Research On Peer-assisted System And Data Availability Prob- Lem In P2P Storage System

With the development of network technology and improvement of bandwidth, applica-tion of the Internet has been changed enormously. Various kinds of data are deployed on the Internet and have attracted lots of users. The traditional single storage server model can hardly serve so many users concurrently. P2P based distributed storage is a reasonable candidate for those applications because of its scalability and high perfor-mance. Lots of research has been done on it.Because of the unpredictable user performance, P2P storage currently can not be used in real world. Some researchers begin to apply P2P storage as assisted system. With the increasing popularity of cloud computing, lots of companies have deployed cloud storage. Some cloud storage providers charge users not only for the occupied s-torage but also for the bandwidth transmitting their data out of cloud. The well-studied P2P distributed storage, as an assisted system, can utilize edge storage and edge band-width to efficiently alleviates cloud storage press and decrease the cost of transmitting data.Currently peer-assisted cloud storage systems mainly utilize unstructured P2P s-torage. In the past few years, lots of structure based P2P storage systems have been deployed. In this work, we investigate how to apply structure based P2P storage into peer-assisted system. In our system, data blocks have been placed sequentially on peers who are organized by Chord. With such store file protocol, we can efficiently retrieve the data. In any kinds of peer-assisted cloud storage, because the storage is composed by two parts, cloud storage and P2P storage, it should determine which part responses user’s request. In structure based P2P storage, without central tracker, it is hard to get the real-time information of system. This makes the bandwidth allocation problem is hard to be solved. In this work, we propose a prediction method to solve this problem.Advantage of developing P2P storage as assisted system is that we don’t need to take care of data availability, cloud storage will take response of it. However, currently, lots of cloud storage systems are based on P2P storage. In such system, cloud storage usually is composed by lots of inexpensive commodity storage machines, so storage failures are the normal phenomenon rather than the exception. When we design the cloud storage, we should take data availability into account.In this work, we provide a group and replication based mechanism to keep data available at any time. Adopting time slot model and assuming peer’s online state is pre-dictable, we present the MCN and FPR problem. We have proved all of those problems are NP-Complete. After this, restricting peers keeping online for a continuous period of time, we have designed a polynomial time algorithm to solve the MCN problem. Meanwhile, with the same restriction, we use PSA algorithm to enumerate the optimal solution of FPR.