Metadata management method and device and computer readable storage medium

Abstract

The embodiment of the invention discloses a metadata management method and device and a computer readable storage medium. The metadata management method comprises the steps of obtaining load values and heat values corresponding to all metadata servers respectively in a clustering system; selecting overload metadata servers and light load metadata servers according to a preset threshold value and the load values corresponding to all the metadata servers; calculating an average load value according to the load values corresponding to all the metadata servers; migrating a corresponding number ofcatalogue subtrees in the overload metadata server to the selected target metadata server according to the average load value, the preset threshold value, the load values and the heat values corresponding to the overload metadata servers and the load values and the heat values corresponding to the light load metadata servers. In the process, the heat conditions of all the metadata servers are fully considered, therefore the catalogue subtrees can be more reasonably migrated, the frequency of migrating the catalogue subtrees is effectively reduced, and the efficiency of load balance is improved.

Description

technical field [0001] The present invention relates to the technical field of storage clusters, in particular to a metadata management method, device and computer-readable storage medium. Background technique [0002] Distributed file system (distributed file system, CEPH) provides unified, large-capacity, high-performance, and high-reliability file services by organizing the resources of multiple machines to meet the requirements of large-scale applications. [0003] CEPH usually consists of a metadata server (MetaDataServer, MDS) cluster and a storage server cluster. Statistics show that in file system accesses, accesses to metadata account for 50-80% of all accesses. In order to cope with a large number of metadata operation requests and ensure good performance and scalability, the management method of metadata and the load balancing strategy of the MDS cluster are extremely important. [0004] In the traditional way, the migration algorithm assumes that the capabiliti...

AI Technical Summary

Problems solved by technology

When migrating the directory subtree corresponding to the metadata from a high-load and strong-capability MDS to a low-load and weak-capability MDS, the weak MDS may accept too much metadata load and quickly become a new system bottleneck; Although in the new round of migration activities, the directory subtree will be migrated from the weak MDS, but a lot of network traffic is still wasted and the load balancing time is prolonged

When the directory subtree is migrated from a high-load and weak-capacity MDS to a low-load and strong-capability machine, although it will not cause waste of network traffic, in fact, more heat can be migrated to the past, and the migration is not fully utilized. Opportunity

Images