Research On Storage Managment Method Of Intelligent Network Disk

With rapid development of networking and digitalization technology, a large number of information resources are provided for all users around the world by means of Internet. It brings to many people's great attention that how to handle and store all of these mass valuable data. Therefore, mass information storage technology has become a great important research orientation and technical hot spot in the area of the computer science and technology in recent years.At present, in many of technical schemes to solve mass information storage technique problem, the network cluster technology is the most popular achievable way. However, storage performance bottle neck and single node failure still exist in network cluster storage system to a certain extent. So, how to efficiently conquer storage performance bottle neck and single node failure in network cluster storage way is a very important research work with major theoretical and practical significance in the research area of computer storage technology now.Intelligent Network Disk (IND) storage system, which was presented by author's reseach group, is a novel intelligent mass information storage architecture. Besides possessing the characteristics of general network cluster storage system, the network storage system of IND make adopt a series of intelligent control algorithms to prevent it from storage performance bottle neck and single node failure, to sustain system load balancing and to achieve intellectualization information storage of network storage system as a whole. Hence, IND storage system is a comparative ideal network cluster storage system.The storage management techniques of IND storage system is studied systematically and deeply in this thesis, such as embedded model of IND storage system, IND intelligent storage management algorithms, compound load balancing strategy suited to IND characteristics, cache regulation mechanism to eliminate system hot spot bottle neck and application of IND storage system in video server and high performance computing etc. Main research results and innovation achievements in this thesis are as follows:1. An IND architecture model is proposed. It can be used to construct storage system flexibly to realize active storage. For realizing storage virtulization, IND storage system is a new way for achieveing mass storage. A series of tests are made in detail and provide adventages for further exploration.2. The intelligent storage management algorithams of IND storage system are provided. To realize coordination among INDs and improve system storage efficiency as a whole, a series of IND intelligent storage management algorithms based on thorough research of storage management on which Bloom filter and Hash table make a notable effect are designed to control process of read and write in system and deal with user's high concurrent requests automatically. Mean while, management strategy of centralized management and decentralized service is adopted to eliminate existing storage performance bottle neck and single node failure. A lots of test show that this technique above is a feasible storage technique for achieveing mass storage at present situation.3. A compound load balancing strategy accorded with IND charactristic is proposed. The load of system is allocated rationally according to IND capacity and task characteristic. Considering time and space factor, a schedule strategy with dynamic regulation load balancing based on statistic of access frequency of read and write request is used to sustain storage system's load balancing and to distribute load evenly among INDs. Load skewed caused by high concurrent read and write request is migrated by regulating a quantity of files among INDs in time. The load balancing strategy with dynamic regulation based on access frequency of read and write request don't disturb system process and don't increase system overhead. Tests show that system run stably and normally.4. A cache regulation mechanism to eliminate system hot spot bottle neck is put forward. While read request is frequent in system such as IND used for Web server and video server, a dynamic regulation scheme with cache is used to eliminate existing bottle neck caused by hot spot, deduce time delay and migration communication overhead and improve storage system performance.5. An IND VoD is designed and run normally and satisfactorily, It reflects the characteristics of IND storage system comprehensively and indicates the application prospect of IND mass storage system. Moreover, a scheme of high performance cluster computing system of IND mount is suggested. By means of storage testing for matrix operation with I/O intensive in high performance computing, the expect result is obtained.The research results above show that storage system of IND make is a comparative ideal network cluster storage system and a good enlightenment from it is gained for further study.