In a clustered installation, how is load balancing achieved?
As a business process is placed in one of the Sterling B2B Integrator queues prior to being executed, an evaluation is made as to whether it should be distributed or executed on the current node.
If the determination is to distribute it, the current node selects another node to execute the business process. The algorithm that makes the determination of whether to try to distribute the load and where to send it is controlled by a scheduling policy. The default policy “FairShareSchedulingPolicy” determines if there is any node that has 10 percent (or more) load than the other nodes. The load factor is an approximation of how long before the business process executes (a factor of the queue depth and the average time that business processes are executing before relinquishing their thread). Each node in an application cluster multicasts its current load to all other participants in the cluster every few seconds.