Staying alive — Network coding for data persistence in volatile networks

Abstract

In volatile network environments, node connectivity and availability changes rapidly. This poses a challenge to efficient repair of failed nodes in distributed storage with erasure-correcting code. The commonly considered bandwidth-efficient data regenerating schemes either rely on availability of specific storage nodes in the network, or require a substantial number of repair nodes. We propose a RLNC storage scheme, which can operate well with highly changing network connectivity, and allows data decoding even after many failures and repairs. We demonstrate its substantial performance gain over the existing regenerating schemes.