Network coding for data-retrieving in cloud storage systems

As the rapid growth of data, many storage systems have used erasure codes instead of replication to reduce the storage cost under the same level of reliability. Maximum-Distance- Separable (MDS) codes have been the most widely adopted, due to their optimal storage efficiency. It is well understood that the application of codes in storage systems, where the data is less frequently accessed. For the data, which is stored in cloud storage systems, is frequently accessed (or so-called “hot data”), the performance of data-retrieving is the key metric. To the best of our knowledge, there has been only a little work on the performance of data-retrieving in cloud storage systems with erasure codes. They combined queuing theory with coding theory to analyze and optimize the effect of MDS codes in the performance of data-retrieving. Their studies were based on MDS codes and they gave the solutions. In this paper, we transfer the perspective of study of based on MDS codes to that of optimizing MDS codes in order to improve the performance of data-retrieving, that is, from optimizing the system retrieving strategies to optimizing the coding schemes. We apply Network Coding to optimize the coding schemes and propose a new family of MDS codes, which reach optimal performance of data-retrieving in theory.