High availability and workload management
High availability and workload management can be achieved using clusters as bus members. Note that messaging engines do not follow the same clustering model that J2EE applications do in clusters.
When you add a cluster to a bus, a single messaging engine is created. The messaging engine is active on only one server within the cluster. In the event of an appserver or messaging engine failure, the messaging engine becomes active on another server in the cluster if one is available.
By default, the messaging engine starts on the first available server in a cluster. If you want to ensure that the messaging engine runs on a particular server, for example, if you have one primary server and one backup server, or if you want the messaging engine to only run on a small group of servers within the cluster, then specifically configure this.
Cluster bus members for workload management
Because a single messaging engine for the cluster is active, there is no workload management by default. To achieve greater throughput of messages, it is beneficial to spread messaging load across multiple servers and, optionally, across multiple hosts. You can achieve this, while maintaining a simple destination model, by creating additional messaging engines for the cluster, each of which has a preference to run on a separate server in the cluster.
You can configure these messaging engines with a preference to run on particular servers within the cluster. This enables a messaging engine to run in every server in the cluster, thus providing every application in the cluster with a messaging engine for local access to the bus. Local access to the bus is always better for messaging performance, especially in the case of queues where the queue is assigned to the bus member from which it is being accessed.
When a queue is assigned to a cluster bus member, the queue will be partitioned across all messaging engines in the cluster.
See also...
- WAS V6 Scalability and Performance Handbook, SG24-6392.
- WAS V6: High Availability Solutions, REDP-3971