Font Size: a A A

High-availability Web-based Floppy Disk Arrays

Posted on:1998-11-24Degree:MasterType:Thesis
Country:ChinaCandidate:M Y ChenFull Text:PDF
GTID:2208360185495485Subject:Computer organization and system architecture
Abstract/Summary:PDF Full Text Request
Hard disk devices are needed in nearly all computer systems. But in recent years the perfonnance of hard disk has been improving very slowly due to the mechanical restrictions and has been left far behind the VLSI and network technology. RAID system , which based on parallel and redundancy technology , can overcome the shortcoming of hard disk to some extent. But the centralized structure restricts its further improvement . In this thesis , a network-based software disk array system in a cluster environment is discussed. By using distributed resource and interface, this system may provide a scalable , high performance , high available secondary storage system.This thesis first reviews the current state of hard disk device and RAID system . Then the current network and cluster technology is introduced. Also the data rate and I/O rate concerning the performance of disk storage system in each system are given .Next, this thesis examines the reliability of multi-error-recoverable software disk array. It is showed that due to the software failure rate a high available software disk array must use an at least two level error recoverable mechanism. And the node failure recovery time should be minimized at the same time.This thesis also introduces some multi-erasure-correcting codes for disk array, including the Reed-Solomon code, array code , and a fast 2-erasure correcting code — EVENODD code. A parity declustering method to avoid large system bottleneck is reported too.Another emphasis of this thesis is the realization of network-based software disk array . First the interface and difficulties are analyzed. Then a fast recover methods based on log information is provided. In order to solve the small write problem in disk array, this thesis analyzes two methods which use data reorganization to reduce random disk I/O times . The author then provides a new method which use network redundant buffer to increase the system performance.Finally , this thesis describes the author's prototype system of an network mirror disk based on NCICHA system . The system structure and some test results are given.
Keywords/Search Tags:High-availability
PDF Full Text Request
Related items