Provenance Based Rebuild: Using Data Provenance to Improve Reliability

Abstract

Traditionally, data preservation and reliability have used error correcting codes (ECCs) to ensure data safety. The development of general data provenance tracking systems provides a new opportunity for data reliability. We present a method that utilizes provenance to determine a datum’s generating process and inputs, and then uses this information to recompute lost data. This method, called Provenance Based Rebuild (PBR) provides a new, complimentary reliability mechanism that integrates with traditional systems to offer a variety of benefits including fine grained prioritized rebuild and parallel rebuild. While PBR offers benefits that address weaknesses in current techniques, it also faces a number of challenges such as data placement, and infrastructure provisioning.