Intelligent management system and method for abnormal workload of tape library file system

Abstract

The invention discloses an intelligent management system for abnormal workloads of a tape library file system, which comprises an abnormal access training unit for simulating and running specific workloads by using a special test environment and carrying out classified training on the workloads, and the abnormal access training unit comprises a client server and an AI training server, the client server comprises a Windows thumbnail / index service module, a MacOS index / Quick Look service module, a malicious scanning program module, a normal program, a storage unvarnished transmission module, an application information collection module and a disk file module, and the AI training server comprises an AI learning engine module and a model file module. According to the method, the unfriendly abnormal workload for the tape file system can be automatically identified and detected and can be automatically processed, and the problems that the performance of a tape library becomes poor and even response cannot be performed due to the abnormal workload can be effectively avoided under the condition that user participation is avoided as much as possible.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of magnetic tape library load exception processing, and particularly to an intelligent management method for abnormal work load of a magnetic tape library file system. BACKGROUND

[0002] A magnetic tape library is an important storage system for mass data storage. After the introduction of a magnetic tape file system (LTFS), users can directly write and read files to and from the magnetic tape in the magnetic tape library, making the use of the magnetic tape library popular in various fields.

[0003] However, the magnetic tape library has its own storage characteristics. For example, after loading a single disk tape, the sequential read-write speed of the tape will be fast, but the operation of loading the tape itself will be slow (a mechanical hand is needed to perform the tape grabbing and releasing operation), and the operation of randomly addressing data in the same disk tape will also be slow (repeated rewinding is needed). If the applied work load causes the magnetic tape library to repeatedly load a large number of tapes to read and write data, or to randomly access data in a disk tape, the performance of the magnetic tape library will be poor, and it may even be unable to respond.

[0004] After the introduction of the magnetic tape file system, many users will directly use the magnetic tape library in the same way as ordinary file storage. However, some common file operations will become abnormal work loads of the magnetic tape library. For example, the Windows and MAC operating systems will scan all the files of the mounted file system to establish an index, which will cause the magnetic tape library to load a large number of tapes in a short period of time, resulting in abnormal performance. In addition, the full-disk scanning process of anti-virus software will also cause similar situations. SUMMARY

[0005] To overcome the above-mentioned shortcomings, the present application aims to provide an intelligent management system and method for abnormal work load of a magnetic tape library file system, which can automatically identify and detect abnormal work load that is not friendly to the magnetic tape file system, and automatically process it. The system and method can effectively avoid the problem of poor performance or even unresponsiveness of the magnetic tape library caused by abnormal work load, and reduce the failure of the magnetic tape library, without the user's participation.

[0006] To achieve the above-mentioned purposes, the technical solution adopted by the present application is as follows: an intelligent management system for abnormal work load of a magnetic tape library file system, comprising an abnormal access training unit for simulating the running of specific work load using a special test environment and performing work load classification training, the abnormal access training unit comprising:

[0007] A client server, comprising: a Windows thumbnail / index service module, a Mac OS index / Quick Look service module, a malicious scanner module simulating a malicious scanning application, a normal program for accessing a file by a normal application, a storage pass-through module for intercepting information related to accessing a file, an application information collection module for collecting additional information of an application process, and a disk file module having an original file, the disk file module being exposed to a user for use through the storage pass-through module;

[0008] An AI training server, comprising: an AI learning engine module for receiving information related to accessing a file intercepted by the storage pass-through module and additional information of an application process collected by the application information collection module, and performing learning and training based on the collected information; and a model file module for saving a model file.

[0009] Preferably, the information related to accessing a file includes at least one of file type, file name, LBA, size, timestamp, process id, uid, and gid information intercepted by a fuse file system pass-through technology.

[0010] Preferably, the additional information refers to additional information of an application process collected based on given process id, uid, and gid information, including a name of the process and a name of an operating system service corresponding to the process.

[0011] Preferably, the AI learning engine module performs machine learning and training based on the collected information related to accessing a file and the additional information of an application process, and the learning algorithm includes a gradient boosting tree algorithm, a random forest algorithm, and a deep learning model algorithm using a one-dimensional convolutional neural network (1D-CNN) and LSTM joint modeling algorithm.

[0012] Preferably, the system further comprises a main work unit, and the main work unit comprises a storage server and a tape library, and the storage server comprises a tape library management module, and the tape library management module comprises:

[0013] The tape library management module comprises storage hierarchical management, abnormal access pattern recognition, and migration and back migration scheduling management, the storage hierarchical management is used to automatically migrate cold data to a tape according to a certain strategy, the abnormal access pattern recognition is used to identify and judge a pattern based on read LBA information of one or more files and a pre-trained and saved model, and the migration and back migration scheduling management is used to schedule a migration and back migration request, and schedule migration or back migration based on free resources of a tape library device and a result of the abnormal access pattern recognition.

[0014] Preferably, the abnormal access mode recognition can be associated with the information of the application information audit module for overall judgment.

[0015] Preferably, the storage server further comprises:

[0016] An LTFS module, as a tape file system management module, is used to format the tape and mount it as an LTFS format file system;

[0017] A tape device driver module, as a SCSI driver of the tape library device, can discover the device of the tape library, including the drive, the tape, and obtain the information of the device;

[0018] A disk storage module, as a data cache of the tape library management module;

[0019] A database is used to save the management information of the migration back to the migration scheduling management; the management information includes information for the migration operation, each operation including the file name, the read LBA, the read size, and the like, and can also contain some additional information, such as the process id, the uid, the gid, and the like.

[0020] Preferably, the main working unit further comprises a main client server, and the main client server comprises:

[0021] An NFS / SMB client;

[0022] A main application information audit module, which is used to record the process id, the service name, the file name, the offset, the size, the timestamp information of the process accessing the NFS or SMB mounted file system;

[0023] And a Windows / Linux / MacOS operating system.

[0024] The application also discloses an intelligent management method for abnormal working load of a tape library file system, which adopts the intelligent management system and comprises the following steps:

[0025] S1, a file access information acquisition step is performed through a storage access layer; through a fuse technology, the information related to all access requests for accessing the tape file (including but not limited to: file name, LBA, size, timestamp, process id, service name, etc.) is recorded on the storage layer; an AI engine is used to classify and identify the working load of the above-mentioned tape file access request; the file access requests identified as the same working load are marked with the working load source; the AI engine simulates the running of a certain specific working load through an abnormal access training unit, then captures the read-write characteristics to train the working load classification, and simultaneously generates the coping strategies in advance according to whether the working load is friendly to the tape library and adds the strategies to the strategy library.

[0026] S2, an abnormal workload identification step, including identification of unknown workload sources and identification of known workload sources;

[0027] S3, an abnormal workload isolation step, putting the file access request R marked as abnormal into a separate queue Q, and marking the queue Q;

[0028] S4, an abnormal workload processing step, for each Q, performing policy matching, the method of policy execution including reducing execution priority, limiting execution speed, suspending execution, canceling execution, queue optimization; if no policy is matched, the default is to take the suspend execution policy, and an alarm is sent to the user and administrator through the alarm system, the user and administrator confirm the abnormal workload, and give the execution policy confirmation; if multiple policies are matched, the highest priority action is taken.

[0029] Preferably, in S3, if the workload does not match the policy and the user confirms it, the label of the workload is automatically added to the policy library with the user's execution behavior to generate the policy.

[0030] Preferably, the abnormal workload identification method for unknown workload sources includes the following steps:

[0031] Step one, when the user accesses the data of the tape file system, the following information can be extracted from each file access request (R) received from the tape file system: tape number, head addressing distance (L) of the tape, data access volume (S), and recording the time T of loading a tape in the tape library m , time T m contains the average time of taking out the existing tape in the drive and loading a new tape in the tape library, the average speed V of the tape library addressing in a tape s , the speed V of the tape sequential read-write rw , T m , V s , V rw can be obtained from the manufacturer according to the model and configuration of the tape library; for example, LTO tape uses linear serpentine recording technology, stores data through multiple parallel data tracks, and each tape contains 4-16 parallel data tracks, so the storage location of a file on the tape is represented by offset (logical offset), which can be converted into physical distance (meters) and access time, for example, 12345678 can be converted into the position of the xth meter of the tape, and under the condition of the average rotating speed of the tape access, the access time can be calculated. The offset of each file on the tape can be converted into the actual distance and expected access time of the tape accessing this address;

[0032] Step two, define the file access requests R from the same workload source in a period of time t as a set R t , the number of which is N;

[0033] Step three, calculate the ideal access time T according to the total amount of data accessed in Rx rw , the speed of tape sequential access V o , and the expected access time T E = T O × α, where α is a performance coefficient that can be pre-set according to the tolerance of performance; where Si is the amount of data accessed per tape library.

[0034]

[0035] Step four, calculate the simulated access time Ts for all access requests R in Rt, where M is the number of different tapes required to be accessed in Rt, and Li is the head addressing distance per request;

[0036]

[0037] Step five, if T s > T E , it is considered that an abnormal workload has occurred.

[0038] Preferably, the abnormal workload identification method for known workload sources is:

[0039] When the workload source matches the index, thumbnail process of Windows / MAC, it is marked as an abnormal workload;

[0040] When the workload source matches the system scan of an antivirus software, it is marked as an abnormal workload. It can also be automatically marked according to the rules of other known workload sources.

[0041] Preferably, for known workload sources, if the workload source can match the corresponding strategy, the corresponding strategy is executed; if it matches the system scan process of Windows / MAC or the scan process of an antivirus software, the request execution is cancelled.

[0042] Preferably, for unknown workload sources, when the Ts / TE calculated by the abnormal workload is more than 5 times and less than 20 times, the priority of the request in Q is continuously reduced according to the multiple ratio; when the Ts / TE calculated by the abnormal workload is more than 20 times, the request execution is suspended; when the Ts / TE calculated by the abnormal workload is less than 5 times, the order optimization of the request tasks in the queue Q is started, such as sorting from small to large according to the L of each R to optimize the addressing of the tapes.

[0043] The intelligent management system and method for abnormal work load of the tape library file system has the following advantages: the intelligent method is used to identify the abnormal work load characteristics which are not friendly to the tape file system, the related abnormal work load is put into a separate task queue, and the data operation task is isolated; different actions such as reducing the execution priority, limiting the execution speed, suspending the execution, canceling the execution and the like are automatically taken according to the strategy for the isolated abnormal work load task queue, so as to avoid the problem that the performance of the tape library is reduced or even cannot be responded due to the execution of the abnormal work load; the related strategy (such as canceling the execution) is automatically adapted through the identification of the work load source (such as the system access of window and MAC); the user and the administrator are notified by the system alarm to confirm the abnormal work load, and the actions (such as reducing the execution priority, limiting the execution speed, suspending the execution, canceling the execution) confirmed by the user and the administrator are executed; the results of the abnormal work load alarm and the user / administrator confirmation are recorded, the abnormal work load characteristics and the corresponding actions are associated, and the results are used as the strategy for the subsequent automatic execution of the system; the abnormal work load which is not friendly to the tape file system can be automatically identified and detected, and automatically processed, so that the problem that the performance of the tape library is reduced or even cannot be responded due to the abnormal work load can be effectively avoided without the participation of the user, and the failure of the tape library is reduced. Then, the tape file system can be smoothly used by the user without the professional knowledge of the running characteristics of the tape library, the use threshold of the tape file system is reduced, and the use scene and range are expanded; the processing mode of the abnormal work load by the high-level user can be automatically learned, the abnormal work load identification means and the processing strategy library can be automatically enriched, the automatic continuous improvement and optimization are realized, and the ability of the abnormal work load detection and processing is continuously improved. BRIEF DESCRIPTION OF DRAWINGS

[0044] Figure 1 The main system module block diagram of the main work unit is shown in FIG. 1;

[0045] Figure 2 The module block diagram of the abnormal access training unit is shown in FIG. 2;

[0046] Figure 3 The main flow of processing the abnormal work load of the tape library is shown in FIG. 3. DETAILED DESCRIPTION

[0047] The preferred embodiments of the present application are described in detail below with reference to the accompanying drawings, so that the advantages and characteristics of the present application can be more easily understood by those skilled in the art, and the protection scope of the present application is more clearly and definitely defined.

[0048] Reference Figure 2As shown, this embodiment discloses an intelligent management system for abnormal workloads of a tape library file system, including an abnormal access training unit that uses a dedicated test environment to simulate running a specific workload and perform workload classification training. The abnormal access training unit includes:

[0049] The client-server includes: a Windows thumbnail / index service module, a MacOS index / QuickLook service module, a malicious scanner module that simulates malicious scanning applications, a normal program for normal applications to access files, a storage transmission module that intercepts information related to accessed files (the relevant information includes but is not limited to intercepting the file type, file name, LBA, size, timestamp, process ID, UID, and GID information of the accessed files using fuse file system transmission technology), an application information collection module for collecting additional information about application processes, and a disk file module with original files. The disk file module is exposed to users through the storage transmission module;

[0050] AI training server, the AI ​​training server includes: an AI learning engine module and a model file module. The AI ​​learning engine module is used to receive and store relevant information intercepted by the transparent transmission module and additional information of the application process collected by the application information collection module, and perform learning and training based on the collected information; the model file module is used to save the model file.

[0051] Additional information refers to: given the process id, uid, and gid information, to collect additional information about the application process, including the name of the process and the name of the operating system service corresponding to the process.

[0052] The AI ​​learning engine module performs machine learning and training based on the collected relevant and additional information. The learning algorithms include the gradient boosting tree algorithm, the random forest algorithm, and the use of deep learning model algorithms by capturing long-term temporal dependencies. The deep learning model algorithm adopts a one-dimensional convolutional neural network (1D-CNN) and LSTM joint modeling algorithm.

[0053] like Figure 1 As shown, the intelligent management system also includes main working units, including storage servers and tape libraries. The storage server includes a tape library management module, and the tape library management module includes:

[0054] The tape library management module includes storage tier management, abnormal access pattern recognition, and migration and re-migration scheduling management. The storage tier management is used to automatically migrate cold data to tapes according to certain strategies. The abnormal access pattern recognition is based on the read LBA information of one or more files, combined with pre-trained and saved models, to perform pattern recognition and judgment. The migration and re-migration scheduling management is used to schedule and manage migration and re-migration requests, and to schedule migration or re-migration based on the idle resources of the tape library equipment and the results of abnormal access pattern recognition.

[0055] Abnormal access pattern identification can be associated with the information of the application information audit module to make an overall judgment.

[0056] The storage server also includes:

[0057] The LTFS module, as a tape file system management module, is used to format tapes and mount them as an LTFS file system;

[0058] The tape device driver module, as the SCSI driver for tape library devices, can discover tape library devices, including drives and tapes, and obtain device information;

[0059] The disk storage module is used as a data cache for the tape library management module;

[0060] The database is used to store management information for migration and re-migration scheduling management; the management information includes information for re-migration operations, each operation includes information such as file name, read LBA, read size, etc., and can also include some additional information such as process ID, UID, GID, etc.

[0061] The main working unit also includes the main client server, which includes:

[0062] NFS / SMB client;

[0063] The main application information audit module is used to record the process ID, service name, file name, offset, size, and timestamp information of accessing the file system mounted by NFS or SMB;

[0064] And Windows / Linux / MacOS operating systems.

[0065] like Figure 3 As shown, an intelligent management method for abnormal workload of a tape library file system adopts the above-mentioned intelligent management system and includes the following steps:

[0066] S1, file access information acquisition step through the storage access layer, through the fuse technology, record the information related to all access requests to the tape file in the storage layer (including but not limited to: file name, LBA, size, timestamp, process id, service name, etc.); Through the AI engine, the above tape file access request is classified and identified as a working load, and the file access request identified as the same working load is marked as the source of the working load. The AI engine simulates the running of a certain specific working load through the abnormal access training unit, then captures the characteristics of its read and write to train the working load classification, and generates the coping strategy in advance according to whether the working load is friendly to the tape library and adds it to the strategy library;

[0067] S2, abnormal working load identification step, including identification of unknown working load source and identification of known working load source;

[0068] S3, abnormal working load isolation step, the file access request R marked as abnormal is put into a separate queue Q, and the queue Q is marked; In S3, if the working load does not match the strategy, and the user confirms, the label of the working load and the behavior of the user executing the strategy are automatically generated and added to the strategy library;

[0069] S4, abnormal working load processing step, for each Q, strategy matching is performed, and the strategy execution method includes reducing execution priority, limiting execution speed, suspending execution, canceling execution, and queue optimization; If no strategy is matched, the default is to take the suspend execution strategy, and an alarm is sent to the user and administrator through the alarm system, the abnormal working load is confirmed by the user and administrator, and the execution strategy confirmation is given; If multiple strategies are matched, the highest priority action is taken.

[0070] Among them, the abnormal working load identification method for unknown working load source includes the following steps:

[0071] Step one, when the user accesses the data of the tape file system, the following information can be extracted from each file access request (R) received from the tape file system: tape number, head addressing distance (L) of the tape, access data volume (S), record the time T of loading a tape in the tape library m , time T m contains the average time of taking out the existing tape in the drive and loading a new tape in the tape library, record the average speed V of addressing in a tape in the tape library s , the speed V of sequential reading and writing of the tape rw , T m , V s , V rw can be obtained from the manufacturer according to the model and configuration of the tape library;

[0072] Step two, define the file access requests R from the same workload source in a period of time t as a set R t , the number of which is N;

[0073] Step three, calculate the ideal access time T according to the total amount of data accessed in Rx rw , the speed of tape sequential access V o , and the expected access time T E = T O × α, where α is the performance coefficient, which can be pre-set according to the tolerance to performance;

[0074]

[0075] Step four, for all access requests R in Rt, calculate the simulated access time Ts, where M is the number of different tapes required to be accessed in Rt

[0076]

[0077] Step five, if T s > T E , it is considered that an abnormal workload occurs.

[0078] Among them, the abnormal workload identification method for known workload sources is:

[0079] When the workload source matches the index, thumbnail process of Windows / MAC, it is marked as an abnormal workload;

[0080] When the workload source matches the system scan of the antivirus software, it is marked as an abnormal workload. It can also be automatically marked according to other rules of known workload sources

[0081] Among them, for known workload sources, if the workload source can match the corresponding strategy, the corresponding strategy is executed; if it matches the system scan process of windows / MAC or the scan process of antivirus software, the request execution is cancelled.

[0082] For unknown workload sources, when the Ts / TE calculated by the abnormal workload is more than 5 times and less than 20 times, the priority of the request in Q is continuously reduced according to the multiple ratio; when the Ts / TE calculated by the abnormal workload is more than 20 times, the request execution is suspended; when the Ts / TE calculated by the abnormal workload is less than 5 times, the order optimization of the request tasks in the queue Q is started, such as sorting from small to large according to the L of each R to optimize the addressing of the tape.

[0083] The above embodiments are only for illustrating the technical concept and characteristics of the present application, and the purpose is to enable those skilled in the art to understand the content of the present application and implement it, and cannot limit the protection scope of the present application, and any equivalent changes or modifications made according to the spirit and essence of the present application shall be covered within the protection scope of the present application.

Claims

1. An intelligent management system for magnetic tape library file system abnormal workloads, characterized by: The abnormal access training unit includes using a special test environment to simulate running specific workloads, and performing workload classification training, the abnormal access training unit includes: A client server, the client server includes: a Windows thumbnail / index service module, a MacOS index / QuickLook service module, a malicious scanner module simulating a malicious scanning application, a normal program that accesses files by a normal application, a storage pass-through module that intercepts access file related information, an application information collection module for collecting additional application process information, and a disk file module with original files, the disk file module is exposed to users through the storage pass-through module; An AI training server, the AI training server includes: an AI learning engine module for receiving related information intercepted by the storage pass-through module and additional application process information collected by the application information collection module, and performing learning and training based on the collected information; and a model file module for saving model files.

2. The intelligent management system for magnetic tape library file system abnormal workloads of claim 1, wherein: The related information includes at least one of the file type, file name, LBA, size, timestamp, process id, uid, and gid information of the accessed file intercepted by the fuse file system pass-through technology.

3. The intelligent management system for magnetic tape library file system abnormal workloads of claim 1, wherein: The additional information refers to collecting additional information of the application process given the process id, uid, and gid information, including the process name and the corresponding operating system service name of the process.

4. The intelligent management system for magnetic tape library file system abnormal workloads of claim 1, wherein: The AI learning engine module performs machine learning and training based on the collected related information and additional information, and the learning algorithm includes a gradient boosting tree algorithm, a random forest algorithm, and a deep learning model algorithm using one-dimensional convolutional neural network (1D-CNN) and LSTM joint modeling algorithm.

5. The intelligent management system for magnetic tape library file system abnormal workloads of claim 1, wherein: It also includes a main work unit, the main work unit includes a storage server and a tape library, the storage server includes a tape library management module, the tape library management module includes: The tape library management module includes storage hierarchical management, abnormal access mode identification, and migration back migration scheduling management, the storage hierarchical management automatically migrates cold data to tapes according to a certain strategy, the abnormal access mode identification identifies and judges the mode according to the read LBA information of one or more files combined with the pre-trained and saved model, and the migration back migration scheduling management schedules the migration back migration request according to the idle of the tape library device resources and the results of the abnormal access mode identification to perform migration or back migration scheduling.

6. The intelligent management system for magnetic tape library file system abnormal workloads of claim 5, wherein: The abnormal access mode identification can be associated with the information of the application information audit module for overall judgment.

7. The intelligent management system for magnetic tape library file system abnormal workloads of claim 5, wherein: The main work unit also includes a main client server, the main client server includes: An NFS / SMB client; A main application information auditing module is used to record process id, service name, file name, offset, size, timestamp information of accessing the NFS or SMB mounted file system; And Windows / Linux / MacOS operating systems.

8. A method for intelligent management of a magnetic tape library file system abnormal workload, the method comprising: The intelligent management system of any one of claims 1-7, comprising the following steps: ​ S1, a file access information acquisition step is performed through a storage access layer, and information related to all access requests for a tape file is recorded at a storage layer through a fuse technology; an AI engine classifies and identifies the tape file access requests as a work load, and marks file access requests identified as the same work load as a work load source; the AI engine simulates running a certain specific work load through an abnormal access training unit, then captures the read-write characteristics of the work load to train the work load classification, and generates a coping strategy in advance according to whether the work load is friendly to the tape library and adds the strategy to a strategy library; S2, an abnormal work load identification step, including identification of an unknown work load source and identification of a known work load source; S3, an abnormal work load isolation step, in which the file access request R marked as abnormal is put into a separate queue Q, and the queue Q is marked; S4, an abnormal work load processing step, in which, for each Q, a strategy is matched, and the strategy execution method includes reducing the execution priority, limiting the execution speed, pausing the execution, canceling the execution, and queue optimization; if no strategy is matched, the pause execution strategy is adopted by default, and an alarm is sent to the user and the administrator through an alarm system, the abnormal work load is confirmed by the user and the administrator, and an execution strategy confirmation is given; if multiple strategies are matched, the highest priority action is taken.

9. The method for intelligent management of magnetic tape library file system abnormal workloads of claim 8, wherein: In S3, if a work load does not match a strategy and is confirmed by the user, the label of the work load and the behavior of the user executing the strategy are automatically generated and added to the strategy library.

10. The method for intelligent management of magnetic tape library file system abnormal workloads of claim 8, wherein: The abnormal work load identification method for the unknown work load source includes the following steps: Step one, when the user accesses the data of the tape file system, the following information can be extracted from each file access request (R) received from the tape file system: the number of tapes, the head addressing distance of the tape (L), the amount of data accessed (S), the time Tm of loading a tape in the tape library, the time Tm including the average time of taking out the existing tape in the drive and loading a new tape in the tape library, the average speed Vs of addressing in a tape in the tape library, the speed Vrw of sequential reading and writing of the tape, Tm, Vs, and Vrw can be obtained from the manufacturer according to the model and configuration of the tape library; Step two, define the file access requests R from the same work load source in a period of time t as a set Rt, and the number is N; Step three, the ideal access time T is calculated according to the total data amount accessed in Rx The ideal access time T is calculated according to the speed Vrw of the sequential access of the magnetic tape o and the expected access time T E = T O x a, where a is a performance coefficient, which can be preset according to the tolerance to performance; Step four, calculate the simulated access time Ts for all access requests R in Rt, where M is the number of different tapes needed to access in Rt; Step five, if Ts>TE, it is considered that an abnormal work load has occurred.

11. The method for intelligent management of magnetic tape library file system abnormal workloads of claim 8, wherein: The abnormal workload identification method for known workload sources is: When the workload source matches the index of Windows / MAC, the thumbnail process, it is marked as an abnormal workload; When the workload source matches the system scan of the antivirus software, it is marked as an abnormal workload.

12. The intelligent management method for magnetic tape library file system abnormal workloads according to claim 10, characterized in that: For known workload sources, if the workload source can match the corresponding strategy, the corresponding strategy is executed. For unknown workload sources, when the Ts / TE calculated by the abnormal workload is more than 5 times and less than 20 times, the priority of the request in the Q is continuously reduced according to the multiple ratio; when the Ts / TE calculated by the abnormal workload is more than 20 times, the request execution is suspended; when the Ts / TE calculated by the abnormal workload is less than 5 times, the order optimization of the request tasks in the queue Q is started.

Images