IMDG Inside Multiple Cloud Foundries

Most large organisations run multiple sites, on-premise or in the cloud, to provide the assurance of business continuity. As far as we're concerned, these organisations will also be running in-memory cache and data grids. And now these organisations are looking towards PAAS platforms for simplified infrastructure management and cost savings.

In this talk we'll look at doing all three at once, using multiple Hazelcast IMDGs hosted in multiple Pivotal Cloud Foundry (PCF) installations.

The core idea is simple. Each Hazelcast IMDG cluster runs as a group of processes within a Pivotal Cloud Foundry somewhere. Using Hazelcast's WAN feature cluster data can be shared with other IMDG instances in other Pivotal Cloud Foundries somewhere else. For a business, this provides the first benefit, user traffic can be split to send users to their geographically nearest instance so providing best performance.

Robustness comes from dynamic scaling. One IMDG cluster can be scaled up to double-size, allowing for an outage on the other. This gives the dual benefit of hitting SLAs for planned outages, without having to pay for more infrastructure than is needed the rest of the time. 

The final piece is of course returning to the original set-up. Reload the returning cluster from the one that had stayed running, then scale the latter back down to normal size.

So what we see in effect is cost-effective business continuity, using the data sharing features of Hazelcast IMDG combined with the application scaling control of Pivotal Cloud Foundry.