Storage-Repair Tradeoff for Hierarchical Distributed Storage Systems.

Real-world distributed storage systems (DSS) such as modern data centers are hierarchical in nature, in which storage nodes are organized in different racks and connected by a network. In such a hierarchical DSS, we focus on establishing the fundamental tradeoff between storage capacity and repair bandwidth. We first propose a new rack model with heterogeneous repair mechanism and construct information flow graph for it. By analyzing the information flow graph, we give a tight lower bound on the min-cut capacity for an arbitrary hierarchical DSS. Moreover, we formulate the problem of identifying storage-repair tradeoff as a bi-objective linear optimization subject to the min-cut constraint and solve it using weighted-sum method.