Research On The Key Techniques Of Storage Network Based On Active Network Techniques

Storage network is the combination of storage techniques and network techniques. With the increase of various new services and applications because of the improvement of digital and network technologies, the storage network need not only more storage capacity, but also the high capability of data access and transfer, management, high reliability, availability and scalability. Since the storage mode has been changed from DAS to NAS and SAN through storage integrated, the more intelligent services need to be deployed in the storage network and the final purpose is to provide storage application service in the storage network. Timely and effective employment of various new services and applications is seriously limited by the lack support within traditional storage network techniques. Active networks represent a quantum step in the evolution of network by providing a programmable network interface in network nodes. The functionality of network nodes can be dynamically constructed and refined, and it could provide the more rich process capability with the packets. Users can program the network on the fly to construct and refine the special services that they need. In this thesis, the key techniques of storage network based on active network techniques have been studied comprehensively and systematically in the aspect of theory and method. The architecture of current active networks has been investigated in-depth. The active network node has been studied detailedly, including node operating system, execution environment and active application. Then some key parts of active networks have been investigated and analyzed, including active network encapsulation protocol, programmable network application programming interface and the packet process in active network node. The relative features have been contrastively analyzed between active networks and traditional packet network in order to provide research method to the key techniques of storage network based on active networks techniques.The architecture and reliability analysis of RAID has been investigated and the necessity of adopting multiple dimensional parity RAIDs in mass storage system has been given too. Then the innovation RAID structure baseded on active network techniques (AN-RAID) has been proposed, which is based on analyzing the fault-tolerance and reliability of multiple dimensional parity RAIDs in commom use. The special storage functions have been programmed and constructed on the active network node deployed on the storage network. Then the active network nodes have combined to form multiple dimensional parity networks RAID 5 structure. It is fit to high reliability and mass data storage requirement by theoretically analyzing fault-tolerance, the reliability and I/O performance of AN-RAID structure.The storage network congestion control model based on active networks techniques has been proposed to solve the problem that TCP congestion control in storage network based on traditional packet-switched network techonlogy, including congestion detect and control based on active networks techniques (ACDC) and active speed adjustment (ASA). ACDC provides congestion detection and congestion control mechanism in active network nodes comparing with traditional TCP congestion control. ACDC can not only precisely detect the network congestion and timely notify the sources, but also provide more effectively congestion detection and control. Therefore ACDC improves the average network throughput and decreases packet delay of storage network and solve the problems in traditional TCP congestion control, such as low use rate of network bandwidth, worse equability, congestion synchronization and worse adaptability of burst data flow. ASA introduces active detection packet in sender and adjust sending speed according to bandwidth of storage network. ASA can not only reduce the possibility of storage network congestion, but also improve the use ratio of storage network bandwidth. The speed adjustment algorithm has been proposed by investigating the transfer speed...