Broker 10.5 | webMethods Broker Documentation | webMethods Broker Messaging Programmer's Guide | JMS Policy Based Client-Side Clustering | Cluster Load Balancing Policies | How Does the Round Robin Cluster Policy Work?
How Does the Round Robin Cluster Policy Work?
The following figure gives an example of how round robin load balancing policy will work with a cluster of four Brokers.
Round Robin Selection Logic Used in Load Balancing
For load balancing, published messages are directed sequentially, that is the first message is sent to the first Broker and the next message to the second Broker. After the last Broker in cluster receives the messages, next message is sent to the first Broker, and the sequence repeats. One message at a time is sent to each Broker in the load-balancing cluster.
If a Broker in the cluster goes offline (indicated by the cross for Broker B), when it is the turn of Broker B to receive a message, the message is re-directed to the next Broker in the cluster, that is Broker C. Broker B will be skipped for load distribution until it is available again.
In an XA transaction, all the published messages are routed to a single Broker determined by the cluster policy in the cluster. The cluster policy applies to a transaction instead of a message.