A dynamic file replication and consistency mechanism for efficient data center operation and its formal verification

Abstract

Data centers are built to provide a highly available and scalable infrastructure on which the applications run. As enterprises grow, so do their application need, along with resources required for additional application-specific services. This increase in bottom line expense heightens the overall resource requirement. This paper provides solution to mitigate the impact of these expenses by proposing a file replication and consistency maintenance mechanism that enhances the manageability, scalability, and high availability of resource in these environments. To keep files consistent, changes made at one replica of the file are reflected on other replicas in minimum possible time. File replica is updated on-demand by only propagating the required partial updates. The results show that as compared to Google File System (GFS), the proposed partial write rate shows an improvement of 38.62% while updating the stale replicas. Time required for invalidating the replicas decreases by 34.04% and the time required to update the replica on FRS reduces by 61.75%. Process algebraic approach has been used to establish the relationship between the formal aspect of the file replicating server and its architectural model.