Font Size: a A A

Design And Implementation Of Metadata Management In Global Distributed Storage System

Posted on:2005-10-16Degree:MasterType:Thesis
Country:ChinaCandidate:Y J JiaFull Text:PDF
GTID:2168360152469138Subject:Computer system architecture
Abstract/Summary:PDF Full Text Request
Data-intensive, high-performance computing applications require the efficientmanagement and transfer of terabytes or petabytes of information in wide-area, distributedcomputing environments. Storage virtualization integrates all kinds of high performance storagesystem to a unit one. It can not only share resources and make fully use of resources, butefficiently avoid conflict of data explode and limited storage ability. To resolve these conflict,After two years of research on storage virtualization, we developed GDSS(Global DistributedStorage System). In GDSS, metadata is very important for it takes charge with characterization of data, soreplica for metadata is a must. Although replica can fault tolerance and load balance, it alsobrings the problem of replica coherence. At the same time, In many systems, such as SRB,hierarchical directory structure is adopted. Generally when the metadata becomes enormous, thesystem employs several metadata servers. There are several limitations for directory servers.One is that it must keep the logic tree among the directory servers; the other is that the directoryservers must cooperate and the result will return from the root node that adds the overload to theroot server; the third is that when the root server is out of service the whole meta server will outof service too, so it is difficult to provide high availability; and the last is that it is difficult toexpand the scale of meta servers. Based on those problems, GDSS introduce a novel hierarchy model MDC(MetaDataController)to manage metadata. In MDC, MatchTable module is charged with communicationof SSP(Storage Service Point) and DS(Directory Server), and cache module is charged withincrease of metadata access efficiency and as a assistant to maintain the coherence of metadatareplica. MatchTable saves name and root of each DS, and cache saves hot metadata andmetadata being updated. Update operation is only done on cache, and DS is updated by replicacoherence module, which can avoid conflict of synchronously access to replicas. If cachedoesn't hit, access a seemly DS through MatchTable. MDC can reach better efficiency,scalability and management.
Keywords/Search Tags:Storage virtualization, Global Naming server, Metadata, Hierarchic Management, MatchTable
PDF Full Text Request
Related items