Resource availability determination mechanism for distributed data storage system

Abstract

A distributed data storage system is provided with a mechanism for determining data storage resource availability during data system failures. When a partition occurs, a coordinator requests all reachable replicas to begin a new epoch and each replica responds with its most recent epoch number and the list of replicas that it believes were active in that epoch. From the responses, the coordinator finds the maximum epoch number and the list of replicas in that epoch is taken as the list from which a quorum must be drawn. The coordinator establishes a quorum when, for each byte in the virtual store, the coordinator receives responses from more than half of the replicas for that byte. The coordinator then determines resource availability and sends out a message to all the reachable replicas with a new epoch number greater than the previous epoch number and the list of replicas active in this new epoch. The replicas receive this message, record the epoch number and the list of replicas, and proceed.

Dependencies are inferred from the claim text.