Scalable Reliability Modelling of RAID Storage Subsystems

TL;DR

This paper introduces a scalable CTMC reliability model for RAID storage systems that accurately includes various components and failure modes, enabling efficient analysis of large configurations with up to 600 disks.

Contribution

The authors develop a scalable CTMC model for RAID systems that incorporates detailed component failure characteristics and uses state-space reduction techniques for practical computation.

Findings

Model scales to systems with up to 600 disks.

Uses Weibull and correlated failure modes for disks.

More practical than Monte Carlo simulations.

Abstract

Reliability modelling of RAID storage systems with its various components such as RAID controllers, enclosures, expanders, interconnects and disks is important from a storage system designer's point of view. A model that can express all the failure characteristics of the whole RAID storage system can be used to evaluate design choices, perform cost reliability trade-offs and conduct sensitivity analyses. However, including such details makes the computational models of reliability quickly infeasible. We present a CTMC reliability model for RAID storage systems that scales to much larger systems than heretofore reported and we try to model all the components as accurately as possible. We use several state-space reduction techniques at the user level, such as aggregating all in-series components and hierarchical decomposition, to reduce the size of our model. To automate computation of reliability, we use the PRISM model checker as a CTMC solver where appropriate. Our modelling techniques using PRISM are more practical (in both time and effort) compared to previously reported Monte-Carlo simulation techniques. Our model for RAID storage systems (that includes, for example, disks, expanders, enclosures) uses Weibull distributions for disks and, where appropriate, correlated failure modes for disks, while we use exponential distributions with independent failure modes for all other components. To use the CTMC solver, we approximate the Weibull distribution for a disk using sum of exponentials and we confirm that this model gives results that are in reasonably good agreement with those from the sequential Monte Carlo simulation methods for RAID disk subsystems reported in literature earlier. Using a combination of scalable techniques, we are able to model and compute reliability for fairly large configurations with upto 600 disks using this model.