Extending the Concept of Voting Structures to Support Path-Based Replication Strategies

Data replication is an important concept in distributed computing to provide fault-tolerant, highly available, and low cost data access operations, since a unique master copy can fail. Although the field of data replication is very well researched, the question of application-optimized replication strategies for specific scenarios still stands. There are many scenarios which cannot be satisfactorily managed by known replication strategies. Recent research has proposed the use of voting structures for generating new replication strategies from known ones, such that a wide range of scenarios can be addressed. However, unfortunately, voting structures cannot represent path-based replication strategies in an intuitive way. But path-based replication strategies have been proven to be very flexible and efficient in providing low operation costs and high availabilities. Thus, some very good characteristics cannot yet be covered by the above-sketched approach. In this paper, we present an extension to the concept of voting structures that allows the design of path-based replication strategies in an easy and intuitive way. Furthermore, we introduce a general design pattern that transforms each path-based replication strategy, given by a grid structure, to an efficient equivalent voting structure of the extended concept. The proposed approach is proven as well as its use demonstrated.