Analysis method of storage system and electronic device

By constructing a resource relationship model and combining event logs with performance metrics data, the system automatically diagnoses storage system performance bottlenecks, solving the problems of low efficiency and poor accuracy in storage system performance analysis, and achieving fast, comprehensive, and accurate performance analysis and optimization.

CN120892310BActive Publication Date: 2026-01-27INSPUR SUZHOU INTELLIGENT TECH CO LTD
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
CN202511415366.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2025-09-29
Publication Date
2026-01-27
Estimated Expiration
2045-09-29

AI Technical Summary

Technical Problem

Existing storage system performance analysis is inefficient and inaccurate, making it difficult to quickly and accurately identify performance anomalies and optimize configurations.

Method used

By constructing a resource relationship model and combining event logs with performance metric data for correlation analysis, the system can automatically diagnose storage system performance bottlenecks and generate performance correlation analysis reports.

Benefits of technology

It enables rapid, comprehensive, and accurate correlation analysis of storage system performance, improving the efficiency and accuracy of performance analysis and guiding operation and maintenance personnel in performance optimization and fault repair.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120892310B_ABST
    Figure CN120892310B_ABST
Patent Text Reader

Abstract

The analysis method of the storage system and the electronic equipment of the application relate to the technical field of storage systems, and through a pre-constructed resource association model containing logical association relationships of all resource objects in a service link, log information analysis in an event dimension and time delay index analysis in a performance dimension are performed on a target object and its associated objects, the problems of low efficiency and poor accuracy of traditional troubleshooting methods are overcome, fast, comprehensive and associated analysis of the performance of the storage system is realized, and the efficiency and accuracy of performance analysis are significantly improved. Therefore, the technical problem of how to quickly and accurately analyze the performance of the storage system can be solved, and the technical effect of improving the efficiency and accuracy of performance analysis of the storage system is achieved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This application relates to the field of storage system technology, and in particular to an analysis method and electronic device for a storage system. Background Technology

[0002] In the performance analysis of storage systems, since storage systems contain multiple hardware and software modules, the key points are how to quickly find performance anomalies, identify the performance correlation between modules, discover problems, and find optimization directions.

[0003] In related technologies, the performance analysis of storage systems usually involves manually checking the log information or performance data of faulty modules and then analyzing the log information or performance data. This results in low efficiency and accuracy of performance analysis. Therefore, how to perform fast and accurate performance analysis of storage systems is an urgent problem to be solved. Summary of the Invention

[0004] This application provides a method and electronic device for analyzing storage systems, in order to at least address the problem of how to quickly and accurately perform performance analysis of storage systems in related technologies.

[0005] This application provides a method for analyzing a storage system, including:

[0006] In response to the analysis command for the target object, based on the resource relationship model, the log information of the target object and at least one associated object is analyzed and processed to obtain the event analysis results. The resource relationship model is a model constructed based on the relationship between each resource object in the business link in the system to be analyzed. It includes at least the logical relationship between each resource object. The associated object is a resource object that has a logical relationship with the target object, as determined by the resource relationship model.

[0007] Based on the resource relationship model, performance indicators are calculated and processed according to the latency data of at least one associated object to obtain the latency indicators of at least one associated object.

[0008] Based on latency metrics, perform performance analysis on at least one related object to obtain performance analysis results;

[0009] Based on the event analysis results and performance analysis results, a performance correlation analysis report is generated.

[0010] This application also provides an analysis apparatus for a storage system, comprising:

[0011] The first analysis unit is used to respond to the analysis command for the target object, and perform information analysis and processing on the log information of the target object and at least one associated object based on the resource relationship model to obtain the event analysis results. The resource relationship model is a model constructed based on the relationship between each resource object in the business link in the system to be analyzed, and includes at least the logical relationship between each resource object. The associated object is a resource object that has a logical relationship with the target object as determined by the resource relationship model.

[0012] The calculation unit is used to perform performance index calculation based on the resource relationship model and the latency data corresponding to at least one associated object, so as to obtain the latency index corresponding to at least one associated object.

[0013] The second analysis unit is used to perform performance analysis on at least one related object based on latency indicators and obtain performance analysis results.

[0014] The generation unit is used to generate a performance correlation analysis report based on the event analysis results and performance analysis results.

[0015] This application also provides an electronic device, comprising: a memory for storing a computer program; and a processor for implementing the analysis method of any of the above-described memory systems when executing the computer program.

[0016] This application also provides a computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the steps of the analysis method of any of the above-described storage systems.

[0017] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the analysis method for any of the above-described storage systems.

[0018] The storage system analysis method and electronic device disclosed in this application, through a pre-built resource association model containing the logical relationships of all resource objects in the business chain, perform event-dimensional log information analysis and performance-dimensional latency index analysis on the target object and its associated objects. This overcomes the inefficiency and poor accuracy of traditional troubleshooting methods, achieving rapid, comprehensive, and correlated analysis of storage system performance, significantly improving the efficiency and accuracy of performance analysis. Therefore, it solves the technical problem of how to quickly and accurately perform performance analysis of storage systems, achieving the technical effect of improving the efficiency and accuracy of storage system performance analysis. Attached Figure Description

[0019] To more clearly illustrate the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 A flowchart illustrating an analysis method for a storage system provided in an embodiment of this application;

[0021] Figure 2 A flowchart illustrating an event-dimensional analysis provided in an embodiment of this application;

[0022] Figure 3 This is a schematic diagram of a performance index calculation process provided in an embodiment of this application;

[0023] Figure 4 A schematic diagram of a performance dimension analysis provided for an embodiment of this application;

[0024] Figure 5 A schematic diagram of the structure of an analysis device for a storage system provided in an embodiment of this application;

[0025] Figure 6 This is a schematic diagram of the structure of an analysis device for another storage system provided in an embodiment of this application. Detailed Implementation

[0026] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the protection scope of this application.

[0027] It should be noted that, in the description of this application, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. The terms "first," "second," etc., in this application are used to distinguish similar objects and are not used to describe a specific order or sequence.

[0028] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0029] The storage system analysis method described in this application comprehensively diagnoses storage system performance bottlenecks by combining event logs with performance metric data for correlation analysis. It systematically organizes and analyzes various resource objects in the storage system through a pre-built resource relationship model, thereby efficiently and accurately locating the root cause of problems.

[0030] Figure 1 This document provides a flowchart illustrating a storage system analysis method as an embodiment of the present application. The method is described in detail below, taking into account the execution flow of the storage system analysis method.

[0031] like Figure 1 As shown, the data processing method includes:

[0032] Step 101: In response to the analysis instruction for the target object, based on the resource relationship model, perform information analysis and processing on the log information of the target object and at least one associated object to obtain the event analysis result. The resource relationship model is a model constructed based on the relationship between each resource object in the business link of the system to be analyzed, and includes at least the logical relationship between each resource object. The associated object is a resource object that has a logical relationship with the target object as determined by the resource relationship model.

[0033] In the embodiments of this application, the resource relationship model is a data structure constructed to accurately depict the static logical associations and dynamic dependencies of all resource objects in the business chain of the system to be analyzed. The system to be analyzed refers to a complete information technology (IT) infrastructure environment containing computing, network, and storage resources, such as a Storage Area Network (SAN) system composed of host servers, switches, and storage arrays. Resource objects are logical or physical entities in the system to be analyzed that can be independently managed, monitored, and have performance metrics, such as storage pools, physical hard drives, logical volumes, host adapter ports, switch ports, servers, etc.

[0034] Logical relationships refer to the various relationships between resource objects beyond physical connections, including those formed based on software configuration, protocol logic, or service mapping. For example, a logical volume belongs to a storage pool, a server is mapped to a logical volume, or multiple switch ports belong to the same Fibre Channel zone. By capturing these diverse logical relationships, the resource relationship model forms a topology map covering the entire storage system path, providing a basis for retrieval and traversal in relational analysis.

[0035] The target object is typically any resource object carried in the analysis command that may present a performance bottleneck or requires in-depth analysis, such as a specific logical volume exhibiting high latency. When the analysis command is triggered, a query is performed using the resource relationship model. Based on the defined logical relationships, the model automatically retrieves and identifies all related objects that are directly or indirectly associated with the target object. For example, for a logical volume as the target object, its related objects might include: its storage pool, other logical volumes within the same storage pool, its Redundant Array of Independent Disks (RAID) group, its corresponding physical hard drive, the mapped host, the host port, the connected switch and its port, and other ports within the same switch zone.

[0036] Next, log information will be collected for the target object and all related objects. Log information refers to file data that records events and state changes generated by various resource objects in the storage system during operation. This includes, but is not limited to: alarm information: immediate messages indicating abnormal situations such as component errors, performance threshold exceeding limits, or configuration conflicts; system logs: operation details, error reports, and state change information recorded by the operating system or device, such as the system log of a Linux system or the error log of a network switch; audit logs: historical records of user or system operations such as adding, deleting, modifying, and querying resource configurations.

[0037] Information analysis and processing involves filtering, merging, and correlating log information. Different analysis time windows are set based on different log types and sources (e.g., reviewing recent configuration change logs within the last 24 hours, and system error logs within the last 2 hours). Pattern matching and keyword filtering are then used to extract information about anomalies. Based on this information, the resource objects involved in the anomalies are located within the target and related objects, resulting in event analysis results. These results summarize the abnormal events, error records, and key operations that occurred in the relevant resource objects throughout the entire business chain before and after the problem occurred, providing potential event-based clues and evidence for performance issues.

[0038] Step 102: Based on the resource relationship model, calculate and process the performance indicators according to the latency data corresponding to at least one associated object to obtain the latency indicators corresponding to at least one associated object.

[0039] In the embodiments of this application, latency data refers to performance time-series data collected from the storage system or associated object itself, reflecting the length of input / output (I / O) operation processing time, such as read / write latency of logical volumes, port latency of switch ports, etc.

[0040] The purpose of performance index calculation and processing is to transform the raw, potentially noisy, time series latency data into one or more scalar values ​​that can represent the overall performance characteristics, i.e., latency indices. These indices can extract key information reflecting the latency characteristics of related objects from the raw data, avoiding analytical difficulties caused by the disorganization of the raw data. For example, to determine the degree of resource consumption of related objects, their latency data needs to be sorted and percentage-calculated; to determine the degree of latency fluctuation of related objects, the standard deviation of the latency data needs to be calculated; and to determine the performance correlation between related objects and the target object, correlation analysis of their latency data is required.

[0041] Meanwhile, the performance index calculation process is not arbitrary, but is determined based on the logical relationship between related objects and target objects in the resource relationship model. For example, for storage pools that have a hierarchical relationship with the target object, there is no need to calculate the index; it is directly determined that they have an impact on the target object. For volumes in the same pool that have a peer relationship with the target object, the latency correlation index between them and the target object is calculated to analyze the volumes in the same pool that have an impact on the target object.

[0042] It should be noted that when calculating performance metrics, only related objects that have a sibling logical relationship with the target object will be calculated. For related objects that have a hierarchical logical relationship with the target object, there is no need to calculate or analyze performance metrics. They will be directly identified as factors affecting the target object and specific resource analysis will be performed.

[0043] Step 103: Perform performance analysis on at least one associated object based on the latency index to obtain the performance analysis results.

[0044] In the embodiments of this application, performance analysis processing refers to comparing and ranking the latency metrics of related objects side-by-side, filtering out the related objects that have the greatest impact on the performance of the target object from a performance perspective, and clarifying the specific characteristics of their performance anomalies. For example, the related objects with the highest I / O latency (i.e., the most active resource consumption points) can be identified by ranking them according to their latency metrics; or statistical correlation analysis can be used to find related objects whose performance fluctuation curves are highly positively or negatively correlated with the target object, thereby indicating potential common bottlenecks or resource competition relationships. The performance analysis results indicate the related objects that may cause performance problems of the target object from a performance perspective.

[0045] Step 104: Generate a performance correlation analysis report based on the event analysis results and performance analysis results.

[0046] In the embodiments of this application, event analysis results reflect event factors that may affect the performance of the target object (e.g., alarms, abnormal operations), while performance analysis results reflect performance factors that may affect the performance of the target object (e.g., resource consumption and performance fluctuations of related objects). Both reveal the potential root causes of performance problems in the target object from different dimensions. However, each type of result has limitations. For example, relying solely on event analysis results cannot determine whether a particular event truly caused the performance anomaly (e.g., a switch port has an error count log, but the error count may not yet have a significant impact on latency); relying solely on performance analysis results cannot determine what event triggered the performance anomaly of a related object (e.g., a storage pool experiences severe resource consumption, but it cannot be determined whether this is due to increased business load or disaster recovery operations). Therefore, integrating the two types of results to generate a performance correlation analysis report enables comprehensive and accurate location of performance problems.

[0047] This application organically integrates system topology, event logs, and performance indicators through a resource relationship model, providing an automated and systematic method for storage system performance analysis. This improves the efficiency and accuracy of performance diagnosis and can quickly generate in-depth analysis reports containing multi-dimensional evidence, effectively guiding operations and maintenance personnel in performance optimization and fault repair.

[0048] In one possible implementation of this application, the construction of a resource relationship model is the foundation for the accurate and comprehensive execution of all analysis steps. Without a clear and complete resource association logic, subsequent log analysis and performance indicator calculations of the target object and associated objects will lack a clear analytical scope and basis. Therefore, before performing event analysis and performance analysis on the target object, it is necessary to first complete the construction of the resource relationship model.

[0049] Regarding the construction of the resource relationship model, the following methods may also be used, but are not limited to: obtaining the first resource on the storage side of the system to be analyzed, and the second resource on the storage network link, wherein the first resource and the second resource include at least each resource object and its association; based on the object identifier corresponding to each resource object, the first resource, and the second resource, constructing a resource relationship model that includes at least logical associations, wherein the logical associations include at least the hierarchical logical associations and the peer logical associations between each resource object.

[0050] In the embodiments of this application, the system to be analyzed includes at least a storage side and storage network links. The storage side refers to the subsystem in the system to be analyzed that provides persistent data storage functions, such as a storage array, disk enclosure, and storage management software running on it. The first resource refers to the various resource objects that can be identified and managed within this storage side and the relationships between them, specifically including but not limited to entities such as logical volumes, storage pools, redundant disk groups (MDiskgrp), physical hard drives, storage controllers, and storage nodes, as well as their ownership, mapping, and other relationships.

[0051] Storage network links refer to the network pathways connecting compute nodes (hosts) and storage, primarily consisting of fiber optic switches, host bus adapters, etc. Secondary resources refer to the resource objects existing on this network link and their relationships, such as the physical ports of the fiber optic switch, the Fibre Channel partitions to which those ports are joined, the host bus adapters, and the host itself. The combination of primary and secondary resources encompasses all the resource objects necessary for end-to-end performance correlation analysis and their interrelationships.

[0052] An object identifier is a unique symbol or code that identifies a resource object, such as a logical volume number, a switch port index, or a host name. It serves as an index in the model that associates different resource objects with each other.

[0053] The constructed resource relationship model must contain at least logical associations, including hierarchical and sibling logical associations between resource objects. Hierarchical logical associations describe vertical, non-peer-to-peer subordinate or containment relationships, defining the position of resource objects within the hierarchical structure. For example, a physical hard drive belongs to a redundant disk group, which in turn belongs to a storage pool, while logical volumes are allocated space by the storage pool; similarly, logical volumes are mapped to hosts, which may belong to a host cluster. This chain of relationships clearly outlines the hierarchical topology of data storage.

[0054] Sibling logical relationships describe horizontal, peer-to-peer, parallel, or collaborative relationships, indicating resource objects located at the same level or logical group that may share resources or influence each other. For example, all logical volumes located in the same storage pool are sibling logical relationships (i.e., pooled volumes), sharing the same set of physical hard disk resources, and may experience I / O contention; another example is all switch ports assigned to the same Fibre Channel zone, which are sibling logical relationships, allowing them to communicate with each other, and an anomaly in any port may affect the performance of other ports within the zone.

[0055] Specifically, the construction of the resource relationship model can be achieved through, but is not limited to, the following methods: 1) The entry point for the resource relationship model can be chosen to start with a volume. The purpose is that in storage services, storage services can only be truly provided by mapping a volume to a specific host. The service also ultimately lands on the actual volume through layers of mapping. Therefore, the volume is the core logical unit that fits the storage service. 2) In addition to saving the upper and lower layer relationships of the volume itself, it is also necessary to save the information of volumes in the same pool and volume copy information. In addition to saving the switch port information in the network link, it is also necessary to save the switch port information of the switch and the switch under the Zone. 3) To ensure real-time accuracy, in addition to using a timer mechanism to refresh the resource relationship model, it is also necessary to bind the resource relationship model to the storage resources. When the storage resources change, such as when they are added, deleted, or modified, the resource relationship model should be refreshed synchronously.

[0056] The structure of the resource relationship model can be represented in, but is not limited to, the following ways:

[0057] [Volume - Possessing Pool - Physical Disk (drive) under Possessing Management Disk (mdisk) - Volume Group - Mapping Host - Mapping Host's Host Group - Mapped Host Port - Mapped Logical Unit Number (LUN) - Switch Port - Same Zone Switch Port Group - Same Pool Volume - Volume Copy - Possessing Node - Possessing Controller - Possessing Cluster].

[0058] This application systematically integrates primary and secondary resources and accurately models their hierarchical and sibling logical relationships. The constructed resource relationship model forms a semantically rich topological graph covering the entire data path stack. This resource relationship model enables the analysis engine to instantly perceive all potentially related objects of any target object, whether vertical dependencies or horizontal influences. It provides a data foundation for in-depth, cross-component event and performance correlation analysis, thereby improving the automation and coverage of storage system performance analysis.

[0059] In one possible implementation of this application embodiment, when performing information analysis processing on the log information of the target object and at least one associated object, the following methods can be used, but are not limited to: responding to an analysis instruction to analyze the performance of the target object within a target time period; extracting the log information corresponding to the target object and at least one associated object within the target time period from a preset database, wherein the target time period is not greater than the analysis time threshold corresponding to the target object set in the preset database, the preset database includes object information corresponding to each resource object, and the object information includes log information; extracting change information from the log information, wherein the change information includes at least the state change status of each resource object in the system to be analyzed; performing retrieval processing in the resource relationship model based on the change information to determine the changed object among the target object and at least one associated object, and obtaining the event analysis result, wherein the changed object is a resource object with a state change.

[0060] In the embodiments of this application, the analysis instruction is a request initiated by a user or system, which explicitly includes the target object to be investigated (e.g., a specific logical volume) and the target time period of interest (e.g., a specific period of performance degradation).

[0061] The default database is a centralized data repository, pre-designed and populated to persistently store and manage data related to various resource objects in the system under analysis. The object information contained in this default database is a comprehensive collection, including not only static configuration attributes of resources but also dynamically changing log information. Log information is continuously collected from each resource object and stored in the database, recording various events, alarms, and status changes during its operation.

[0062] It should be noted that the length of the requested target time period cannot exceed the preset analysis time threshold set in the database for that target object or its type. This analysis time threshold is pre-configured based on the validity and relevance of different types of events; for example, it might be set to 24 hours for configuration change logs and 2 hours for system error logs. This ensures that the analysis focuses on the most recent events that are most likely related to the current performance issue, avoiding the performance overhead and information interference caused by processing massive amounts of historical data.

[0063] After successfully extracting log information, the process enters the information refinement stage, which involves extracting change information from the log information. Change information is a record that represents a change in the system state, parsed and filtered from the original log entries. It includes at least the state changes of each resource object in the system to be analyzed. State changes include, but are not limited to, any changes to the configuration state, running state, or health state of resource objects. Specifically, it includes, but is not limited to: a) Configuration changes: such as the creation and deletion of storage volumes, adjustment of LUN mappings, changes in host cluster members, updates to switch zone configurations, etc.; b) Health state changes: such as a physical hard drive changing from online to faulty, a switch port changing from normal to disabled due to errors, warnings triggered by any component, error or severity level alarms, etc.; c) Running state changes: such as system service restarts, path switching (in a multi-path environment), failover of redundant components, etc.

[0064] Finally, the extracted change information is retrieved and processed within the resource relationship model to identify the target object and at least one related object that has undergone changes, thus obtaining the final event analysis results. The retrieval process is based on the query and association process of the resource relationship model graph. Utilizing the hierarchical and sibling logical relationships defined in the resource relationship model, all resource objects that have undergone status changes within the target time period are quickly located. The identified resource objects with status changes are the changed objects. For example, analysis reveals that around the time the target object (a logical volume) experiences performance anomalies, a port on a fiber optic switch associated with it generates a large number of Cyclic Redundancy Check (CRC) errors (health status change). Simultaneously, another member volume in its storage pool (a sibling logical relationship) has just completed a snapshot operation (configuration change). This changed object and its specific change information together constitute the content of the event analysis results.

[0065] Specifically, the information analysis and processing of log information, namely the process of performing event-dimensional correlation analysis on the storage system, can be achieved through, but is not limited to, the following methods: Event-dimensional analysis mainly analyzes log information, including relevant alarms, system logs, and operation logs on the storage side, host, and switch. Log information is divided into the following categories: ① Storage-side events: alarm information generated by the storage itself; CRUD operation logs of the target object and related objects; disaster recovery operation content of the target object and related objects, such as: snapshots, remote replication, disaster recovery, local replication, and volume copy information. The analysis window for this type of information is within 24 hours. ② Switch events: Alarm information generated by the switch itself, analyzed within 24 hours before and after; error counts of the current port and ports within the same zone of the switch (e.g., the portterrshow command), this type of information can be analyzed for current output. ③ Host events: Host system logs, including system logs (message) and kernel logs (dmesg), with an analysis window of 2 hours; custom logs on the host, mainly for scenarios with host multipath configuration. Usually, if there is a host mapping service, there will be such logs. If open-source multipath is used, the logs may be integrated into the message and dmesg. If the storage vendor's self-developed multipath is used, it will be recorded in the storage vendor's custom file, with a retrieval window of 2 hours.

[0066] Specifically, regarding log information, Table 1 provides an example:

[0067] Table 1

[0068]

[0069] Furthermore, to facilitate understanding of the specific operational process of event dimension analysis, this application embodiment also provides a flowchart of event dimension analysis, such as... Figure 2 As shown, it includes:

[0070] S1: When a user initiates performance association, they specify the object to be analyzed and the time period; S2: Based on the specified time, retrieve the event alarm logs of the specified object and related objects in the corresponding time window of the preset database; S3: According to the change information in the event alarm logs, search in the resource association model and extract all changed objects; S4: Finally, obtain the associated event analysis results.

[0071] This application, by combining with a resource relationship model, achieves a transformation from large-scale log information retrieval to precise change event localization. It automates the tedious log correlation and investigation process, not only improving retrieval efficiency but also ensuring the comprehensiveness of correlations through the model, avoiding omissions, and providing an event evidence chain for accurately and quickly locating the root cause of performance problems.

[0072] In one possible implementation of this application embodiment, regarding the process of calculating performance indicators based on latency data, this application embodiment also provides a schematic flowchart of the performance indicator calculation process, such as... Figure 3 As shown, it includes:

[0073] Step 301: Obtain multiple delay data corresponding to at least one first associated object within the target time period, wherein the first associated object is an associated object that has a logical relationship with the target object at the same level among at least one associated object.

[0074] In the embodiments of this application, the first associated object is a specific subset of objects, defined as an associated object among at least one associated object that has a sibling logical relationship with the target object. A sibling logical relationship refers to objects located at the same level or logical group that may share resources or have competing relationships, such as other logical volumes under the same storage pool, or other switch ports under the same Fibre Channel zone. The first associated object is selected for analysis because it shares the same underlying resources with the target object (e.g., the same physical hard disk group, network bandwidth, or processor resources), and its performance often has the strongest comparability and correlation, making it key to diagnosing resource competition or common bottlenecks. Latency data is raw time-series data collected from the performance monitoring system within a specified target time period, reflecting the input / output operation processing latency of each first associated object.

[0075] Step 302: Perform data selection processing on multiple time-delay data to determine the activity value corresponding to at least one first associated object.

[0076] In the embodiments of this application, the data selection processing is a statistical screening method, the purpose of which is to identify indicators that can represent the level of performance loss suffered by an object during the vast majority of normal operations. Specifically, the activity value can be defined by a percentile-based calculation method. The activity value characterizes the critical point of a high latency level for a first associated object in its latency data sample. The higher the value, the greater the consumption of shared resources by the first associated object within a specified time period, or the more saturated its own processing capacity, that is, the higher its activity level, and the more likely it is to become the focus of resource competition and affect the performance of the target object.

[0077] Step 303: Perform standard deviation calculation on multiple time delay data to obtain the oscillation value corresponding to at least one first associated object.

[0078] In the embodiments of this application, the standard deviation calculation is used to quantify the degree to which a set of data points deviate from their mean. The oscillation value is the result of this calculation and is a non-negative scalar value. The magnitude of the oscillation value directly reflects the stability of the performance of the corresponding first associated object. A high oscillation value indicates that the latency data of the first associated object fluctuates very drastically, and its performance is extremely unstable, sometimes good and sometimes bad; while a low oscillation value indicates that its performance is very stable. Identifying objects with high oscillation values ​​is crucial for discovering intermittent performance problems, unstable hardware components, or resources subjected to sudden interference.

[0079] Specifically, the calculation of oscillation values ​​can also be achieved in the following ways, but not limited to: First, select multiple time-delay data points for each first associated object within the target time period, and calculate the average value of the time-delay data corresponding to each first associated object. Next, calculate the variance of the latency data corresponding to each first associated object, and then the standard deviation of the latency data corresponding to each first associated object. (i.e., oscillation value).

[0080] Specifically, it can be calculated using, but is not limited to, formula (1):

[0081] Formula (1)

[0082] in, The first associated object of the target One latency data, The number of latency data corresponding to the first associated object of the target. The oscillation value corresponding to the first associated object of the target. The average value of the latency data corresponding to the first associated object of the target.

[0083] Step 304: Perform correlation coefficient analysis on the target object with at least one first associated object to obtain the correlation coefficient values ​​between the target object and at least one associated object. The time delay indicators include the activity value, the oscillation value and the correlation coefficient value.

[0084] In the embodiments of this application, correlation coefficient analysis is a method used to measure the strength and direction of the linear relationship between two variables. Specifically, it calculates the correlation coefficient between the latency data of the target object and the latency data of each first associated object. The calculated correlation coefficient value is a value between -1 and 1. The sign of the correlation coefficient value indicates the direction of the relationship (positive or negative correlation), while the magnitude of the absolute value indicates the strength of the relationship. Ultimately, the latency index serves as a comprehensive performance profile, including the calculated activity value, oscillation value, and correlation coefficient value.

[0085] Specifically, the calculation of the correlation coefficient can be achieved in the following ways, but not limited to: using the Pearson correlation coefficient for analysis. The Pearson correlation coefficient can reflect the correlation analysis between two linear indicators. The value is [-1, 1]. A value of 1 indicates that the two objects show consistency and similar performance trends, while -1 indicates that the two objects show exactly opposite performance.

[0086] This application transforms raw, coarse latency data into informative and comparable latency metrics by performing calculations on objects with logical relationships at the same level across three different dimensions. It reveals unstable associated objects, quantifies the correlation strength between the performance fluctuations of associated objects and the target object, and provides data support for the final performance correlation analysis report, making performance localization more accurate and efficient.

[0087] In one possible implementation of this application embodiment, when determining the active value corresponding to at least one associated object, the following methods can be used, but are not limited to: sorting multiple target latency data corresponding to the target first associated object in ascending order to obtain sorted latency data, and obtaining the data volume of the multiple target latency data, wherein the target first associated object is any first associated object; performing data calculation based on the data volume, a preset value, and a preset ratio using a first preset algorithm to obtain a first value, and performing data calculation based on the first value, the preset value, and the data volume using a second preset algorithm to obtain a target index value; selecting the target latency data corresponding to the target index value as the active value from the sorted latency data in ascending order.

[0088] In the embodiments of this application, the target latency data refers to multiple latency data corresponding to the first associated object within the target time period. Ascending sorting refers to rearranging a series of unordered values ​​(multiple target latency data) in ascending order, resulting in an ordered sequence, i.e., sorted latency data. Simultaneously, the total number of target latency data points, i.e., the data volume, is acquired and recorded, which can be represented by the symbol... express.

[0089] The first preset algorithm is a predefined set of mathematical operation rules arranged to achieve a specific computational goal. The input parameters of this algorithm include the amount of data. The algorithm uses predefined preset values ​​(e.g., 100) and preset ratios (e.g., 0.8, representing 80%). The algorithm's output is an intermediate result, i.e., the first value.

[0090] The first numerical calculation process can be represented by, but is not limited to, formula (2):

[0091] Formula (2)

[0092] in, That is, the first value.

[0093] After obtaining the first numerical value, the data is then processed further using a second preset algorithm. This second preset algorithm also follows predefined mathematical rules, and its input parameters are the first numerical value, the preset numerical value, and the amount of data. The purpose of the second preset algorithm is to perform necessary adjustments and transformations on the first value, and its output is the target index value. The target index value is an integer that explicitly indicates the target latency data at the corresponding position in the sequence of sorted latency data that will be selected as the active value. The calculation of this active value ensures that the selected target latency data is located near a specific percentile of the sample (e.g., the 80th percentile), thus effectively representing the critical point of a higher latency level.

[0094] Specifically, the calculation of the target index value can be expressed by, but is not limited to, formula (3):

[0095] Formula (3)

[0096] in, The target index value.

[0097] Finally, based on the calculated target index value, the target latency data located at the index position corresponding to the target index value is selected from the sorted latency data. This selected target latency data is defined as the active value corresponding to the first associated object of that target.

[0098] This application determines the active value through a statistical strategy based on sorting and percentile calculation, which can effectively resist the interference of extreme outliers and accurately capture the performance status of the target first associated object. This provides a reliable quantitative standard for subsequent comparison of the resource consumption of multiple first associated objects and screening out the associated objects that have the greatest impact on the performance of the target object.

[0099] In one possible implementation of this application embodiment, when performing performance analysis based on latency metrics, the following methods can be used, but are not limited to: selecting from at least one first associated object based on activity value to obtain at least one target active object corresponding to the target object; selecting from at least one first associated object based on oscillation value to obtain at least one target oscillating object corresponding to the target object; selecting from at least one first associated object based on correlation coefficient value to obtain at least one target related object corresponding to the target object; wherein, the performance analysis result includes at least target active object, target related object, target oscillating object, and second associated object, and the second associated object is an associated object among at least one associated object that has a hierarchical logical relationship with the target object.

[0100] In the embodiments of this application, the selection process refers to a sorting and filtering algorithm based on specific values, which sorts all first associated objects in descending order according to their respective activity values. The higher the activity value, the greater the consumption of shared resources by the first associated object within the target time period, and its performance is at a higher load level. Subsequently, the top-ranked first associated objects (e.g., the Top N, i.e., the N first associated objects with the highest activity values) are selected, and the selected first associated objects are the target active objects corresponding to the target object. The list of target active objects intuitively describes the busiest associated objects among the peer resources of the target object, the ones most likely to compete for resources with it and cause its performance to decline.

[0101] In parallel, the same first associated objects are selected based on their oscillation values. These first associated objects are then sorted in descending order of their respective oscillation values; a higher oscillation value indicates more unstable performance and greater volatility. Similarly, the top-ranked first associated objects are selected, and these are the target oscillation objects corresponding to the target object. The list of target oscillation objects represents the associated objects with the most unstable performance, potential for intermittent failures, or those currently experiencing sudden interference, among the resources of the target object's peer group.

[0102] Furthermore, the first associated objects are selected based on their correlation coefficient values. The correlation coefficient values ​​reflect the strength and direction of the linear correlation between each first associated object and the target object regarding performance fluctuation trends. First, the objects are sorted according to the absolute values ​​of their correlation coefficients to measure the strength of the correlation. Then, first associated objects with strong positive correlation (e.g., values ​​in the range [0.6, 1]) and strong negative correlation (e.g., values ​​in the range [-1, -0.6]) are selected. These selected first associated objects are the target associated objects corresponding to the target object. The list of target associated objects, especially the strongly positively correlated objects, indicates that they may share the same performance bottleneck source with the target object; while the strongly negatively correlated objects reveal the impact of a certain resource allocation dynamic or load balancing strategy.

[0103] Ultimately, the generated performance analysis results are a comprehensive collection, including at least the selected target active objects, target related objects, and target oscillating objects. Furthermore, the performance analysis results also include second-level related objects. A second-level related object is defined as at least one related object that has a hierarchical logical relationship with the target object, such as the storage pool to which the target object (a logical volume) belongs, or the host it is mapped to. Although the second-level related object is not a peer resource of the target object and does not directly participate in peer-based sorting, as a direct parent or child of the target object in the hierarchical structure, its own performance status (e.g., excessive load on the entire storage pool, processor bottlenecks on the host) also has a decisive impact on the performance of the target object. Therefore, including the second-level related objects in the performance analysis results ensures the completeness of the analysis scope, providing operations personnel with a comprehensive view of performance hotspots covering vertical dependencies and horizontal competition.

[0104] This application integrates multi-dimensional objects through a multi-dimensional, algorithmic screening mechanism, making the performance analysis results both targeted (focusing on key influencing objects and avoiding redundancy) and comprehensive (covering peers and hierarchical relationships and avoiding omissions). This provides clear object clues for subsequent cross-validation based on event analysis results (such as status change information of changed objects) and precise location of the root cause of performance problems, significantly improving the efficiency and accuracy of troubleshooting storage system performance issues.

[0105] In one possible implementation of this application embodiment, when selecting from the first associated objects based on the activity value and the oscillation value, the following methods can be used, but are not limited to: sorting at least one first associated object in descending order according to its corresponding activity value to obtain a first sorted associated object; selecting a first preset number of first associated objects from the first sorted associated objects as target active objects; sorting at least one first associated object in descending order according to its corresponding oscillation value to obtain a second sorted associated object; and selecting a second preset number of first associated objects from the second sorted associated objects as target oscillation objects.

[0106] In the embodiments of this application, at least one first associated object (i.e., all objects with a sibling logical relationship to the target object) is sorted in descending order according to their respective activity values, i.e., arranged from largest to smallest activity value, resulting in an ordered sequence, namely the first sorted associated objects. In the first sorted associated objects, the object with the highest activity value is ranked first, and the activity values ​​of subsequent objects decrease sequentially. This sorting result intuitively shows the relative ranking of the degree of shared resource consumption of all first associated objects within a specified target time period; a larger value indicates a heavier business load and a greater potential impact from resource competition.

[0107] Subsequently, from the already sorted sequence of related objects, a first preset number of related objects are selected. This first preset number is an integer parameter pre-configured based on actual operational experience and analysis granularity requirements; for example, it can be set to 3, 5, or 10. It determines how many of the most active objects are ultimately included in the results report. The selected first related objects are formally defined as the target active objects corresponding to the target objects.

[0108] Similarly, a completely parallel filtering logic is applied to the oscillation values, sorting at least one first-order related object in descending order based on its corresponding oscillation value, thus obtaining another ordered sequence, namely the second-order related objects. In the second-order related objects, the object with the largest oscillation value is ranked first, exhibiting the most volatile performance, with the stability of subsequent objects increasing sequentially.

[0109] Next, from the second sequence of related objects, a second preset number of first related objects are selected. This second preset number is also an independently configurable system parameter; its value can be the same as or different from the first preset number, depending on the emphasis placed on activity and stability indicators. The selected second related objects are defined as the target oscillation objects corresponding to the target objects.

[0110] This application achieves an efficient, clear, and configurable object filtering method through two independent descending sorts and a selection mechanism based on a preset number of Top N (i.e., the multiple associated objects with the highest activity / oscillation values). It transforms the abstract performance metrics of activity value and oscillation value into concrete and ordered lists of target active objects and target oscillating objects. This enhances the intuitiveness and operability of the performance analysis results.

[0111] In one possible implementation of this application embodiment, when selecting from the first associated objects based on the correlation coefficient value, the following methods can be used, but are not limited to: determining a preset numerical range in which the correlation coefficient value is located, wherein the preset numerical range includes a preset positive correlation numerical range and a preset negative correlation numerical range; determining the first associated objects corresponding to the correlation coefficient values ​​in the preset positive correlation numerical range as positively correlated objects, and determining the first associated objects corresponding to the correlation coefficient values ​​in the preset negative correlation numerical range as negatively correlated objects; sorting at least one positively correlated object in descending order according to its corresponding correlation coefficient value to obtain a third sorted associated object, and sorting at least one negatively correlated object in ascending order according to its corresponding correlation coefficient value to obtain a fourth sorted associated object; selecting a third preset number of positively correlated objects ranked first from the third sorted associated objects as target positively correlated objects, and selecting a fourth preset number of negatively correlated objects ranked first from the fourth sorted associated objects as target negatively correlated objects, wherein the target associated objects include both positively correlated objects and negatively correlated objects.

[0112] In the embodiments of this application, it is necessary to determine the preset numerical range in which each correlation coefficient value falls. The preset numerical range is a value range defined in advance based on statistical common sense and practical operational experience, used to determine the significance level of the correlation strength. The preset numerical range includes at least two types: preset positive correlation numerical ranges and preset negative correlation numerical ranges. Preset positive correlation numerical ranges are typically set as a closed interval with a lower bound greater than 0.5 and close to 1, for example: [0.6, 1]. The correlation coefficient values ​​within this preset positive correlation numerical range indicate that the performance indicators of the two objects show a significant positive synchronicity, that is, when the latency of one object increases, the latency of the other object also tends to increase. Conversely, preset negative correlation numerical ranges are typically set as a closed interval with an upper bound less than -0.5 and close to -1, for example: [-1, -0.6]. The values ​​within this preset negative correlation numerical range indicate that the performance indicators of the two objects show a significant inverse or complementary trend, that is, when the latency of one object increases, the latency of the other object tends to decrease.

[0113] All objects whose correlation coefficients fall within a preset positive correlation range are identified as positively correlated objects. Positively correlated objects and the target object exhibit performance fluctuations in tandem, indicating they may share the same performance bottleneck or be affected by the same external factor.

[0114] Similarly, the first associated objects corresponding to all correlation coefficient values ​​within the preset negative correlation range are identified as negatively correlated objects. The inverse relationship between the performance of negatively correlated objects and the target object reveals a certain resource allocation strategy (such as load balancing), resource competition, or mutual interference between different workload types within the system.

[0115] Next, positively correlated and negatively correlated objects are sorted within their respective groups to distinguish the relative strength of their correlation. For the set of positively correlated objects, they are sorted in descending order based on their respective correlation coefficient values. Descending order means arranging the values ​​from largest to smallest. After this processing, a third-order sequence of correlated objects is obtained. In this third-order sequence, the first positively correlated object has a correlation coefficient value closest to 1, indicating that it has the strongest positive synchronicity with the target object, and the positive correlation strength of the subsequent objects decreases sequentially.

[0116] For the set of negatively correlated objects, the sorting logic is different. They are sorted in ascending order based on their respective correlation coefficient values. Ascending order means arranging the values ​​from smallest to largest. After this process, a fourth sequence of correlated objects is obtained. Since the correlation coefficient values ​​of all negatively correlated objects are negative, after ascending order, the object at the top has a correlation coefficient value closest to -1 (i.e., the largest absolute value of negative correlation), indicating that it has the strongest inverse correlation with the target object. The negative correlation strength (absolute value) of subsequent objects decreases sequentially. This sorting strategy ensures that within each group, the objects at the top of the sequence are the most correlated objects in their respective categories.

[0117] Finally, the most representative objects are selected from the sorted sequences. From the third sorted related objects (i.e., the sorted group of positively correlated objects), the top three preset number of positively correlated objects are selected as target positively correlated objects. Similarly, from the fourth sorted related objects (i.e., the sorted group of negatively correlated objects), the top four preset number of negatively correlated objects are selected as target negatively correlated objects. The third and fourth preset numbers are integer parameters that can be preset by the system to control the upper limit of the number of each type of object in the final output. Ultimately, the target related objects are a comprehensive set that includes both the selected target positively correlated objects and the target negatively correlated objects.

[0118] This application identifies objects that are strongly correlated with the target object by identifying relevant objects, and further reveals the direction of the correlation, providing operations and maintenance personnel with in-depth and highly valuable insights into the causes of performance problems.

[0119] It should be noted that the process of determining the target active object, the target oscillation object, and the target related object is an independent process. That is, when selecting the target active object, the target oscillation object, and the target related object, they are all selected from the initial set of first related objects. The selected target active object, target oscillation object, and target related object are also independent of each other. In other words, a first related object can be any one or more of the target active object, the target oscillation object, and the target related object.

[0120] Furthermore, to facilitate the specific operational process of performance dimension analysis, this application embodiment also provides a flowchart of performance dimension analysis, such as... Figure 4 As shown, it includes:

[0121] Step 1: When starting performance correlation, specify the object and time to be analyzed (i.e., the target object and the target time period); Step 2: Retrieve the resource correlation model according to the target object, and take the union of the obtained result fields to obtain multiple correlated objects; Step 3: For each correlated object, retrieve latency data according to the target time period; Step 4: For each object group (i.e., correlated objects with the same logical relationship as the target object), calculate its activity value / correlation coefficient value / oscillation value. The activity value and oscillation value only need to be calculated for each object in the object group, while the correlation coefficient needs to compare the correlation between each object in the object group and the target object. Step 5: For the activity value / correlation coefficient value / oscillation value, select correlated objects. The most active and most volatile objects can be directly selected based on the activity value / oscillation value, ranking them first. For correlation, two parts need to be selected: Top N positively correlated objects (i.e., the multiple positively correlated objects with the largest correlation coefficient values) and Top N negatively correlated objects (i.e., the multiple negatively correlated objects with the largest absolute values ​​of correlation coefficient values).

[0122] In one possible implementation of this application, after constructing a resource relationship model covering the storage system and network links, and realizing automated correlation analysis of events and performance in two dimensions based on the model, the system can also perform deep fusion and intelligent reasoning of multidimensional data based on the intelligent correlation and root cause inference of the rule engine, and finally generate a deep analysis report with clear causal inference and optimization suggestions, thereby realizing the leap from data presentation to intelligent diagnosis.

[0123] Specifically, but not limited to the following methods can also be used: taking the changed objects and their status changes in the event analysis results, as well as the target active objects, target oscillating objects, target related objects, and second associated objects in the performance analysis results as input facts; matching the input facts with a pre-set rule knowledge base, which stores multiple scenario pattern rules, each scenario pattern rule defining the conditional logic of a performance failure scenario, as well as the corresponding inference conclusions and handling suggestions after matching; generating and outputting a performance correlation analysis report containing the inference conclusions and handling suggestions corresponding to the successfully matched scenario pattern rules.

[0124] Specifically, when generating the final performance correlation analysis report, a rule engine (including a rule knowledge base) with predefined root cause scenario patterns is invoked. The input to this rule engine is the event analysis results (i.e., all retrieved changed objects and their state changes) and the performance analysis results (i.e., the filtered target active objects, target oscillating objects, target positive / negative correlated objects, etc.).

[0125] The rule engine embeds pre-defined inference logic. Each rule describes a specific failure or bottleneck pattern. For example:

[0126] A rule might be defined as:

[0127] IF (Latency spikes for the target object)

[0128] AND (one of the target positively correlated objects is a volume in the same pool)

[0129] AND (The target's positively correlated objects have change information related to snapshot creation operations at similar time points)

[0130] THEN Root Cause Confidence: High;

[0131] Conclusion: Bursting background tasks (snapshots) within the same storage pool lead to resource contention;

[0132] Recommendation: Adjust the snapshot strategy or business scheduling time.

[0133] The rule engine matches the current event analysis and performance analysis results with scenarios in the rule knowledge base. The matching process considers not only the existence of events and performance anomalies, but also their temporal sequence relationships, intensity, and topological associations between objects (e.g., same pool, same zone). It calculates the confidence score for each matching rule and ultimately outputs one or more most probable root cause inferences and their corresponding optimization suggestions.

[0134] By formalizing knowledge into computable rules, intelligent fusion and interpretation of multi-source, heterogeneous performance diagnostic data are achieved. The resulting benefits include: significantly improved accuracy and reliability of analysis results; freeing operations and maintenance personnel from tedious data correlation analysis; and directly providing actionable directions for fault handling and performance optimization, thus achieving a high-level leap from passive monitoring to intelligent operations and maintenance.

[0135] In one possible implementation of this application, the resource relationship model is the foundation of event analysis and performance analysis, and its accuracy and timeliness directly determine the reliability of the analysis results. However, the business links of the storage system are not static. As business needs are adjusted, resource configurations are changed, or hardware status changes, the relationships between various resource objects will continuously change dynamically. For example, adding volumes to the storage pool, adjusting switch zone partition configurations, and modifying the mapping relationship between volumes and hosts. If the resource relationship model is not updated for a long time after its construction, the association logic in the model will become disconnected from the actual business links, which will lead to deviations in the subsequent analysis results based on the model (e.g., missing newly added volumes in the same pool, or misjudging deleted switch port associations). Therefore, after constructing a resource relationship model that at least includes logical relationships, an effective model update mechanism needs to be established.

[0136] Specifically, regarding the model update mechanism, the following methods may also be used, but are not limited to: obtaining the association relationship of each resource object in the business link based on a preset period, and updating the resource relationship model based on the association relationship; or, in response to a state change of any resource object among the resource objects, obtaining the changed association relationship of each resource object in the business link, and updating the resource relationship model based on the changed association relationship.

[0137] In the embodiments of this application, at least two parallel, optional model update strategies are included, specifically including but not limited to: The first strategy is a time-cycle-based passive polling update, which actively and periodically initiates information collection tasks based on a preset cycle. The preset cycle is a time interval parameter pre-configured according to the frequency of environmental changes, for example, it can be set to execute once every 24 hours or every 12 hours. When this cycle is triggered, the current association relationship of each resource object in the business link of the system to be analyzed will be obtained again. This process is similar to that when initially building the model, and the current status and connection relationship of the first resource on the storage side and the second resource on the storage network link will be rediscovered and collected. After obtaining the latest relationship data, the existing resource relationship model is updated based on the latest association relationship. The update process refers to incrementally or fully replacing the outdated information in the model with the newly obtained data, so that the model keeps synchronized with the actual state of the system. The periodic update method is suitable for environments with infrequent changes and can continuously correct model drift that may be caused by previously uncaptured configuration fine-tuning with controllable system overhead.

[0138] The second strategy is event-driven proactive updates, a real-time and efficient update mechanism. It continuously monitors the status of each resource object, and immediately triggers an update process once a status change is detected for any resource object. Status changes include, but are not limited to: resource creation and deletion (e.g., creating a new logical volume), modification of resource configuration (e.g., adjusting LUN mappings, changing Zone members), and changes in resource running status (e.g., port error disabling, hard disk failure). When such events occur, the system accurately obtains the changed relationships between each resource object in the business chain. For example, if a switch port is moved out of a Zone, the system obtains the latest port member list for that Zone. Subsequently, the resource relationship model is immediately updated based on these changed relationships. This event-driven approach enables near real-time model updates, ensuring that the model is almost always up-to-date at the moment of analysis, thus greatly improving the timeliness and accuracy of the analysis results.

[0139] The model maintenance mechanism defined in this application constructs an adaptive and self-updating resource relationship model through two complementary methods: pre-set periodic polling and event triggering in response to state changes. This effectively addresses the challenges posed by the dynamic nature of storage systems, ensuring that the model, which forms the basis of the entire analysis method, always possesses high real-time accuracy and reliability, thus providing a data foundation for performance correlation analysis.

[0140] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method.

[0141] Embodiments of this application also provide an analysis apparatus for a storage system. Figure 5 A schematic diagram of the structure of an analysis device for a storage system provided in this application is shown below. Figure 5 As shown, it includes:

[0142] The first analysis unit 51 is used to respond to the analysis command for the target object, and perform information analysis and processing on the log information of the target object and at least one associated object based on the resource relationship model to obtain the event analysis result. The resource relationship model is a model constructed based on the relationship between each resource object in the business link in the system to be analyzed, and includes at least the logical relationship between each resource object. The associated object is a resource object that has a logical relationship with the target object as determined by the resource relationship model.

[0143] The calculation unit 52 is used to perform performance index calculation based on the resource relationship model and the latency data corresponding to at least one associated object, so as to obtain the latency index corresponding to at least one associated object.

[0144] The second analysis unit 53 is used to perform performance analysis on at least one associated object based on the latency index and obtain the performance analysis results.

[0145] The generation unit 54 is used to generate a performance correlation analysis report based on the event analysis results and performance analysis results.

[0146] In one embodiment of this application, such as Figure 6 As shown, the analysis apparatus for the storage system further includes a construction unit 55, wherein the construction unit 55 is used for:

[0147] Obtain the first resource on the storage side of the system to be analyzed, and the second resource on the storage network link. The first resource and the second resource include at least each resource object and its relationship.

[0148] Based on the object identifier, first resource, and second resource corresponding to each resource object, construct a resource relationship model that includes at least logical relationships, where the logical relationships include at least the hierarchical logical relationships and the sibling logical relationships between each resource object.

[0149] In one embodiment of this application, the first analysis unit 51 is specifically used for:

[0150] Responding to analysis commands instructing the analysis of the target object's performance within a target time period;

[0151] Extract log information corresponding to the target object and at least one associated object within the target time period from the preset database. The target time period is no longer than the analysis time threshold corresponding to the target object set in the preset database. The preset database includes object information corresponding to each resource object, and the object information includes log information.

[0152] Extract change information from the log information, which includes at least the status changes of each resource object in the system to be analyzed;

[0153] Based on the change information, a retrieval process is performed in the resource relationship model to identify the target object and at least one related object that has undergone a change, thereby obtaining the event analysis results. The changed object is a resource object whose state has changed.

[0154] In one embodiment of this application, the computing unit 52 is specifically used for:

[0155] Obtain multiple time delay data corresponding to at least one first associated object within the target time period, wherein the first associated object is an associated object that has a logical relationship with the target object at the same level among at least one associated object;

[0156] Perform data selection processing on multiple time-delay data to determine the activity value corresponding to at least one first associated object;

[0157] The standard deviation of multiple time-delay data is calculated to obtain the oscillation value corresponding to at least one first associated object;

[0158] The target object is subjected to correlation coefficient analysis with at least one first associated object to obtain the correlation coefficient value between the target object and at least one associated object. The time delay index includes the activity value, the oscillation value and the correlation coefficient value.

[0159] In one embodiment of this application, the computing unit 52 is further configured to:

[0160] Sort the multiple target delay data corresponding to the first associated object of the target in ascending order to obtain sorted delay data, and obtain the data volume of multiple target delay data, wherein the first associated object of the target is any first associated object;

[0161] The first preset algorithm calculates the data based on the data volume, preset value, and preset ratio to obtain the first value, and the second preset algorithm calculates the data based on the first value, preset value, and data volume to obtain the target index value.

[0162] From the sorted latency data, select the target latency data corresponding to the target index value as the active value in ascending order.

[0163] In one embodiment of this application, the second analysis unit 53 is specifically used for:

[0164] Based on the activity value, at least one target active object is selected from at least one first associated object to obtain at least one target active object corresponding to the target object; and based on the oscillation value, at least one target oscillating object is selected from at least one first associated object to obtain at least one target oscillating object corresponding to the target object.

[0165] Based on the correlation coefficient value, at least one target related object is selected from at least one first associated object to obtain at least one target related object corresponding to the target object;

[0166] The performance analysis results include at least the target active object, the target related object, the target oscillating object, and the second associated object. The second associated object is at least one associated object that has a hierarchical logical relationship with the target object.

[0167] In one embodiment of this application, the second analysis unit 53 is further configured to:

[0168] At least one first associated object is sorted in descending order according to its corresponding activity value to obtain the first sorted associated object;

[0169] From the first sorted associated objects, select the first preset number of first associated objects that are ranked first as the target active objects;

[0170] At least one first associated object is sorted in descending order according to its corresponding oscillation value to obtain a second sorted associated object;

[0171] From the second sorted associated objects, select the first associated object with the second preset number of sorted items as the target oscillation object.

[0172] In one embodiment of this application, the second analysis unit 53 is further configured to:

[0173] Determine the preset numerical range in which the correlation coefficient value lies, wherein the preset numerical range includes a preset positive correlation numerical range and a preset negative correlation numerical range;

[0174] The first associated object corresponding to the correlation coefficient value in the preset positive correlation range is determined as a positively correlated object, and the first associated object corresponding to the correlation coefficient value in the preset negative correlation range is determined as a negatively correlated object.

[0175] At least one positively correlated object is sorted in descending order according to its corresponding correlation coefficient value to obtain the third sorted associated object; at least one negatively correlated object is sorted in ascending order according to its corresponding correlation coefficient value to obtain the fourth sorted associated object.

[0176] From the third sorted related objects, select the third preset number of positively related objects that are ranked first as target positively related objects. From the fourth sorted related objects, select the fourth preset number of negatively related objects that are ranked first as target negatively related objects. The target related objects include target positively related objects and target negatively related objects.

[0177] In one embodiment of this application, such as Figure 6 As shown, the analysis apparatus for the storage system further includes an update unit 56, wherein the update unit 56 is used for:

[0178] The relationships between each resource object in the business chain are obtained based on a preset period, and the resource relationship model is updated based on the relationships.

[0179] Alternatively, in response to a state change of any resource object among the resource objects, the changed association relationship of each resource object in the business chain is obtained, and the resource relationship model is updated based on the changed association relationship.

[0180] For a description of the features in the embodiment corresponding to the analysis device of the storage system, please refer to the relevant description of the embodiment corresponding to the analysis method of the storage system, which will not be repeated here.

[0181] Embodiments of this application also provide an electronic device, including a memory and a processor, wherein the memory stores a computer program and the processor is configured to run the computer program to perform the steps in any of the above-described storage system analysis method embodiments.

[0182] Embodiments of this application also provide a computer-readable storage medium storing a computer program, wherein the computer program is configured to execute the steps in any of the above-described storage system analysis method embodiments when run.

[0183] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard disk, magnetic disk, or optical disk.

[0184] Embodiments of this application also provide a computer program product, which includes a computer program that, when executed by a processor, implements the steps in any of the above-described storage system analysis method embodiments.

[0185] Embodiments of this application also provide another computer program product, including a non-volatile computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps in the analysis method embodiments of any of the above storage systems.

[0186] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented 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 this application.

[0187] The analysis method and electronic device for a storage system provided in this application have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are only for the purpose of helping to understand the method and core ideas of this application. It should be noted that those skilled in the art can make several improvements and modifications to this application without departing from the principles of this application, and these improvements and modifications also fall within the protection scope of the claims of this application.

Claims

1. A method for analyzing a storage system, characterized in that, include: In response to the analysis command for the target object, based on the resource relationship model, information analysis and processing are performed on the log information of the target object and at least one associated object to obtain the event analysis result. The resource relationship model is a model constructed based on the association relationship of each resource object in the business link in the system to be analyzed, and includes at least the logical association relationship between each resource object. The associated object is a resource object that has the logical association relationship with the target object based on the resource relationship model. The logical association relationship includes at least the hierarchical logical association relationship and the sibling logical association relationship between each resource object. Based on the resource relationship model, performance indicators are calculated and processed according to the latency data corresponding to at least one of the associated objects to obtain the latency indicators corresponding to at least one of the associated objects. Based on the latency index, at least one of the associated objects is subjected to performance analysis processing to obtain performance analysis results; Based on the event analysis results and the performance analysis results, a performance correlation analysis report is generated; The process of analyzing and processing the log information of the target object and at least one associated object based on the resource relationship model to obtain event analysis results includes: Extract change information from the log information, wherein the change information includes at least the status change status of each resource object in the system to be analyzed; Based on the change information, a retrieval process is performed in the resource relationship model to determine the target object and at least one changed object among the associated objects, thereby obtaining the event analysis result. The changed object is a resource object whose status has changed. The step of calculating and processing performance indicators based on the resource relationship model and the latency data corresponding to at least one of the associated objects to obtain the latency indicators corresponding to each of the at least one associated object includes: Obtain multiple delay data corresponding to at least one first associated object within a target time period, wherein the first associated object is an associated object among at least one of the associated objects that has a logical association with the target object at the same level; Data selection processing is performed on multiple latency data to determine the activity value corresponding to at least one of the first associated objects; The standard deviation of the multiple time delay data is calculated to obtain the oscillation value corresponding to at least one of the first associated objects; The target object is subjected to correlation coefficient analysis with at least one of the first associated objects to obtain the correlation coefficient values ​​between the target object and at least one of the associated objects, wherein the time delay index includes the activity value, the oscillation value and the correlation coefficient value.

2. The analysis method for the storage system according to claim 1, characterized in that, The method further includes: Obtain the first resource on the storage side of the system to be analyzed, and the second resource on the storage network link, wherein the first resource and the second resource include at least the resource objects and the association relationship; Based on the object identifiers corresponding to each resource object, the first resource, and the second resource, a resource relationship model is constructed that includes at least the logical association relationships. In response to the analysis command instructing the analysis of the performance of the target object within the target time period; Log information corresponding to the target object and at least one associated object within the target time period is extracted from a preset database, wherein the target time period is not greater than the analysis time threshold corresponding to the target object set in the preset database, and the preset database includes object information corresponding to each resource object, and the object information includes the log information.

3. The analysis method for a storage system according to claim 1, characterized in that, The step of performing data selection processing on multiple latency data to determine the activity value corresponding to at least one of the associated objects includes: The target delay data corresponding to the first associated object of the target are sorted in ascending order to obtain sorted delay data, and the data volume of the multiple target delay data is obtained, wherein the first associated object of the target is any one of the first associated objects; The first preset algorithm performs data calculation based on the data volume, preset value, and preset ratio to obtain a first value, and the second preset algorithm performs data calculation processing based on the first value, the preset value, and the data volume to obtain a target index value. From the sorted latency data, the target latency data corresponding to the target index value is selected as the active value in ascending order.

4. The method for analyzing a storage system according to claim 1, characterized in that, The step of performing performance analysis on at least one of the associated objects based on the latency index to obtain performance analysis results includes: Based on the activity value, at least one target active object is selected from at least one first associated object to obtain at least one target active object corresponding to the target object; and based on the oscillation value, at least one target oscillating object is selected from at least one first associated object to obtain at least one target oscillating object corresponding to the target object. Based on the correlation coefficient value, at least one target related object is selected from at least one first associated object to obtain at least one target related object corresponding to the target object; The performance analysis results include at least the target active object, the target related object, the target oscillating object, and the second associated object. The second associated object is at least one of the associated objects that has a hierarchical logical relationship with the target object.

5. The method for analyzing a storage system according to claim 4, characterized in that, The step of selecting from at least one first associated object based on the activity value to obtain at least one target active object corresponding to the target object, and selecting from at least one first associated object based on the oscillation value to obtain at least one target oscillating object corresponding to the target object, includes: At least one of the first associated objects is sorted in descending order according to their respective activity values ​​to obtain the first sorted associated objects; From the first sorted associated objects, select the first preset number of the first associated objects that are sorted first as the target active objects; At least one of the first associated objects is sorted in descending order according to its corresponding oscillation value to obtain the second sorted associated objects; From the second sorted associated objects, select the first associated objects that are ranked first by a second preset number as the target oscillation objects.

6. The method for analyzing a storage system according to claim 5, characterized in that, The step of selecting from at least one first associated object based on the correlation coefficient value to obtain at least one target associated object corresponding to the target object includes: Determine the preset numerical range in which the correlation coefficient value lies, wherein the preset numerical range includes a preset positive correlation numerical range and a preset negative correlation numerical range; The first associated object corresponding to the correlation coefficient value that is in the preset positive correlation value range is determined as a positively correlated object, and the first associated object corresponding to the correlation coefficient value that is in the preset negative correlation value range is determined as a negatively correlated object; At least one of the positively correlated objects is sorted in descending order according to its corresponding correlation coefficient value to obtain a third sorted associated object; at least one of the negatively correlated objects is sorted in ascending order according to its corresponding correlation coefficient value to obtain a fourth sorted associated object. From the third sorted associated objects, a third preset number of positively correlated objects are selected as target positively correlated objects, and from the fourth sorted associated objects, a fourth preset number of negatively correlated objects are selected as target negatively correlated objects, wherein the target correlated objects include the target positively correlated objects and the target negatively correlated objects.

7. The method for analyzing a storage system according to claim 2, characterized in that, After constructing the resource relationship model that includes at least the logical associations, the method further includes: The association relationships of each resource object in the business chain are obtained based on a preset period, and the resource relationship model is updated based on the association relationships. Alternatively, in response to a state change in any of the resource objects, the changed associations of the resource objects in the business chain are obtained, and the resource relationship model is updated based on the changed associations.

8. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor, configured to implement the steps of the analysis method for the storage system as described in any one of claims 1 to 7 when executing the computer program.

Citation Information

Patent Citations

  • Intelligent operation and maintenance method and device based on knowledge graph and large model and electronic equipment

    CN120596306A