Tuning the Performance of a Clustered Deployment
- Former user (Deleted)
This section describes how to tune the performance of the Message Broker profile in terms of deployment in a clustered environment.
| Improvement area | Parameter | Description | Location | Default/Recommended Value |
|---|---|---|---|---|
| Memory | -Xms256m -Xmx1024m -XX:MaxPermSize=256m | The memory allocated for the broker. Â Â Â | For Windows: For Linux: | Generally, at least 2 GB memory is recommended for production instances. |
| Cluster health | hazelcast.max.no.heartbeat.seconds | The maximum time period that should elapse between pings received from a worker node in a broker cluster before acknowledging that worker node to be dead. This value is specified in seconds. This parameter prevents the allocation of resources to inactive worker nodes, thereby avoiding unnecessary system overheads. A short time duration can be specified in environments with a very high message flow and it is important to reallocate resources from inactive nodes to active nodes fast. A longer time duration can be specified in environments with lower message flows. | <EI_HOME>/wso2/broker/conf/hazlecast.properties file | 600 |
| Cluster Recovery | concurrentStorageQueueReads | The number of storage queue reads carried out concurrently at a given time. This number should be set based on the number of storage queues that currently exist. | <EI_HOME>/wso2/broker/conf/broker.xml file | 5 |