• PUB Student: Lucian Cancescu
• PUB Advisor: Alexandru Costan
• INRIA Advisor: Alexandra Carpen-Amarie, Bogdan Nicolae

Currently, BlobSeer uses a static replication factor associated with stored data for fault tolerance purposes. Users set this factor when writing new data into the system. While this approach proves reliable and simple to implement in uniform environments, its efficiency is arguable when dealing with large amounts of distributed data, over heterogenous resources with dynamic access patterns.

This project enhanced the BlobSeer system with self adaptive and autonomic capabilities by means of dynamically changing and maintaining the replication factors of the data. This is achieved by using real-time monitoring information related to the number of accesses to data providers, BLOBs and versions. Hence, when a certain BLOB or version of a BLOB becomes more popular (in terms of reads), the system automatically increases its replication factor and handles all the necessary data transfers to this end. In contrast, when some data is less (or never) used, its replication factor is transparently reduced.

More details: Thesis