Enhancing Availability for Distributed Replicated Services Considering Network Edge Availability

Abstract

Mechanism to improve data or service availability is critical for an enterprise to ensure the quality of service in terms of availability. Replication has been used to improve system availability. The number and location of the replicas are two impact factors on availability. In this paper, we will consider the impact of the node and network edge failures on the availability of replicated data or services. The Effective availability modeling approach is designed and efficient availability computing algorithms are developed to model and compute availability of replicated services for systems with the tree topology. The availability enhancement problem (maximizing the objective function) is transformed to a p-median problem (minimizing the objective function) through re-define the availability enhancement problem. An efficient replica allocation algorithm is developed to improve data availability in tree networks, with a runtime complexity of O(K|V|2), where K is the number of replicas and |V| is the number of nodes in the tree network. Finally, experimental studies have been conducted to evaluate how efficient and effective the proposed availability computing algorithm and the availability enhancement algorithm on improving the availability of replicated data or services. The results show that the proposed solutions are efficient and effective on availability computing and availability enhancement