Font Size: a A A

Client Design And Load Balance Research Of Object-Based Storage System

Posted on:2008-08-14Degree:MasterType:Thesis
Country:ChinaCandidate:L ChenFull Text:PDF
GTID:2178360272468975Subject:Computer system architecture
Abstract/Summary:PDF Full Text Request
The global information storage capacity get a 30% growth rate per year, and online storage capacity is doubling every nine months. The bottleneck of traditional Direct storage mode makes network storage becoming an inevitable trend. Object was used as an basic transfer unit in Object Based Storage model, which separates the data storage and metadata management As the middle course of traditional block interface and file interface, the object interface congregated the advantage of those two interfaces, and Object-Based Storage get a much better performance in security, usability and expansibility. Object-Based Storage System gathered the object interface and intelligence storage device, can get a petabyte scale storage capacity.Based on the analyses of current network storage architecture, analyzed Object-based Storage technology and T10 SCSI OSD standard command, constructed an T10 SCSI OSD standard accord storage system. Developed the user mode of storage access software, which can support the tripartite communications of OBS architecture, analysis the OSD standard T10 command, and finish an object-based data transmission through a iSCSI channel. Completed the buffer mechanism, multi-thread ,dynamic equipment acquisition design of client software. The performance of user mode client software is correspond with the POSIX semantic client.Realized an object storage device based load balance method, resolved the common problem of load imbalance in a mass storage system. After deep analysis of the load characteristic of Object-based storage nodes, get the real time load information of the storage node, designed an device select arithmetic and I/O schedule, test results shows the arithmetic make the Request Queue Length difference declined form 300% to 125%, the load difference declined from 247.2% to 122.2%, which means system loads were distributed uniformly among storage nodes.
Keywords/Search Tags:Object-based Storage System, Client Node, Load Balance
PDF Full Text Request
Related items