| As traditional internet services, networked multimedia and emerging network storage, network gaming, internet of things develops; human demand for data grows rapidly. For example, in the pharmaceutical industry, medical imaging data keeps growing with 15% -20% annual growth rates. And in the next 10-15 years, it will reach PB-level. Meanwhile, the growth rate of media, finance data is also very impressive. With the data requirements of industry continues a rapid development, PB storage service is imminent. According to CCID Consulting's research and analysis, in 2009 China's market demand for storage will grow up to $ 7.38 billion, with an annual growth of 18.5%.Almost all applications require the cluster to provide uninterrupted service. High availability cluster refers to both on the condition of the manually backup, system upgrades, maintenance, configuration changes, or passive hardware error, power outages, network failures and other cases, cluster will be able to continue to provide services, or to reduce downtime to a minimum.The research objectives of this project is to design corresponding high-availability security programs for object-based parallel mass storage system, which makes the same mass storage system continue to provide a stable storage service. Failures in the system can quickly self-recover to minimize system downtime. The most important issue for the management server and data server program to achieve high availability and specific implementation mechanism, and tested. Experimental results show that the solution meet the mass storage system management server and data server for high availability requirements.Firstly with the basic knowledge of studies conditions at home and abroad, and the products status, the paper researches on the network storage system architecture model and development. And the traditional high-availability clustering technology is summarized, including the disk data availability, network availability cluster, Cluster Server availability and other aspects.Based on the analysis of network storage, the paper explores the traditional mass storage systems, and the recent object-based parallel mass storage system, and analyzes advantages and system architecture features of the new object-based storage model compared to parallel storage systems for NAS storage cluster architecture and SAN shared storage architectures on the performance and scalability . The paper makes a detailed analysis of its architecture and mechanism for object-based storage model of parallel mass storage system.The project designed solutions to protect the corresponding high-availability for the mass storage system: programs on the disk using RAID 5 data redundancy; twin-engine by means of the switch to a single point to enhance the availability of equipment; through the redundant switches, redundant routers, redundant power systems, port redundancy, redundant twisted pair, fiber optic redundancy to guarantee high network availability; by the active/passive, N-way method of protection to insure management server, data server high availability.The paper then achieves high availability for server; through open source heart rate monitoring software and server resource management software, monitoring the server and the service to open and close, service migration managing. In a virtual machine environment, the project builds an experimental environment. Through the management server and data server downtime test, when the server is down it generally needs 5-6 seconds to restore the service, even this time can basically guarantee server availability. |
PDF Full Text Request