Deployment recommendations

Learn about the recommended ways to deploy SRM in your system.

SRM can be deployed in two different ways:
  • Co-located with existing Kafka brokers.
  • On dedicated machines outside of existing Kafka brokers.

Co-located deployment

Use a co-located deployment approach when your existing Kafka Brokers are not heavily utilized. For this deployment approach, configure each instance of SRM to only write to the cluster that it is running on. This avoids scenarios in which SRM clients deployed on Cluster 1 would write to brokers in Cluster 2. Depending on the available network bandwidth between the two clusters, this approach can increase the replication performance.
Figure 1. Co-located deployment

Dedicated deployment

SRM mainly consumes memory and network bandwidth on the hosts it is installed on. If your existing Kafka Brokers do not have a lot of headroom left for these resources, consider deploying SRM on dedicated machines or expanding your Kafka deployments.

If you have data centers with multiple Kafka clusters, a dedicated SRM deployment in each data center can help to simplify management of the SRM nodes. In this scenario, configure each instance SRM to write to all Kafka clusters running in the local data center.

Figure 2. Dedicated deployment