An on line social network usually requires many servers to store its user data so that a large number of users can be served. The data should also be replicated across the servers to provide a high degree of availability in case failures occur. In this paper we discuss the importance of social locality in the replication procedure. To substantiate this importance, we propose and evaluate a replication method that can significantly improve the system efficiency by leveraging the social relationships of the data.
Published in:
Distributed Computing Systems Workshops (ICDCSW), 2011 31st International Conference on
Date of Conference: 20-24 June 2011