Cluster Sizing Estimates for Hazelcast Jet

| 7 pages

Hazelcast Jet is an event stream processing engine built into the Hazelcast In-Memory Computing Platform, which also features an in-memory data grid (Hazelcast IMDG) to allow extremely fast processing for both batch and streaming data sources. The platform is typically deployed across a series of hardware nodes in a cluster, either in an on-premises or cloud environment. This scale-out architecture lets you incrementally and efficiently add more nodes to handle more capacity as your load grows.

One exercise as part of deploying Jet is to determine a good estimate of the number of computing resources you need to optimally run your Jet application(s). Since there are many different use cases that can be handled by Jet, there is no one-size-fits-all strategy for estimating cluster sizing. Sizing can be complicated and uncertain, but providing the characteristics of an example deployment is a good place to start.

This guide will walk you through specific environments as examples which can then be extrapolated for your own specific workloads. These examples are based on deployments set up by Hazelcast, with workloads using published source code, which should give you more information on how you can compare your actual workloads.