寄生存储机制及底层实现技术-计算机系统结构专业论文.docxVIP

II II Abstract* The Internet is a dynamic complex system with many kinds of network protocols. While transferring data over the network by using the communication mechanisms of these protocols, the time taken for each message sent and echoed between local and remote nodes is nonzero. So by repeatedly sending and receiving data, one can make data sustained on the network for a period of time and thus form a novel storage mode—parasitic storage. Unlike traditional storage, parasitic storage needs no additional storage facilities. It just makes use of the existing resources to achieve high-capacity data storage without being sensed by the network. According to the characteristics of complex network, a novel networking storage model has been designed and implemented over Internet based on the protocol ICMP. By a highly scalable architecture, parasitic storage fully utilizes the available resources to build a free of cost, large space and short term data storage. For the dynamic of the network, a distributed data storage model adapted for parasitic storage is presented. The model defines several performance parameters to describe data distribution policy according to the basic requirement of parasitic storage, and establishes an estimation and decision mechanism for parasitic storage. Simultaneity, the data fault-tolerance strategy and dynamic replication management mechanism are also discussed. Then, the paper presents a novel decentralized dynamic replication management mechanism, which improves the data reliability, depress the read-write latency and keeps parasitic storage’s stabilization and high-effect. However, parasitic storage may affect the network to some extent and vary itself with the fluctuations of the network. To analyze the interaction between them, an analysis model has been constructed. With this model, we have researched the impact of parasitic storage on network and the dynamic characteristic of parasitic storage on complex network in the sense of pr