Abstract

Many current online services are deployed over geographically distributed sites (i.e., datacenters). Such services call for geo-replicated storage, that is, storage distributed and replicated among many sites. Geographical distribution and replication can improve locality and availability of a service. Locality is achieved by moving data closer to the users. High availability is attained by replicating data in multiple servers and sites. In this talk I will consider a class of highly available storage systems based on deferred update replication. Deferred update replication ensures strong consistency and is at the core of several highly available storage systems. I will first show how the performance of deferred update replication can be made to scale to dozens and potentially hundreds of nodes. I will then consider the implications of the approach on user-perceived latency in geographically distributed environments and propose solutions. The talk will conclude with some open problems in the design of strongly consistent storage systems.