Efficient And Semantic Olap Aggregate Queries In Peer To Peer Networks

Today, the ultimate way to retrieve and share resources in a distributed system is the use of Online Analytical Processing (OLAP) tools. OLAP is a multi-dimensional analysis and decision support tool, which built on the basis of data warehouse. OLAP is also the primary means of user’s of access to decision support. A data warehouse is a large repository of information containing data collected from different data sources over a long period of time. The query performance in such an environment is a critical challenge as most of the associated queries are complex and adhoc in nature and require huge volumes of data to be processed. Database systems have traditionally used a Client-Server architecture, where clients send queries to a database server. In this architecture, the server may become overloaded, leading to clients experiencing an increase in query response time. OLAP technology architecture need be further improved to better meet the needs of decision analysis of users. In the other hand, in the domain of file sharing, P2P technology has come as a solution to overcome server load, where peer shares file or piece of file between each over so sharing the load. The potential benefits of a P2P database system include performance and scalability; allowing peers to answer each others’ queries will reduce the load on the database server and so could overcome the problem of a busy server becoming overloaded. Other potential benefits are fault tolerance and cost reduction. Our work is based on query performance issue.Based on the above mentioned issues, this thesis carried out the followed researches:We propose a TrackerCache system, which provides a central component called tracker, which acts as a directory, keeping track of all submitted queries and a list of peers caching their results. The key idea is to enable peers to share their local caches, to form an efficient and semantic distributed cache sharing system. For this issue, a query forwarding policy is introduced.Then, to face inconsistent information in the client side cache, we proposed a knowledge replacement policy based on Binary Decision Tree (BDT) to evict unwanted data.This thesis proposed an idea that all submitted OLAP queries and their results are cached, so that the new incoming queries are fast served. We observed that the system enhances the efficiency of decision analysis and reducing the OLAP server’s load.