Router supported data regeneration protocols in distributed storage systems

Distributed storage systems provide reliable data storage by keeping redundancy over a large number of storage nodes in the network. Due to the large number of storage nodes, they can not be regarded as always available. Compared with replications, erasure codes can achieve higher data integrity, such that any k among n storage nodes can recover the original data. Moreover, in order to maintain a consistent degree of redundancy, it is necessary to regenerate new data in a replacement node after a node failure. However, erasure codes require k blocks to be transmitted from other storage nodes to generate a new block, resulting in a high volume of network traffic and a significant bandwidth bottleneck in the network. In this paper, we present a novel idea that we can both save the traffic and time consumed during the regeneration with the support of routers. With the enhanced computing ability, some reconstruction operations in the regeneration process can be performed by supporting routers rather than by the replacement node solely. We also design two router supported data regeneration protocols that make supporting routers provide support transparently to storage nodes and utilize the bandwidth diversity in the network, respectively. Simulation results show that with the help of supporting routers and our protocols, the traffic and time spent during the regeneration can be both reduced significantly.