Fault Tolerant Methods for Intermitted Failures in Virtual Large Scale Disks

Recently, the demand of low cost large scale storages increases. We developed VLSD (Virtual Large Scale Disks) toolkit for constructing virtual disk based distributed storages, which aggregate free spaces of individual disks. However, current implementation of VLSD can mask only stop failure but cannot mask other kinds of failures such as intermitted failure. In this paper, we introduce two classes to VLSD in order to increase the intermitted fault tolerance. One is Retry Disk which retries to read/write at failures, another is VotedRAID1 which masks failures by majority voting. In this paper, we describe these classes in detail and evaluate their fault tolerance.