LT-Hitchhiker's XOR framework for cloud storage systems

Abstract

Cloud storage is an important service of cloud computing where the data owners store their data in cloud servers so that users can access the data from cloud servers anywhere irrespective of their device and location such as thin clients with limited memory capabilities, CPU and limited bandwidth. The owners can also reduce the maintenance and burden of cloud storage. However this cloud storage is unreliable due to failures which occur in the cloud frequently. These failures are to be tolerated to prevent data loss. Although triple replication can be used to tolerate data loss, but it suffers from a very low storage efficiency. Erasure coding techniques like Reed Solomon Codes used to prevent data loss in distributed storage systems. These codes can achieve lower storage costs compared to replica but are inefficient for data reconstruction in cloud storage systems because of high repair costs. In this paper, we construct a new LT-Hitchhiker's framework that uses these both Luby transform(LT) codes and Hitchhiker's XOR codes together for efficient data reconstruction in cloud storage systems. These LT- Hitchhiker's XOR erasure codes can reduce the amount of data required for reconstruction than Hitchhiker's XOR codes and Reed Solomon Codes with no additional storage.