Erasure Codes-Based Storage-Aware Protocol For Preserving Cloud Data Under Correlation Failures

Abstract

Distributed storage of cloud data imposes a great challenge for assuring data availability and reliability in the face of stochastic failures and attacks. Although redundancy scheme such as erasure codes or network coding schemes can be used to improve storage efficiency, they do not consider the scenario that the network may be split up into several regions due to correlated failures of storage components, which may not successfully reconstruct the original data. Furthermore, the stored data on the failed node need to be recovered to keep the same redundancy of network data while introducing low overhead. In view of this, we first devise an efficient algorithm that guarantees residual network has enough data segments to recreate the original data under correlated failures. Second, an efficient data recovery scheme is presented to repair the data stored on the failed nodes in case of Byzantine failures or pollution attacks. The proposed protocol benefits from several key features such as less storage cost, competitive communication cost, and much better data robustness comparing to the state of the art distributed storage systems.