Method for testing and adjusting cluster storage system performance based on synchronous strategy

Abstract

The invention provides a method for testing and adjusting cluster storage system performance based on a synchronous strategy. The method comprises a data node module group, a control module group and a client node module group, wherein the three module groups cooperate and communicate with each other, and take the control module group as the core to finish the system test and monitoring tasks together; the control module group is used for scheduling load and balancing all modules, and collecting the test result and synchronizing the time of all modules; the data node module group is used for collecting the data node monitoring information and realizing time synchronization; and the client node module group is used for simulating the read/write operation, collecting the monitored data of the client node and realizing time synchronization.

Description

technical field [0001] The invention relates to the field of computer application technology, in particular to a method for testing and adjusting the performance of a cluster storage system based on a synchronization strategy. The invention is applicable to cluster storage systems popular in the industry at present. Background technique [0002] A computer cluster, referred to as a cluster, is a computer system that is connected through a group of loosely integrated computer software and hardware to work closely together to complete computing tasks. At present, due to the massive growth of personal and enterprise data, the cluster storage system is developing rapidly. However, for the system performance test of the cluster storage system, the performance test method of a single node is basically followed, and then the performance test results of each node are accumulated. method to measure the performance of a clustered storage system. This method, although simple, is not r...

AI Technical Summary

Problems solved by technology

This method, although simple, is not rigorous and scientific enough

[0003] At present, popular performance testing tools such as iozone, iometer, etc., although capable of multi-threading and multi-process testing of read and write bandwidth, also provide more testing options, which can test the performance of random read, random write, sequential read, and sequential write Indicators, however, these tools only test the read and write bandwidth of a single node

If you need to test the performance of the entire cluster system, if you simply accumulate the performance test results of multiple nodes, the test results are often not accurate enough due to the asynchronous time and the asynchronous end of the process of each node

Images