DYNAMICALLY QUANTIFYING AND IMPROVING THE RELIABILITY OF DISTRIBUTED DATA STORAGE SYSTEMS

摘要

Data is stored in a distributed data storage system comprising a plurality of disks. When a disk fails, system reliability is restored by executing a set of reconstructions according to a schedule. System reliability is characterized by a dynamic Normalcy Deviation Score. The schedule for executing the set of reconstructions is determined by a minimum intersection policy. A set of reconstructions is received and divided into a set of queues rank-ordered by redundancy level ranging from a lowest redundancy level to a highest redundancy level. For reconstructions in each queue, an intersection matrix is calculated. Diskscores for each disk are calculated. The schedule for the set of reconstructions is based at least in part on the intersection matrices, the Normal Deviation Scores, and the diskscores.