Robust data sharing with key-value stores

Abstract

A key-value store (KVS) offers functions for storing and retrieving values associated with unique keys. KVSs have become widely used as shared storage solutions for Internet-scale distributed applications. We present a fault-tolerant wait-free efficient algorithm that emulates a multi-reader multi-writer register from a set of KVS replicas in an asynchronous environment. Our implementation serves an unbounded number of clients that use the storage. It tolerates crashes of a minority of the KVSs and crashes of any number of clients. We provide two variants of our algorithm: one implementing an atomic register and one implementing a regular register; the latter does not require read operations to store data at the underlying KVSs. We note that applying state-of-the-art reliable storage solutions to this scenario is either impossible or prohibitively inefficient. © 2011 Authors.