Reliability and performance of RAIDs

摘要

We define and analyze the ACATS declustering scheme for use with the classic Level 5 as well as MDS RAIDs. ACATS provides a paradigm for distributed sparing and checking as well. Using Markov models we calculate the reliability of classic Level 5 as well as MDS RAIDs both with and without ACATS. Our results demonstrate, for fixed hardware configurations, trade-offs between performance and reliability within these RAID organizations. Our analysis assumes a transaction oriented environment.