A CPU effective load collection normalization method for system load evaluation

Abstract

The application discloses a CPU effective load collection normalization method for system load evaluation, relates to the technical field of computer operation monitoring, and comprises the following steps: combining and identifying key path weighted load segments and state modeling are carried out, execution behavior state feature sets are formed, real execution efficiency grade division is carried out according to the execution behavior state feature sets, and real execution efficiency labeled segments are generated; according to processor occupation in the real execution efficiency labeled segments, separation, hierarchical accumulation and differentiated inclusion are carried out, load structured results are generated, consistency verification is carried out, and unified verification load results are generated. The application realizes processor occupation attribution correction and execution quality distinction, improves the accuracy of business main chain occupation identification and the accuracy of effective load expression, realizes unified verification load result generation, and improves the reliability and comparability of system load evaluation results.

Description

Technical Field

[0001] This invention relates to the field of computer operation monitoring technology, and in particular to a normalized method for collecting CPU effective load for system load assessment. Background Technology

[0002] With the continuous development of cloud computing, container orchestration, microservice collaboration, edge computing, and heterogeneous computing platforms, load monitoring technologies for computing platform operation have evolved from single-machine performance monitoring to a comprehensive analysis stage encompassing cross-process, cross-thread, cross-container, and cross-business links. In related research and engineering practice, processor utilization has long been an important indicator for measuring host busyness, resource utilization, and operational pressure. Multiple implementation paths have emerged around the collection, merging, recording, and comparison of processor utilization, based on methods such as runtime record collection, call chain organization, occupancy interval segmentation, behavior segment identification, and result normalization.

[0003] While existing technologies can collect processor usage data, analyze call chains, and monitor runtime behavior, they still have two shortcomings in the normalization of CPU effective load data collection for system load assessment: First, existing technologies mostly summarize processor usage based on processes, threads, or containers, lacking a layer-by-layer correlation analysis of the business request propagation process and identification of critical path contributions, which can easily lead to inaccurate attribution of processor usage. Second, existing technologies mostly express processor usage based on time-based results, rarely combining the relationships between usage segments, execution behavior combinations, and execution state changes to distinguish the true execution efficiency, resulting in insufficient accuracy in expressing effective load. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a normalized method for CPU effective load acquisition in system load assessment to solve the problems of inaccurate processor occupancy attribution and insufficient accuracy in effective load representation.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] This invention provides a method for normalizing CPU effective load data collection for system load assessment, comprising:

[0008] Collect raw load sampling fragments and perform layer-by-layer association to establish a propagation chain mapping from business requests to execution objects, forming business-related load fragments;

[0009] Based on the propagation chain records in the business-related load segments, the association validity is verified, the path continuity is compared, and short-term bounce is identified. Redundant propagation relationships, abnormal bounce relationships, and invalid subordinate relationships are removed to generate candidate propagation path segments.

[0010] Perform path contribution analysis on candidate propagation path segments, identify critical paths, and redistribute propagation workload according to the contribution of each execution object in the critical path to the overall execution process, generating critical path weighted workload segments.

[0011] The weighted load segments of the critical path are combined, identified, and modeled to form an execution behavior state feature set. Based on the execution behavior state feature set, the actual execution efficiency level is divided, and the actual execution efficiency labeled segments are generated.

[0012] Based on the processor usage in the actual execution efficiency annotation segments, the data is separated, layered, accumulated, and differentially incorporated to generate a structured load result. Consistency verification is then performed to generate a unified verified load result.

[0013] As a preferred embodiment of the CPU effective load collection normalization method for system load assessment described in this invention, the generation of service-related load segments is specifically as follows:

[0014] The original load sampling segments are spliced ​​together to form business request threads, and the order is made consistent to generate a set of request-related segments;

[0015] Based on the set of request-related fragments, layer-by-layer association processing, propagation direction identification, and splicing of the beginning and end of the association links are performed to generate a propagation chain mapping, forming business-related load fragments.

[0016] As a preferred embodiment of the CPU effective load collection normalization method for system load assessment described in this invention, the step of performing correlation validity verification is as follows:

[0017] Perform head-to-tail connection checks on adjacent propagation chain records in the business-related load segments and attach them to the same link to be checked to form a propagation chain record group sequence;

[0018] The validity of the propagation chain record grouping sequence is verified, and a propagation chain record set is generated.

[0019] As a preferred embodiment of the CPU effective load collection normalization method for system load assessment described in this invention, the steps of performing association validity verification, path continuity comparison, and short-time bounce identification are as follows:

[0020] The propagation chain record set is checked for continuity status and backtracking features by performing path continuity comparison operations, thereby generating path continuity comparison records.

[0021] Short-term bounce identification and cross-checking are performed on the path continuity comparison records to obtain abnormal bounce relationships, redundant propagation relationships and invalid subordinate relationships. The relationships are then consolidated to generate a set of propagation paths to be purified.

[0022] As a preferred embodiment of the CPU effective load collection normalization method for system load assessment described in this invention, the generation of candidate propagation path segments is specifically as follows:

[0023] The propagation chain records corresponding to redundant propagation relationships, abnormal bounce relationships, and invalid subordinate relationships in the set of propagation paths to be purified are deleted in sequence to obtain the set of purified propagation path records.

[0024] The set of records of purification propagation paths is segmented into fragments to generate candidate propagation path fragments.

[0025] As a preferred embodiment of the CPU effective load acquisition normalization method for system load assessment described in this invention, the identification of critical paths specifically includes the following:

[0026] The candidate propagation path segments are marked to form a candidate propagation path sequence;

[0027] Path contribution analysis and sequence aggregation are performed based on the execution object link order in the candidate propagation path sequence to generate path contribution sorting fragments.

[0028] As a preferred embodiment of the CPU effective load collection normalization method for system load assessment described in this invention, the generation of critical path weighted load segments is specifically as follows:

[0029] Based on the path contribution ranking segments, key paths are identified, and corresponding markers are added to generate a set of propagation paths;

[0030] The propagation service load is redistributed across the propagation path set, generating a path load record set.

[0031] The path load record set is fragmented and continuous intervals are divided to generate critical path weighted load fragments.

[0032] As a preferred embodiment of the CPU effective load collection normalization method for system load assessment described in this invention, the formation of the execution behavior state feature set is specifically as follows:

[0033] Expand the critical path weighted load segment to locate different occupied segments one by one, forming a critical path weighted load sequence;

[0034] Segment relationships are identified in the weighted load sequence of the critical path to generate a sequence of combined execution behaviors;

[0035] Based on the sequence of execution behaviors, an execution state modeling mapping relationship is established, and then merged and refined to generate a set of execution behavior state features.

[0036] As a preferred embodiment of the CPU effective load collection normalization method for system load assessment described in this invention, the generation of the actual execution efficiency labeled segment is specifically as follows:

[0037] The different state chain segments in the execution behavior state feature set are divided into actual execution efficiency levels, and corresponding tags are added to generate an execution state record set;

[0038] The execution status record set is labeled, organized, and concatenated to obtain the actual execution efficiency labeled record set. Then, it is segmented into fragments to generate actual execution efficiency labeled fragments.

[0039] As a preferred embodiment of the CPU effective load collection normalization method for system load assessment described in this invention, the generation of unified verification load results is specifically as follows:

[0040] Based on the processor usage in the actual execution efficiency labeled segments, the processor usage is grouped, linked together, and separated to generate a set of processor usage records;

[0041] The processor usage record set is hierarchically accumulated and differentiated, and continuous load segments are spliced ​​to obtain a structured load result;

[0042] The structured load results are validated and corrected for consistency, generating a unified validated load result.

[0043] The beneficial effects of this invention are as follows: by performing path contribution analysis, critical path identification, propagation business load redistribution, and actual execution efficiency level classification steps, processor occupancy attribution correction and execution quality differentiation are achieved, thereby improving the accuracy of business main chain occupancy identification and effective load expression; by performing processor occupancy separation, layered accumulation, differentiated inclusion, and consistency verification steps, unified load verification result generation is achieved, thereby improving the reliability and comparability of system load assessment results. Attached Figure Description

[0044] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0045] Figure 1This is a flowchart illustrating a normalized method for collecting CPU effective load data for system load assessment.

[0046] Figure 2 A flowchart for generating candidate propagation path segments.

[0047] Figure 3 A flowchart for generating weighted load segments for the critical path.

[0048] Figure 4 A flowchart for generating segments labeled with actual execution efficiency and unified verification load results. Detailed Implementation

[0049] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0050] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0051] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0052] Reference Figures 1-4 This is one embodiment of the present invention, which provides a method for normalizing CPU effective load collection for system load assessment, including the following steps:

[0053] S1: Collect raw load sampling fragments and perform layer-by-layer association to establish a propagation chain mapping from business requests to execution objects, forming business-related load fragments;

[0054] S1.1: The original load sampling segments are spliced ​​together to form business request threads, and the order is made consistent to generate a set of request-related segments;

[0055] Furthermore, the original load sampling segments are grouped and retrieved according to business request identifier, connection identifier, session identifier, service entry identifier, and sampling time. Original load sampling segments containing the same business request identifier or having connection succession relationship, session continuation relationship, or service entry correspondence relationship are spliced ​​to form a business request thread. Around each original load sampling segment in the business request thread, the sampling time sequence relationship, connection continuation relationship, and request-response correspondence relationship are checked. Segments with reversed time, broken succession segments, and misaligned responses are readjusted to their corresponding positions, and consecutive succession segments with consistent order are merged and organized to generate a request-related segment set.

[0056] S1.2: Based on the set of request-related fragments, perform layer-by-layer association processing, propagation direction identification, and splicing the beginning and end of the association links to generate a propagation chain mapping, forming business-related load fragments.

[0057] Furthermore, the request-related fragment set is processed layer by layer according to the business request identifier and fragment acceptance order. The process derivation relationship, thread acceptance relationship, container call relationship, cross-container access relationship, and resource access relationship are checked downwards from the request start position. Execution objects with consecutive acceptance records are attached to the same related link. The call initiation position, call arrival position, and response return position between adjacent execution objects are compared. The access direction from the upstream execution object to the downstream execution object is identified as the forward propagation direction, and the response return direction from the downstream execution object to the upstream execution object is identified as the back propagation direction. Related links with consistent propagation directions, consecutive execution object acceptance, and consecutive acceptance positions are spliced ​​together. The end execution object of the previous related link is connected to the starting execution object of the next related link according to the link order, forming a propagation chain mapping. Then, based on the link order, acceptance time interval, and processor occupancy interval in the propagation chain mapping, fragmentation is performed to form business-related load fragments.

[0058] It should be noted that the propagation chain mapping includes at least the business request identifier, execution object identifier, propagation direction marker, link level position, acceptance time interval, processor occupancy interval, and resource access position.

[0059] S2: Based on the propagation chain records in the business-related load segments, perform association validity verification, path continuity comparison, and short-term bounce identification, and eliminate redundant propagation relationships, abnormal bounce relationships, and invalid subordinate relationships to generate candidate propagation path segments;

[0060] S2.1: Perform a head-to-tail connection check along adjacent propagation chain records in the business-related load segment and attach them to the same link to be checked to form a propagation chain record group sequence;

[0061] Furthermore, the propagation chain records in the business-related load segments are grouped and arranged according to the business request identifier, propagation direction marker, link level position, and acceptance time interval. Propagation chain records under the same business request identifier and adjacent to each other in the acceptance time interval are paired sequentially. For each group of adjacent propagation chain records, it is checked whether there is a succession relationship between the end execution object of the previous propagation chain record and the start execution object of the next propagation chain record, whether there is a continuous relationship between the end time of the previous propagation chain record and the start time of the next propagation chain record, and whether the propagation direction marker of the previous propagation chain record is consistent with the propagation direction marker of the next propagation chain record. Adjacent propagation chain records that meet the head-to-tail connection conditions are attached to the same link to be verified, and are sequentially connected in the attachment order to form a propagation chain record group sequence.

[0062] S2.2: Perform association validity verification on the propagation chain record grouping sequence and generate a propagation chain record set;

[0063] Furthermore, the validity of each group of adjacent propagation chain records in the propagation chain record group sequence is verified. The verification is carried out item by item, focusing on the connection relationship between upstream and downstream execution objects, the continuity relationship between connection time intervals, the connection relationship between processor occupation intervals, and the correspondence between resource access positions. Propagation chain records with clear connection relationships, continuous time intervals, connected occupation intervals, and corresponding access are retained as valid propagation chain records. Propagation chain records with broken connection, time jump, occupation separation, or access mismatch are removed or removed from the current group sequence. The retained records are then reorganized according to the original link order to generate a propagation chain record set.

[0064] S2.3: Perform continuity status checks and backtracking feature checks on the propagation chain record set through path continuity comparison operations to generate path continuity comparison records;

[0065] Furthermore, the propagation chain record set is arranged sequentially according to the business request identifier, link level position, propagation direction marker, and acceptance time interval. A continuity status check is performed segment by segment along adjacent propagation chain records to verify the continuity between upstream and downstream execution objects, the connection between acceptance time intervals, and the continuity between processor-occupied intervals. Propagation chain records with continuous acceptance, matching times, and continuous occupancy are classified into continuous propagation paths. Then, a backtracking feature check is performed segment by segment along adjacent propagation chain records in the continuous propagation path to verify whether there is a sudden drop or rise in the link level position, whether there is a backtracking switch in the propagation direction marker, and whether there is a short-term interchange in the execution object order. Propagation chain records showing backtracking signs are written into the backtracking feature record segment, and a path continuity comparison record is generated.

[0066] S2.4: Perform short-term bounce identification and cross-checking on the path continuity comparison records to obtain abnormal bounce relationships, redundant propagation relationships and invalid subordinate relationships, and perform relationship consolidation to generate a set of propagation paths to be purified;

[0067] Furthermore, the backtracking characteristic record segments in the path continuity comparison record are checked segment by segment for the acceptance time interval, the order of execution objects, the position of propagation direction switching, and the position of processor occupation change. Links with excessively short acceptance intervals, rapid backtracking in direction, and repeated falling and rising of occupation position are identified as abnormal backtracking relationships. Cross-checking is performed on adjacent propagation chain records in the continuous propagation path. Links that are repeatedly attached, repeatedly covered, or only attached to the main chain without forming an independent acceptance and advancement are identified as redundant propagation relationships and invalid subordinate relationships, respectively. Then, abnormal backtracking relationships, redundant propagation relationships, and invalid subordinate relationships are sorted according to business request identifier, link level position, and propagation direction mark to generate a set of propagation paths to be purified.

[0068] It should be noted that if the acceptance interval is too short, it is judged that the records between adjacent propagation chains have not crossed the range of adjacent sampling periods and have not formed an independent propagation segment. Links with acceptance time intervals within the range of adjacent sampling periods, rapid back-and-forth propagation directions, and repeated fall and rise of processor occupancy positions are identified as abnormal bounce relationships.

[0069] S2.5: Delete the propagation chain records corresponding to redundant propagation relationships, abnormal bounce relationships, and invalid subordinate relationships in the set of propagation paths to be purified in sequence to obtain the set of purified propagation path records;

[0070] Furthermore, the propagation chain records in the set of propagation paths to be purified are rearranged according to the business request identifier, link level position, propagation direction mark, and acceptance time interval. The propagation chain records corresponding to redundant propagation relationships, abnormal bounce relationships, and invalid subordinate relationships are located in sequence and deleted one by one according to the relationship type. The link start and end acceptance relationship, acceptance time continuity relationship, and processor occupation connection relationship are checked around the retained propagation chain records. The propagation chain records that can continue to be connected are reassembled to obtain the set of purified propagation path records.

[0071] S2.6: Fragment the set of records for the purification propagation path to generate candidate propagation path fragments.

[0072] Furthermore, the purification propagation path record set is arranged sequentially according to the business request identifier, link level position, propagation direction mark, acceptance time interval, and processor occupation interval. Each adjacent purification propagation path record is checked segment by segment to verify whether the beginning and end of the link are continuous, whether the acceptance time interval is connected, whether the processor occupation interval remains continuous, and whether the propagation direction mark is consistent. Purification propagation path records with continuous link acceptance, clear time connection, complete occupation continuity, and consistent propagation direction are merged into the same continuous propagation segment. The locations where the link acceptance is interrupted, the time jump is changed, the occupation is broken, or the propagation direction is switched are determined as the segmentation positions. Then, the purification propagation path record set is segmented according to each segmentation position, and each continuous propagation segment is organized into an independent path segment. The corresponding business request identifier, execution object link order, propagation direction mark, and acceptance time interval are added to generate candidate propagation path segments.

[0073] S3: Perform path contribution analysis on candidate propagation path segments, identify critical paths, redistribute propagation workload according to the contribution of each execution object in the critical path to the overall execution process, and generate critical path weighted load segments.

[0074] S3.1: Mark the candidate propagation path segments to form a candidate propagation path sequence;

[0075] Furthermore, the candidate propagation path segments are grouped and organized according to the business request identifier, the execution object link order, the acceptance time interval, and the propagation direction mark. Candidate propagation path segments belonging to the same business request identifier, with consecutive acceptance time intervals and consecutive execution object link orders are arranged sequentially. The starting position, continuation position, convergence position, and branch position of each candidate propagation path segment are checked one by one, and the corresponding positions are written into the path mark record. Then, based on the path mark record, the candidate propagation path segments that are adjacent to each other and have a clear path position relationship are concatenated end to end, and the concatenation order is arranged to form a candidate propagation path sequence.

[0076] S3.2: Perform path contribution analysis and sequence aggregation based on the execution object link order in the candidate propagation path sequence to generate path contribution sorting fragments;

[0077] Furthermore, the continuity, processor occupancy, hierarchical progression, and propagation direction of each execution object in the candidate propagation path sequence are examined segment by segment, from the start to the end of the link. The continuous continuity, occupancy, and progression of each execution object in the corresponding segment are recorded in the path contribution record. Based on the path contribution record, the execution objects are aggregated sequentially according to their link order, with continuously connected and fully covered execution objects placed first, and partially stopped, less covered, or less progressive execution objects placed last. The path contribution ranking value is calculated. Based on the path contribution ranking value, the continuous segments in the candidate propagation path sequence are judged by size and ordered. Continuous segments with higher path contribution ranking values ​​and continuous continuity and full coverage are assigned to the preceding segment, while continuous segments with lower path contribution ranking values ​​and partially stopped, less covered, or less progressive continuous segments are assigned to the following segment. The segments are then fragmented according to the boundary positions between the preceding and following segments to obtain the path contribution ranking segments.

[0078] The expression for calculating the path contribution ranking value is:

[0079] ;

[0080] in, Contribute ranking values ​​to the path. For the first The continuity and succession of candidate propagation path sequences. For the first Processor usage coverage of candidate propagation path sequences For the first The link-level progression of the candidate propagation path sequence. For the first The propagation direction of each candidate propagation path sequence is maintained. No. The completeness of the sequence of candidate propagation paths. For the first The degree of continuity of the candidate propagation path sequence.

[0081] It should be noted that when there is a sorting conflict between the continuous acceptance action, the continuous occupation action, and the continuous advancement action on the same execution object, the order of their positions is determined by prioritizing the continuous acceptance action, followed by the continuous occupation action, and then the continuous advancement action, so as to ensure that the path contribution sorting segment reflects the integrity of the link acceptance in a priority manner.

[0082] S3.3: Based on the path contribution ranking segments, identify critical paths, add corresponding markers, and generate a set of propagation paths;

[0083] Furthermore, based on the path contribution ranking fragment, each candidate propagation path sequence under the same business request identifier is compared one by one according to the execution object link order, the degree of acceptance, the degree of processor occupation continuity, and the degree of link advancement continuity. The candidate propagation path sequence with complete acceptance link, continuous occupation coverage, and stable advancement position is identified as the critical path. The critical path marker, critical path order marker, and critical path contribution marker are added around each execution object in the critical path. The non-critical path marker is added around the candidate propagation path sequence that is not identified as a critical path but still retains the link acceptance relationship. The propagation paths are then merged and organized according to the business request identifier and the link order to generate a propagation path set.

[0084] It should be noted that the stability of the propagation position is determined by the continuous progression of the link-level position without any short-term backtracking. Candidate propagation path sequences that maintain continuous advancement of the link-level position, continuous processor occupation coverage, and complete succession relationships are identified as critical paths. The integrity of the succession link is determined by the existence of a succession relationship between adjacent execution objects in the candidate propagation path sequence and the continuity of the start and end of the succession time interval. Continuous occupation coverage is determined by the fact that the processor occupation interval is connected along the execution object link sequence without any occupation gaps. The stability of the propagation position is determined by the continuous progression of the link-level position, the consistency of the propagation direction marking, and the absence of short-term backtracking in the execution object link sequence.

[0085] S3.4: Perform propagation service load redistribution on the propagation path set and generate a path load record set;

[0086] Furthermore, the propagation path set is grouped and organized according to business request identifier, critical path marker, execution object link order, and processor occupancy interval. The processor occupancy corresponding to the execution object in the critical path is written into the core load segment in sequence according to the critical path order marker and the critical path contribution marker. The processor occupancy of non-critical paths that are parallel to the critical path, short-term attached, or repeatedly covered is written into the auxiliary load segment and the load reduction segment, respectively. The execution link acceptance verification and occupancy interval merging are performed around each load segment under the same business request identifier. The continuously accepted and continuously occupied load segments are reattached and the allocation result marker is written. The path load record set is generated by organizing according to the execution object link order.

[0087] S3.5: Perform fragmentation and continuous interval segmentation on the path load record set to generate critical path weighted load fragments.

[0088] Furthermore, the path load records are segmented according to the business request identifier, execution object link order, allocation result marker, and processor occupancy interval in the path load record set. The path load records with consecutive processor occupancy intervals, continuous execution object link order, and consistent allocation result markers are fragmented and spliced ​​into continuous load segments. The maintenance status of critical path sequence markers, critical path contribution markers, and propagation direction markers is checked around each continuous load segment. The continuous load segments with consistent markers and continuous link continuity are segmented according to time continuous intervals and link continuous intervals to generate critical path weighted load segments.

[0089] S4: Combine and identify the weighted load segments of the critical path and model their states to form an execution behavior state feature set. Based on the execution behavior state feature set, classify the actual execution efficiency level and generate actual execution efficiency labeled segments.

[0090] S4.1: Expand the critical path weighted load segment to locate different occupied segments one by one, forming a critical path weighted load sequence;

[0091] Furthermore, when expanding the critical path weighted load segments, they are arranged sequentially according to the business request identifier, the execution object link order, the critical path order marker, and the processor occupancy interval. Adjacent critical path weighted load segments with continuous link continuity are sequentially concatenated. The processor occupancy interval change trend, critical path contribution marker change position, and propagation direction marker maintenance status are checked segment by segment around the concatenated segments. Segments with continuously increasing processor occupancy are positioned as continuously occupied segments, segments with fluctuating processor occupancy are positioned as occupancy fluctuation segments, segments with significantly decreasing processor occupancy are positioned as occupancy decrease segments, and segments with recovering processor occupancy and resuming connection are positioned as occupancy recovery segments. The segments are then arranged in order to form a critical path weighted load sequence.

[0092] S4.2: Identify segment relationships in the weighted load sequence of the critical path and generate a sequence of execution behavior combinations;

[0093] Furthermore, adjacent occupied segments in the critical path weighted load sequence are compared segment by segment according to the execution object link order, the change position of the processor occupied interval, the change position of the critical path contribution mark, and the position where the propagation direction mark remains unchanged. Adjacent occupied segments with synchronous changes in processor occupied interval and consistent critical path contribution mark are identified as synchronous segments; adjacent occupied segments with alternating fluctuations in processor occupied interval and continuous execution object link order are identified as alternating segments; and adjacent occupied segments with continuously extending processor occupied interval and unchanged propagation direction mark are identified as continuing segments. Then, synchronous, alternating, and continuing segments are concatenated and arranged according to the order of segment appearance to generate a sequence of execution behavior combinations.

[0094] S4.3: Based on the sequence of execution behaviors, establish the execution state modeling mapping relationship, and merge and refine it to generate an execution behavior state feature set;

[0095] Furthermore, based on the synchronous, alternating, and continuous segments in the execution behavior combination sequence, corresponding connections are made according to the execution object link order, the processor occupancy interval connection position, the critical path contribution marker change position, and the propagation direction marker maintenance position. The segments with continuous succession and clear change relationship are connected end to end to establish an execution state modeling mapping relationship. Around each segment in the execution state modeling mapping relationship, execution state records with the same change direction, the same succession position, and the same occupancy evolution position are merged, and continuous advancement features, local dwell features, repeated coverage features, and fallback recovery features are extracted to generate an execution behavior state feature set.

[0096] S4.4: Divide the different state chain segments in the execution behavior state feature set into actual execution efficiency levels, and add corresponding tags to generate an execution state record set;

[0097] Furthermore, different state chains within the execution behavior state feature set are checked separately according to the degree of continuous progress, the degree of processor usage continuity, the degree of repetition coverage, the degree of fallback recovery, and the degree of retention of critical path contribution markers. The results of each sub-item check are written into the execution efficiency judgment record. A comprehensive comparison is made of the execution efficiency judgment records within the same state chain segment. State chains with high degree of continuous progress, high degree of processor usage continuity, low degree of repetition coverage, low degree of fallback recovery, and stable critical path contribution markers are classified into the high true execution efficiency level. State chains with medium degree of continuous progress, continuous processor usage, limited degree of repetition coverage, and controllable degree of fallback recovery are classified into the medium true execution efficiency level. State chains with low degree of continuous progress, insufficient degree of processor usage continuity, significant degree of repetition coverage, or high degree of fallback recovery are classified into the low true execution efficiency level. Then, true execution efficiency level markers and level switching position markers are added to each state chain segment, and the execution state record set is generated by organizing them according to the execution object link order.

[0098] It should be noted that the degree of continuous advancement is determined by the coverage of continuous advancement segments in all segments; the degree of processor occupation continuity is determined by the connection between the beginning and end of the processor occupation interval; the degree of repeated coverage is determined by the proportion of repeatedly occupied segments in all segments; the degree of fallback recovery is determined by the segments traversed by the state chain segment from the occupation fallback position to the continuous acceptance position; state chain segments that meet the high-level judgment conditions are classified as high actual execution efficiency level, state chain segments that meet the medium-level judgment conditions are classified as medium actual execution efficiency level, and the remaining state chain segments are classified as low actual execution efficiency level.

[0099] S4.5: The execution status record set is labeled, organized, and concatenated to obtain the actual execution efficiency labeled record set, and then segmented into fragments to generate actual execution efficiency labeled fragments.

[0100] Furthermore, the execution status record set is arranged sequentially according to the business request identifier, execution object link order, actual execution efficiency level mark, and processor occupancy interval. Adjacent execution status records with consistent actual execution efficiency level marks, continuous execution object link order, and connected processor occupancy intervals are annotated, organized, and spliced ​​together to obtain the actual execution efficiency annotated record set. Around the continuous annotated segments in the actual execution efficiency annotated record set, the preservation status of critical path sequence marks, critical path contribution marks, and propagation direction marks is checked. Fragmentation is performed according to the time continuity position and link continuity position to generate actual execution efficiency annotated segments.

[0101] S5: Based on the processor usage in the actual execution efficiency annotation segments, perform separation, hierarchical accumulation and differential inclusion to generate a structured load result, and perform consistency verification to generate a unified verified load result.

[0102] S5.1: Based on the actual execution efficiency, the processor usage in the labeled segments is grouped, linked together, and separated to generate a set of processor usage records;

[0103] Furthermore, based on the business request identifier, execution object link order, actual execution efficiency level marker, critical path order marker, critical path contribution marker, and processor occupancy interval in the actual execution efficiency labeled segments, the actual execution efficiency labeled segments belonging to the same business request identifier and with consecutive execution object link orders are rearranged according to the order of processor occupancy intervals, and adjacent actual execution efficiency labeled segments that are consecutive and have clear occupancy connections are joined together; around each consecutive segment after joining, the actual execution efficiency level marker, critical path order marker, and critical path contribution marker are separated and organized, and the processor occupancy corresponding to different occupancy segments is written into the corresponding record segments to generate a processor occupancy record set.

[0104] S5.2: The processor occupancy record set is hierarchically accumulated and differentially included, and continuous load segments are spliced ​​to obtain the load structured result;

[0105] Furthermore, the processor occupancy record set is hierarchically categorized according to business request identifier, execution object link order, actual execution efficiency level marker, and processor occupancy range. Records with continuous processor occupancy ranges and consistent execution object link order within the same level are cumulatively organized to form continuous occupancy segments. Around continuous occupancy segments at different levels, differentiated inclusion is performed based on actual execution efficiency level marker and critical path contribution marker: high-level occupancy segments are included as core inclusion segments, medium-level occupancy segments as auxiliary inclusion segments, and low-level occupancy segments as restricted inclusion segments. Core inclusion segments are written to the load consolidation segment with full retention, auxiliary inclusion segments are written to the load consolidation segment with proportional retention, and restricted inclusion segments are written to the load consolidation segment with reduced retention. The load consolidation segments are then concatenated into continuous load segments according to business request identifier and execution object link order, and the concatenated continuous load segments are mapped to a unified normalized range to obtain the load structured result.

[0106] It should be noted that the same level is determined by consistent actual execution efficiency level marking and corresponding execution object link order; when the same continuous occupied segment meets multiple level inclusion conditions at the same time, differentiated inclusion is performed in the order of core inclusion segment first, auxiliary inclusion segment second, and restricted inclusion segment third; the full retention method means that the corresponding processor occupation is fully included in the load consolidation segment, the proportional retention method means that the corresponding processor occupation is included in the load consolidation segment according to the preset retention ratio, and the reduced retention method means that the corresponding processor occupation is included in the load consolidation segment according to the preset reduction ratio; the unified normalization interval is determined according to the preset benchmark load range to ensure the comparability of load structured results under different business request identifiers.

[0107] S5.3: Perform consistency verification and correction on the structured load results to generate a unified verified load result.

[0108] Furthermore, the structured load results are subjected to consistency verification based on business request identifiers, execution object link order, actual execution efficiency level markers, critical path order markers, and processor occupancy intervals. Each segment is checked to verify whether the execution object link order is continuous, whether the actual execution efficiency level markers correspond to the critical path contribution markers, whether the processor occupancy intervals are connected end to end, and whether there are hierarchical misalignments and boundary conflicts between continuous load segments. For load records with sequence misalignments, marker conflicts, or interval breaks, the execution position is called back, the boundary is straightened, and the hierarchy is corrected. The verified load records and the corrected load records are then reassembled and reorganized in a unified order to generate a unified verified load result.

[0109] It should be noted that the hierarchical correction is carried out by recalibrating the actual receiving position, inclusion level and critical path level of the continuous load segment, and moving the continuous load segments with misaligned, overlapping or out-of-place hierarchical positions back to their corresponding hierarchical positions, so that each continuous load segment maintains hierarchical consistency.

[0110] In summary, this invention achieves processor occupancy attribution correction and execution quality differentiation through path contribution analysis, critical path identification, propagation business load redistribution, and actual execution efficiency level classification, thereby improving the accuracy of business main chain occupancy identification and effective load expression. Through processor occupancy separation, layered accumulation, differentiated inclusion, and consistency verification, it achieves unified load verification result generation, thereby improving the reliability and comparability of system load assessment results.

[0111] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for normalizing CPU effective load data collection for system load assessment, characterized in that, include: Collect raw load sampling fragments and perform layer-by-layer association to establish a propagation chain mapping from business requests to execution objects, forming business-related load fragments; Based on the propagation chain records in the business-related load segments, the association validity is verified, the path continuity is compared, and short-term bounce is identified. Redundant propagation relationships, abnormal bounce relationships, and invalid subordinate relationships are removed to generate candidate propagation path segments. Perform path contribution analysis on candidate propagation path segments, identify critical paths, and redistribute propagation workload according to the contribution of each execution object in the critical path to the overall execution process, generating critical path weighted workload segments. The weighted load segments of the critical path are combined, identified, and modeled to form an execution behavior state feature set. Based on the execution behavior state feature set, the actual execution efficiency level is divided, and the actual execution efficiency labeled segments are generated. Based on the processor usage in the actual execution efficiency annotation segments, the data is separated, layered, accumulated, and differentially incorporated to generate a structured load result. Consistency verification is then performed to generate a unified verified load result.

2. The CPU effective load acquisition normalization method for system load assessment as described in claim 1, characterized in that, The formation of the business-related load fragment is as follows: The original load sampling segments are spliced ​​together to form business request threads, and the order is made consistent to generate a set of request-related segments; Based on the set of request-related fragments, layer-by-layer association processing, propagation direction identification, and splicing of the beginning and end of the association links are performed to generate a propagation chain mapping, forming business-related load fragments.

3. The CPU effective load acquisition normalization method for system load assessment as described in claim 1, characterized in that, The specific steps for performing the association validity check are as follows: Perform head-to-tail connection checks on adjacent propagation chain records in the business-related load segments and attach them to the same link to be checked to form a propagation chain record group sequence; The validity of the propagation chain record grouping sequence is verified, and a propagation chain record set is generated.

4. The CPU effective load acquisition normalization method for system load assessment as described in claim 1 or 3, characterized in that, The process of performing association validity verification, path continuity comparison, and short-term bounce identification is as follows: The propagation chain record set is checked for continuity status and backtracking features by performing path continuity comparison operations, thereby generating path continuity comparison records. Short-term bounce identification and cross-checking are performed on the path continuity comparison records to obtain abnormal bounce relationships, redundant propagation relationships and invalid subordinate relationships. The relationships are then consolidated to generate a set of propagation paths to be purified.

5. The CPU effective load acquisition normalization method for system load assessment as described in claim 4, characterized in that, The generation of candidate propagation path segments is as follows: The propagation chain records corresponding to redundant propagation relationships, abnormal bounce relationships, and invalid subordinate relationships in the set of propagation paths to be purified are deleted in sequence to obtain the set of purified propagation path records. The set of records of purification propagation paths is segmented into fragments to generate candidate propagation path fragments.

6. The CPU effective load acquisition normalization method for system load assessment as described in claim 1, characterized in that, The identification of the key path is as follows: The candidate propagation path segments are marked to form a candidate propagation path sequence; Path contribution analysis and sequence aggregation are performed based on the execution object link order in the candidate propagation path sequence to generate path contribution sorting fragments.

7. The CPU effective load acquisition normalization method for system load assessment as described in claim 6, characterized in that, The generation of the critical path weighted load segment is as follows: Based on the path contribution ranking segments, key paths are identified, and corresponding markers are added to generate a set of propagation paths; The propagation service load is redistributed across the propagation path set, generating a path load record set. The path load record set is fragmented and continuous intervals are divided to generate critical path weighted load fragments.

8. The CPU effective load acquisition normalization method for system load assessment as described in claim 7, characterized in that, The formation of the execution behavior state feature set is as follows: Expand the critical path weighted load segment to locate different occupied segments one by one, forming a critical path weighted load sequence; Segment relationships are identified in the weighted load sequence of the critical path to generate a sequence of combined execution behaviors; Based on the sequence of execution behaviors, an execution state modeling mapping relationship is established, and then merged and refined to generate a set of execution behavior state features.

9. The CPU effective load acquisition normalization method for system load assessment as described in claim 1 or 8, characterized in that, The specific steps for generating the actual execution efficiency labeled segment are as follows: The different state chain segments in the execution behavior state feature set are divided into actual execution efficiency levels, and corresponding tags are added to generate an execution state record set; The execution status record set is labeled, organized, and concatenated to obtain the actual execution efficiency labeled record set. Then, it is segmented into fragments to generate actual execution efficiency labeled fragments.

10. The CPU effective load acquisition normalization method for system load assessment as described in claim 9, characterized in that, The generation of the unified verification load result is as follows: Based on the processor usage in the actual execution efficiency labeled segments, the processor usage is grouped, linked together, and separated to generate a set of processor usage records; The processor usage record set is hierarchically accumulated and differentiated, and continuous load segments are spliced ​​to obtain a structured load result; The structured load results are validated and corrected for consistency, generating a unified validated load result.

Images