Scaling Data Services with Pivotal GemFire

Chapter 6. Multisite Topologies Using the WAN Gateway

Mike Stolz

You can deploy GemFire clusters in a single site or in a cloud deployment across nearby availability zones. But often the need arises for having two separate clusters in distinct locations. You can meet this need by replicating data between distant sites using GemFire’s Asynchronous WAN Gateway, a bidirectional, fault-tolerant, store-and-forward replication subsystem. It is asynchronous because it is intended for use in long-haul replication situations, for example, New York, London, Tokyo. You can build mesh topologies wherein each site replicates data that originates there to all the other sites, or you can limit the locations to which data is replicated.

Example Use Cases for Multisite

The simplest and most often used design pattern for multisite deployment of two clusters is active/passive disaster recovery (DR). DR is still a very popular design pattern, and if you have several business units that update different data sets it is very easy to make this pattern active/active. Let’s say you have an equity trading business and a fixed-income trading business. They trade different instruments, so they will never collide with each other. This means that you could host the equity business in one datacenter and the fixed-income business in the other datacenter, with both active, and both backing each other up for DR purposes, as illustrated in Figure 6-1.

Figure 6-1. A simple bidirectional active/passive WAN Gateway ...