Redundancy and Aging of Efficient Multidimensional MDS Parity-Protected Distributed Storage Systems

Abstract

The effect of redundancy on the aging of an efficient Maximum Distance Separable (MDS) parity--protected distributed storage system that consists of multidimensional arrays of storage units is explored. In light of the experimental evidences and survey data, this paper develops generalized expressions for the reliability of array storage systems based on more realistic time to failure distributions such as Weibull. For instance, a distributed disk array system is considered in which the array components are disseminated across the network and are subject to independent failure rates. Based on such, generalized closed form hazard rate expressions are derived. These expressions are extended to estimate the asymptotical reliability behavior of large scale storage networks equipped with MDS parity-based protection. Unlike previous studies, a generic hazard rate function is assumed, a generic MDS code for parity generation is used, and an evaluation of the implications of adjustable redundancy level for an efficient distributed storage system is presented. Results of this study are applicable to any erasure correction code as long as it is accompanied with a suitable structure and an appropriate encoding/decoding algorithm such that the MDS property is maintained.