Optimal Replication Strategy Using Rough Set Theory for Dynamic Content Replication

Replication strategy is used to minimize the recovery time and data loss at an operational system and it attempts to reduce the system load, access latency and network congestion in a distributed system. Replication improves the reliability and scalability of the services. To increase availability we divide large files into small, medium and large fragments. When there are more number of hits for a data and when there is more delay in accessing a data we need to replicate it more. However these methods lead to high overhead for unnecessary file replication, fragmentation and consistency maintenance. Thus we place data on reliable servers. Any distributed database has three properties, consistency, availability and partition tolerance (CAP). In practice, at most two of these properties can be satisfied for any shared data. Thus a mathematical method called, Rough Set Theory (RST) can be used, that deal with vagueness and uncertainty in data and decision making. When available information in the database is insufficient, lower and upper approximations can be used to determine the exact value of a given set. Rough sets are created based on attributes and allocate memory for the sets. Based on ranking we decide whether it is fixed or variable size set, as we try to maintain time consistent data. The objective of this paper is to determine optimal data replicas using Rough Set Theory methodology, which groups data into sets of different granule size and generates a set of decision rules. This helps in determining which data set should be replicated more in order to provide better availability and consistent data. Keywords— Replication strategy, rough set theory, consistency of data