Method, device and medium for adjusting main pg equalization of distributed storage system

Abstract

The invention discloses a main PG balance adjustment method for a distributed storage system, Devices and media, the method presets a desired number of primary PGs for each OSD, the initial value of the weight value of the expected deviation and the primary PG affinity, and then determining whether to adjust the weight value of the current primary PG affinity of the OSD according to the relationship between the absolute difference between the current primary PG number of each OSD and the expected primary PG number and the expected deviation, The primary PG number distribution on each OSD is adjusted by adjusting the weight value of the current primary PG affinity until the weight value of the current primary PG affinity for each OSD does not need to be adjusted. Therefore, compared with the prior art method of randomly distributing the main PG number on each OSD, the method applied to each OSD can gradually approach the expected main PG number, and the maximum deviation is the expecteddeviation, thereby realizing the equalization adjustment of the main PG.

Description

technical field [0001] The present invention relates to the field of distributed storage systems, in particular to a method, device and medium for balancing and adjusting master PGs of a distributed storage system. Background technique [0002] As distributed storage systems are widely used in video surveillance, high performance, broadcast media, cloud computing, big data, virtualization and other fields, customers have higher and higher performance requirements for distributed storage systems. The performance of the storage system is becoming more and more important. [0003] PG (Placement Group) is a very important concept in distributed storage systems. It can be regarded as a virtual node in consistent hashing, which maintains a part of data and is the smallest unit of data migration and change. It plays a very important role in a distributed storage system. In a distributed storage system, there are a certain number of PGs (which can be dynamically increased or decrea...

AI Technical Summary

Problems solved by technology

For example, when the read hit rate of some OSDs in a distributed storage system reaches 100%, there are some OSDs whose read hit rate is only 70% or even lower; generally, the lower the read hit rate of an OSD, the lower its average delay. High, so these OSDs with low read hit rate become the performance bottleneck of the entire distributed storage system

Images