High-Reliability Distributed ISCSI Storage System Based On Double-Stage RAID

Global economy has entered the era of a post-industrial revolution, the information revolution era. Information as an important carrier of this revolution begins more and more important. Companies and individuals in their daily life generate lots of data, which are kept in the storage systems. With the explosive growth of storage system, system reliability becomes increasingly important. Larger system leads to greater likehood of damage of disks in the same time.911 terrorist attacks in the United States, Japanese Tsunami and other large man-made or natural disasters make people understand the devastating impact to the data by disasters more deeply and pay more and more attention to disaster tolerance of storage system.This dissertation focuses on the storage system reliability and system availability, and describes advantages and disadvantages of the RAID, which is commonly used to improve reading and writing speed, to avoid bottlenecks and to offer redundancy protection to storage system. Current main technology which are NAS, SAN, and IP-SAN is analysised, especially the iSCSI technology in IP-SAN.This dissertation discusses High-Reliability Distributed Network Storage System Based on iSCSI and Double-Stage RAID. As it is distributed storage, it can avoid unrecoverable loss by earthquakes and other disasters.The main work is divided into the following two aspects:1. This dissertation introduces the factors that determine the performance of the storage system, and puts forward to Markov Chain mathematical model of High-Reliability Distributed Network Storage System Based on iSCSI and Double-Stage RAID. The mathematical model is finally transformed into Matlab symbolic computation program. The reliability value of the system is ultimately calculated to demonstrate that the DRAID system has a higher level of reliability compared to the previous system.2. Linux system is used to achieve the implementation. In this dissertation, IOMeter testing tools is used to measure reading and writing performance. Through experiments the stripe unit size, the child node bandwidth are analysised on system performance. The system concurrency is also tested. Finally the fileds where the DRAID system can be used are described. Compared with other systems, the system has a higher level of redundancy and good performance.