Repair-efficient distributed storage codes with heterogeneous reliability requirements

The digital contents in a large scale distributed storage systems usually have different reliability requirements (e.g., new customer billing records vs. 10-year-old office document backup), and for this reason, erasure codes with different strengths should be utilized to achieve the best storage efficiency. On the other hand, in such large scale distributed storage systems, nodes fail on a regular basis, and the contents stored on them need to be regenerated and stored on other healthy nodes, the efficiency of which is an important factor affecting the overall quality of service. In this work, repair-efficient data storage codes are considered in systems with heterogeneous reliability requirements. We formulate the problem of multi-reliability regenerating (MRR) codes and investigate the optimal storage vs. repair-bandwidth tradeoff. One key question is whether contents with different reliability requirements need to be “mixed” in the optimal solution, for which we show that such a mixing can strictly improve upon the non-mixing solution.