Evaluating the Reliability of Storage Systems

Modern storage systems are often large complex distributed systems. Current techniques for evaluating their reliability function require the solution of a system of differential equations. We present a more elementary, intuitive approach that focuses on the steady-state behavior of each storage organization when it goes through repeated cycles of failures succeeded by repairs. As a result, our approach provides immediately a purely algebraic method for computing both the average failure rate and mean time to failure. We show how to apply our technique to model the high infant mortality of disk drives and the behavior of the so-called S.M.A.R.T. drives, which can warn users of impending disk failures. 1 J.-F. Pâris is with the Department of Computer Science, University of Houston, Houston, TX 77204, USA. 2 T. J. Schwarz is with the Department of Computer Engineering, Santa Clara University, Santa Clara, CA 95053, USA. 3 D. D. E. Long is with the Department of Computer Science, University of California, Santa Cruz, CA 95064, USA. 1 EVALUATING THE RELIABILITY OF STORAGE SYSTEMS J.-F. Pâris 1 , T. J. E. Schwarz 2 , and Darrell D. E. Long 3 Abstract Modern storage systems are often large complex distributed systems. Current techniques for evaluating their reliability function require the solution of a system of differential equations. We present a more elementary, intuitive approach that focuses on the steady-state behavior of each storage organization when it goes through repeated cycles of failures succeeded by repairs. As a result, our approach provides immediately a purely algebraic method for computing both the average failure rate and mean time to failure. We show how to apply our technique to model the high infant mortality of disk drives and the behavior of the so-called S.M.A.R.T. drives, which can warn users of impending disk failures.Modern storage systems are often large complex distributed systems. Current techniques for evaluating their reliability function require the solution of a system of differential equations. We present a more elementary, intuitive approach that focuses on the steady-state behavior of each storage organization when it goes through repeated cycles of failures succeeded by repairs. As a result, our approach provides immediately a purely algebraic method for computing both the average failure rate and mean time to failure. We show how to apply our technique to model the high infant mortality of disk drives and the behavior of the so-called S.M.A.R.T. drives, which can warn users of impending disk failures. Index Terms Fault-tolerant systems, storage systems, repairable systems, k-out-of-n systems.