The storage vs repair bandwidth trade-off for multiple failures in clustered storage networks

Abstract

We study the trade-off between storage overhead and inter-cluster repair bandwidth in clustered storage systems, while recovering from multiple node failures within a cluster. A cluster is a collection of m nodes, and there are n clusters. For data collection, we download the entire content from any k clusters. For repair of t ≥ 2 nodes within a cluster, we take help from local nodes, as well as d helper clusters. We characterize the optimal trade-off under functional repair, and also under exact repair for the minimum storage and minimum inter-cluster bandwidth (MBR) operating points. Our bounds show the following interesting facts: 1) When t|(m− ) the tradeoff is the same as that under t = 1, and thus there is no advantage in jointly repairing multiple nodes, 2) When t (m − ), the optimal file-size at the MBR point under exact repair can be strictly less than that under functional repair. 3) Unlike the case of t = 1, increasing the number of local helper nodes does not necessarily increase the system capacity under functional repair.

Rack-Aware Cooperative Regenerating Codes

“…A related version of storage-repair bandwidth tradeoff has been studied in the context of clustered distributed storage systems in [9]. The tradeoff for the case of multiple node failures where all the node failures are in a single cluster have been dealt with in [10].…”

Storage and repair bandwidth tradeoff for heterogeneous cluster distributed storage systems

“…In [17,24], the authors proposed algorithms for the cluster DSS model corresponding to the above two steps under the assumption that the newcomer connects to all the alive nodes, namely, d = n − 1. However, this assumption cannot be used to analyse the cases with multiple node failures, which is also frequent in realistic storage systems [20,32]. Thus we investigate more flexible situations in [22,28], where the applicability of the algorithms are proved under more general assumption…”

“…without changing the final tradeoff bound. Let α * denote the minimum α satisfying (20), the tradeoff bound is formulated as follows with similar methods introduced in [6,28]. For i = 2, .…”

“…In [1,18,19], the authors focused on minimizing the cross-cluster bandwidth in the multi-cluster model, where the relay node in each cluster was used to collect data from intra-cluster nodes and transmit to the newcomer in the repair procedure. In [16,20], Prakash et al analyzed the tradeoff bounds in detail when the relay nodes were used both in the repair and data collection procedures and connected to different amounts of intra-cluster nodes. On the other hand, the authors of [17,21,22] considered the cluster DSS model without relay nodes, where the intra-cluster and cross-cluster repair bandwidth were analyzed as two parameters in the repair procedure.…”

Storage-Repair Bandwidth Trade-off for Wireless Caching with Partial Failure and Broadcast Repair

“…In [11], the authors consider a centralized model of multiple-node failures that is equivalent to the broadcast repair model in [10]. Reference [12] investigates the storage-repair bandwidth trade-off for clustered storage networks, where multiple nodes within a cluster fail. This is close to the partial failure model that we consider since each cluster could model multiple memory units within a node, and multiple memory units failing is equivalent to partial failure in a node.…”