulable window, and further combines the task scheduling information to map the corresponding time period to the scheduling period index allocated to the task, completes the extraction of the space-time correspondence relationship of the task node, and finally establishes the node scheduling adaptive section list.

[0113] Please refer to Figure 5 , the acquisition step of the multi-task matching node mapping table is specifically:

[0114] S401: Based on the node scheduling adaptive section list, extract the executable state interval of each node in the task request period, collect the current remaining CPU core number and unallocated memory capacity of each node, record the node resource status information, and establish the node adjustable resource parameter set; ​​​

[0115] Based on the node scheduling adaptation segment list, the scheduling cycle range and node number corresponding to each task are read sequentially. The node corresponding to the task is extracted from the scheduling node mapping table, and the status record within the task request cycle is retrieved according to the node number. The corresponding cycle index field is located, and the executable identifier field in the node cycle status data is extracted, with the interval continuously marked "1" as the effective adjustable segment of the node. On this basis, the remaining resource status of the node within this time period is further extracted, specifically the currently available CPU cores and unallocated memory capacity, and the field values ​​are extracted respectively. , In the example, if a node N1 has 6 remaining CPU cores and 18GB of remaining memory during a task request period, this parameter information is associated with the task mapping table to construct the adjustable resource supply set for that task during that period, and the task number, node number, and other parameters are used to determine the resource supply set. , The four items are structured and combined into entries, and all task node combinations are processed in sequence to finally establish a set of adjustable resource parameters for nodes.

[0116] S402: Based on the set of adjustable resource parameters of the node, obtain the number of CPU core requests and the memory capacity request value of each task, calculate the ratio with the remaining number of CPU cores and the remaining memory capacity of the node respectively, and combine the node task response latency value and the duration of occupation for joint adjustment, calculate the resource adaptation strength value between the task and the node, compare it with the set resource adaptation threshold, filter the task node pairs that exceed the threshold, and obtain the resource adaptation matching result;

[0117] The formula for calculating the resource compatibility strength value between tasks and nodes is as follows:

[0118] ;

[0119] in, Indicates the first The resource adaptation strength value for each task. Indicates the first The number of CPU cores for each task request. Indicates the first The number of CPU cores currently available for allocation on each node. Indicates the first The memory capacity requested by each task. Indicates the first The current remaining allocatable memory capacity of each node. Indicates the first The number of durations occupied by each task. Indicates the first Normalized response latency of each task under the assigned node;

[0120] Resource adaptation strength value is used to quantify the adaptation tightness between each task and target node resources, and measure whether it is suitable to schedule and execute the task. It comprehensively considers the CPU core number matching degree, memory capacity adaptation tightness, and the coupling effect of task life cycle and node response delay. The lower the value is, the higher the task adaptation to node resources is, and it is more suitable for current scheduling matching; when it is higher than the threshold value, it means that the resource matching burden is too heavy or the scheduling delay is unreasonable;

[0121] The first term : measures the CPU core matching strength, which directly reflects whether it can be carried;

[0122] The second term : represents the coupling error of memory usage efficiency and task life cycle, which amplifies the influence of mismatching by weighting the node response delay. This formula combines resource load ratio, life cycle tightness, and execution delay sensitivity, and can more truly reflect the task scheduling adaptation in heterogeneous resource environment, with the advantages of engineering practicability and algorithm precision balance;

[0123] According to the resource matching relationship between tasks and nodes in the node adjustable resource parameter set, the resource demand parameters of each task are called, and the CPU core request number and memory capacity request value are read respectively 、 At the same time, the number of occupied continuous periods and the normalized value of task response delay on the node are extracted, and these parameters are substituted into the resource adaptation strength formula:

[0124] ;

[0125] The calculation is as follows: in the actual example, if task T5 requests 4 core CPU, 8 GB memory, occupies a period of 3, and the current resources of node N1 are 6 core CPU, 18 GB memory, and the normalized value of response delay is 0.36, then the calculation is as follows:

[0126] ;

[0127] ;

[0128] ;

[0129] The intermediate difference is , and the final calculation result is:

[0130] ;

[0131] The system sets the resource adaptation threshold value The formula yields a task T5 fit value of 0.7834, which is lower than the set threshold. Therefore, the task is considered to be a valid match with the node. The task number and node number are then compared... Record all results together, and finally output the resource adaptation and matching results;

[0132] S403: Call the resource adaptation matching result, summarize and organize the number, parameters and matching status between tasks and nodes according to the task node pair identifier, eliminate matching combinations that do not meet the conditions, and establish a multi-task matching node mapping table.

[0133] Retrieve all task and node combination identifiers from the resource adaptation matching results, perform a dual primary key matching operation between task number and node number, and filter and remove all resource adaptation values. Exceeding the set threshold The combination retains tasks that meet the pairing conditions. The node numbers and resource configuration parameters are mapped and aggregated according to the task number. For each task record, the corresponding matching node number, number of CPU cores used, memory capacity value, and resource adaptation strength value fields are summarized into a task mapping information structure. In the example, if task T5 matches node N1, then the entry T5→N1{CPU:4,MEM:8,} is generated. :0.7834}, thereby completing the pairing and organization of all tasks, and finally establishing a multi-task matching node mapping table.

[0134] Please see Figure 6 The steps for obtaining the task scheduling and execution information are as follows:

[0135] S501: Based on the multi-task matching node mapping table, according to the established task and node pairing information, extract the transmission link of each task-dependent node, collect the minimum round-trip time and maximum round-trip time of the link in the last three monitoring periods, construct the round-trip time ratio sequence in each period, and obtain the round-trip time variation ratio sequence set.

[0136] Based on the task and node pairing information already established in the multi-task matching node mapping table, the transmission link number of each task-dependent node is extracted. The communication path identifier between each task node is read according to the node topology table. Link performance data recorded by the network performance monitoring module is then retrieved to obtain the minimum round-trip time of the target link within the last three scheduling monitoring periods. With maximum round-trip delay Periodic index Perform a ratio calculation operation for each cycle to construct the round-trip delay ratio. In the example, if the minimum and maximum round-trip times of a link in three periods are: Period 1: 5ms and 9ms, Period 2: 6ms and 12ms, Period 3: 7ms and 14ms, then its ratio sequence is as follows: 、 、 store the ratio sequence of the link into the structured data table, perform the same processing for all links, and finally build the round-trip delay variation ratio sequence set corresponding to the task link;

[0137] S502: Call the round-trip delay variation ratio sequence set, select the corresponding positions of the current period and the previous period, calculate the difference value of the corresponding ratios of the two periods, call the set round-trip delay variation threshold value, compare the difference value, mark the links greater than the threshold value as unstable paths, and generate link fluctuation state marking information;

[0138] Call the ratio sequence of each link in the round-trip delay variation ratio sequence set, compare the ratio items of the current monitoring period (set as the 3rd period) and the previous period (the 2nd period), that is, perform difference calculation for each link , and compare the difference value range with the set threshold value . The threshold value is set according to the link type. If the link is an inter-node backbone channel, the tolerance is set to 0.3. If it is an edge node or a cross-domain path, the threshold value is set to 0.15. In the example, if the link number L7 has a ratio of 2.0 and 1.6 in the two periods, the difference is 0.4, and its type is set as a backbone channel, then , the link is marked as unstable, and a field structure is established for the link number and stability flag pair. All links greater than the threshold value are marked as “unstable”, and the rest are marked as “stable”. Complete state records are output, and link fluctuation state marking values are generated;

[0139] S503: Call the link fluctuation state marking information, remove the corresponding link task pairs from the multi-task matching node mapping table, extract the corresponding links of the remaining matching task pairs as data transmission paths and complete task data forwarding, and establish task scheduling execution information;

[0140] According to the link number marked as “stable” in the link fluctuation state marking value, perform filtering operation on all task pairs in the multi-task matching node mapping table, and perform one-to-one screening on the mapping table according to the link identification corresponding to the task pair. Remove all combinations in which the link relied by the task is marked as “unstable”, and only keep the task node mapping items corresponding to the stable path. Then, the link number in the remaining combination is used as the data transmission path in the task scheduling stage. Record the corresponding task number, link number, path communication average delay, and task transmission data amount fields in the transmission stage. In the example, if the matching path of task T3 is L5, L5 is determined to be a stable path, the transmission data amount is 12MB, and the average round-trip delay is 8.5ms, then the task scheduling execution information is recorded as T3→L5{12MB, 8.5ms}. Finally, the mapping information in the task scheduling stage is arranged, and the task scheduling execution information is established.

[0141] Please refer to Figure 7 A multi-task processing system in an algorithm network fusion environment, the multi-task processing system in the algorithm network fusion environment is used for executing the multi-task processing method in the algorithm network fusion environment, and the system comprises:

[0142] The periodic load analysis module obtains the CPU usage rate, memory occupation rate and task scheduling log of the registered task in the algorithm network fusion architecture in the continuous three resource allocation periods, calculates the resource usage concentration degree of the task in each period, and generates a periodic load compact record set;

[0143] The task priority evaluation module calculates the length of the monotonically increasing segment of the sequence of the resource usage concentration degree of the task based on the periodic load compact record set, compares the length with the set resource density threshold value, and if the length meets the condition of increasing in the continuous three periods and exceeding the threshold value, the task is listed as having a high access tendency, and a priority in-list task list is obtained;

[0144] The node scheduling analysis module extracts the CPU peak period in each period based on the priority in-list task list, calculates the interval between adjacent two peaks of the node, calculates the fluctuation degree of the sequence, judges whether the fluctuation is lower than the set stability threshold value, and generates a node scheduling adaptation section list;

[0145] The task matching evaluation module calculates the resource adaptation ratio according to the task request parameter based on the node scheduling adaptation section list, and if the resource adaptation ratio is greater than the set resource adaptation threshold value, marks it as a matching success, and obtains a multi-task matching node mapping table;

[0146] The scheduling execution module constructs a round-trip delay variation ratio sequence according to the transmission link of the node based on the multi-task matching node mapping table, calculates the difference value at the corresponding position of the sequence of the previous period, and if the difference value is greater than the round-trip delay variation threshold value, the path is removed, the remaining link is used to implement the task data transmission processing, and task scheduling execution information is obtained.

[0147] It should be understood that the term "and / or" herein is only a description of the association relationship of the associated objects, which means that there can be three relationships, for example, A and / or B can represent: A exists alone, A and B exist together, and B exists alone, wherein A and B can be singular or plural. In addition, the character " / " herein generally represents an "or" relationship between the front and rear associated objects, but it can also represent an "and / or" relationship, which can be understood according to the context before and after.

[0148] In the present application, "at least one" means one or more, and "multiple" means two or more. "At least one of the following" or the like means any combination of the items, including a single item or any combination of multiple items. For example, at least one of a, b, or c can mean a, b, c, a-b, a-c, b-c, or a-b-c, where a, b, and c can be single or multiple.

[0149] It should be understood that the size of the sequence number of the above-mentioned processes does not mean the order of execution in various embodiments of the present application. The execution order of the processes should be determined according to their functions and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0150] Those of ordinary skill in the art can realize that the units and algorithm steps of the examples described in conjunction with the embodiments disclosed herein can be realized in electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0151] Those skilled in the art can clearly understand that, for the convenience and brevity of the description, the specific working processes of the devices, apparatuses and units described above can refer to the corresponding processes in the foregoing method embodiments, which will not be repeated here.

[0152] In several embodiments provided by the present application, it should be understood that the disclosed devices, apparatuses and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely schematic, for example, the division of the units is only a logical function division, and actual implementation can have another division manner, for example, multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the displayed or discussed units can be indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0153] The units described as separate components can or can not be physically separated, and the components shown as units can or can not be physical units, i.e. they can be located in one place or distributed on multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the present embodiment.

[0154] In addition, each function unit in each embodiment of the present application can be integrated in one processing unit, or each unit can be physically present separately, or two or more units can be integrated in one unit.

[0155] If the functions are realized in the form of software function units and sold or used as independent products, they can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application or the part of the prior art that essentially contributes or the part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The foregoing storage medium includes a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0156] The above is only a specific embodiment of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or replacements within the technical range disclosed by the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A multi-task processing method in a cloud computing environment, characterized in that, The method comprises the following steps: S1: obtaining CPU usage rate, memory occupation rate and task scheduling log of registered tasks in the computing network fusion architecture within three continuous resource allocation periods, calculating resource usage concentration of the tasks in each period, and generating a period load compact record set; S2: based on the period load compact record set, calculating the length of the monotonically increasing segment of the sequence of resource usage concentration of the tasks, and comparing it with the set resource density threshold value, if the increasing is continuous within three periods and exceeds the threshold value, the task is listed as high access tendency, and a priority entry task list is obtained; S3: based on the priority entry task list, extracting the CPU peak period in each period, calculating the interval between adjacent two peaks, calculating the fluctuation degree of the sequence, judging whether the fluctuation is lower than the set stability threshold value, and generating a node scheduling adaptive segment list; S4: based on the node scheduling adaptive segment list, calculating the resource adaptation ratio according to the task request parameters, if the resource adaptation ratio is greater than the set resource adaptation threshold value, marking as matching success, and obtaining a multi-task matching node mapping table.

2. The multi-task processing method in the cloud computing environment according to claim 1, wherein, The period load compact record set comprises task period resource scheduling continuity parameters, unit period peak resource pressure occupation factor and period average resource load balancing index, the priority entry task list specifically comprises task access demand trend label, period priority identification state value and density trend stability mark item, the node scheduling adaptive segment list comprises period fluctuation mean factor, period standard deviation stability mark and adjustable segment boundary index, and the multi-task matching node mapping table specifically refers to node resource remaining capacity index, task resource demand mapping relationship and matching task identification comparison number.

3. The multi-task processing method in the computing and network fusion environment according to claim 2, characterized in that, The acquisition step of the period load compact record set is specifically: S101: obtaining CPU usage rate, memory occupation rate and task scheduling log of registered tasks in the computing network fusion architecture within three continuous resource allocation periods, extracting CPU usage rate sequence and memory occupation rate sequence corresponding to the tasks in each unit time period according to task trigger time, end time and resource usage paragraph identification in the task scheduling log, and establishing a period resource usage basic data set corresponding to the tasks; S102: based on the period resource usage basic data set, extracting the maximum value of the CPU usage rate sequence of each task in the unit time period as the CPU occupation rate peak, calling the length of the time period in the memory occupation rate sequence which is not less than the set threshold value, calculating the resource usage interval concentration of the task in the period, and generating a resource concentration expression value list; S103: based on the resource concentration expression value list, performing weighted average operation on the resource interval concentration numerical value sequence of the same task in each period, combining the task existence identification and execution segment coverage in the period, obtaining the period load convergence intensity of the task, and establishing a period load compact record set.

4. The multi-task processing method in the computing and network fusion environment according to claim 3, characterized in that, The acquisition step of the priority entry task list is specifically: S201: Based on the periodic load compact record set, the resource utilization concentration value of each task in three consecutive scheduling periods is extracted, arranged in time sequence as a period sequence, the adjacent items in the period sequence are judged for greater relationship and the number of continuous paragraphs is counted, the resource utilization growth trend length of the task in the corresponding period section is obtained, and the resource concentration incremental section length value is generated; S202: Based on the resource concentration incremental section length value, the set resource density threshold value is obtained according to the growth section length of each task, the resource utilization concentration sequence of the task in three periods is called, the greater relationship is executed for each value, the number of items meeting the threshold condition is counted and compared with the growth section length, the task identification meeting the three-period incremental and higher threshold value is obtained, and the resource trend compliant task number set is generated; S203: The resource trend compliant task number set is called, the original task list is filtered according to the task identification information, the task items not meeting the continuous growth and high resource density conditions are screened out, the task set meeting the access tendency judgment requirement is obtained, and the priority in list task list is established.

5. The multi-task processing method in the computing and network fusion environment according to claim 4, characterized in that, The obtaining step of the node scheduling adaptation section list is specifically: S301: Based on the priority in list task list, the CPU instruction cycle completion number and memory usage percentage of the node in the last three scheduling periods are collected according to the assigned node corresponding to each task, the time section corresponding to the maximum value of the CPU utilization rate sequence in each period is extracted and recorded as the period CPU peak time section, and the period peak time index set is generated; S302: Based on the period peak time index set, the interval on the time axis is calculated according to the adjacent two CPU peak time sections, and the time interval sequence is constructed in order, the adjacent interval sequence is calculated by calling the node memory usage percentage sequence in the corresponding period, and the scheduling period stability scale value is obtained; S303: According to the scheduling period stability scale value, it is judged whether it is lower than the set stability threshold range, the period index section meeting the scheduling condition is obtained, which is marked as a schedulable window, the period time section range consistent with the task mapping relationship is extracted, and the node scheduling adaptation section list is established.

6. The multi-task processing method in the computing and network fusion environment according to claim 5, characterized in that, The formula for calculating the scheduling period stability scale value is specifically: ； in, Represents the stability scale value of the scheduling cycle. Representing the Normalized value of each CPU peak time point Representing the Normalized value of each CPU peak time point Representing the Percentage of memory usage of a node within a scheduling cycle. This represents the mean of the normalized time interval divided by the corresponding memory metric. This represents the total number of scheduling cycles.

7. The multi-task processing method in a cloud computing environment according to claim 6, wherein, The obtaining step of the multi-task matching node mapping table is specifically: S401: Based on the node scheduling adaptation section list, the executable state interval of each node in the task request period is extracted, the current remaining CPU core number and unallocated memory capacity of each node are collected, the node resource present situation information is recorded, and the node adjustable resource parameter set is established; S402: According to the node adjustable resource parameter set, the CPU core request number and memory capacity request value of each task are obtained, the ratio calculation is performed with the node remaining CPU core number and remaining memory capacity respectively, and the joint adjustment is performed by fusing the node task response delay value and the occupied duration length, the resource adaptation strength value between the task and the node is obtained by operation, and the resource adaptation matching result is obtained by comparing the set resource adaptation threshold value and screening the task node pairs exceeding the threshold value; The formula for obtaining the resource adaptation strength value between the task and the node is specifically: ； wherein, represents a resource adaptation intensity value of the th task, represents a CPU core number requested by the th task, represents a currently allocable CPU core number of the th node, represents a memory capacity requested by the th task, represents a currently remaining allocable memory capacity of the th node, represents an occupation duration period number of the th task, represents a response latency normalization value of the th task under the assigned node. S403: According to the task node pair identifier, the number, parameters and matching state between the task and the node are summarized and arranged by calling the resource adaptation matching result, and the matching combination that does not meet the condition is eliminated, and a multi-task matching node mapping table is established.

8. The multi-task processing method in the computing and web fusion environment according to claim 7, wherein, The method further comprises the following steps: S5: Based on the multi-task matching node mapping table, a round-trip delay variation ratio sequence is constructed according to the transmission link of the node, and a difference value is calculated at the corresponding position of the previous period sequence, if the difference value is greater than the round-trip delay variation threshold, the path is eliminated, the remaining link is used for task data transmission processing, and task scheduling execution information is obtained; The task scheduling execution information includes link stability state markers, path connectivity level classification items and node path screening index sets.

9. The multi-task processing method in a cloud computing environment according to claim 8, wherein, The task scheduling execution information obtaining step is specifically: S501: Based on the multi-task matching node mapping table, the transmission link of each task dependent node is extracted according to the established task and node pairing information, the minimum round-trip delay and the maximum round-trip delay of the link in the last three monitoring periods are collected, the round-trip delay ratio sequence in each period is constructed respectively, and a round-trip delay variation ratio sequence set is obtained; S502: The round-trip delay variation ratio sequence set is called, the corresponding positions of the current period and the previous period are selected, the difference value of the corresponding ratios of the two periods is calculated, the set round-trip delay variation threshold is called, the difference value is compared, and the link greater than the threshold is marked as an unstable path, and link fluctuation state marker information is generated; S503: The link fluctuation state marker information is called, the corresponding link task pair in the multi-task matching node mapping table is eliminated, the corresponding link of the remaining matching task is extracted as a data transmission path and the task data forwarding is completed, and task scheduling execution information is established.

10. A multi-task processing system in a computer-network converged environment, characterized in that, The system is used to realize the multi-task processing method in the algorithm network fusion environment as claimed in any one of claims 1-9, and the system comprises: The period load analysis module obtains the CPU usage rate, memory occupancy rate and task scheduling log of the registered task in the continuous three resource allocation periods in the algorithm network fusion architecture, calculates the resource usage concentration degree of the task in each period, and generates a period load compact record set; The task priority evaluation module calculates the length of the monotonically increasing segment of the sequence according to the resource usage concentration degree sequence of the task based on the period load compact record set, and compares the length with the set resource density threshold value, if the length meets the conditions of increasing in the continuous three periods and exceeding the threshold value, the task is listed as a high access tendency, and a priority in list task list is obtained; The node scheduling analysis module extracts the CPU peak period in each period based on the priority in list task list, calculates the interval between adjacent two peaks of the node, calculates the fluctuation degree of the sequence, judges whether the fluctuation is lower than the set stability threshold value, and generates a node scheduling adaptation section list; The task matching evaluation module calculates the resource adaptation ratio according to the task request parameters based on the node scheduling adaptation section list, if the resource adaptation ratio is greater than the set resource adaptation threshold value, the task matching is marked as successful, and a multi-task matching node mapping table is obtained; The scheduling execution module constructs a round-trip delay variation ratio sequence according to the transmission link of the node based on the multi-task matching node mapping table, and performs difference calculation with the corresponding position of the previous period sequence, and if the difference is greater than the round-trip delay variation threshold, the path is removed, the remaining link is used to implement the task data transmission processing, and the task scheduling execution information is obtained.

Images