Font Size: a A A

Design And Implement Of Data Channel In Object-based Storage System

Posted on:2009-05-22Degree:MasterType:Thesis
Country:ChinaCandidate:W ZhangFull Text:PDF
GTID:2178360275471906Subject:Computer system architecture
Abstract/Summary:PDF Full Text Request
With explosive growth of the digital information and continuous improvement of application requirements, the traditional network storage systems face a huge challenge in capacity, performance, scalability, security, quality of service and so. Object Storage Technology ,which uses a new interface objects, is considered as the next generation network storage technology standards. It has become a new hotspot that how to design an efficient Object Storage Technology systems ,which can give full play to the advantages of object storage technology to meet the growing needs of the mass data storage. The Object Storage Technology systems is formed by four parts: MDS ,OSD, Client and the Internet.In the Object Storage Systems, the data channel, where the communications between the file system of Client and OSD go through, is one of the keys which affects the whole system performance, which including the transmission protocol used by the communications between the two and the buffer strategy of internal use. After researching operating mechanism of key part of HOSS, we use a smaller granularity of buffer strategy and revenue UNH iSCSI protocol modules to re-design and implement the data channel and the buffer.The comparison results of the between the Object Storage Systems before and after the HOSS improving shows, the improved system performance in the writing increased 1.65 times of the largest when file size is 52KB,as the reading increase of 85% when file size is 256MB. It Also compares the impacts to the read / write performance by packet length and number of OSD .It was also proposed algorithms and ideas to solve the cache consistency problem of existing system HOSS.
Keywords/Search Tags:Object Storage, internet Small Computer System Interface, Buffer Strategy, Cache Consistency
PDF Full Text Request
Related items