WLM stateful session placement adds support to WLM queue/server management. WLM supports round-robin scheduling of queue manager requests across active server regions.

WebSphere informs WLM about stateful work requests through statefull session EJBs and servlets with HTTPSession objects. It establishes a “temporal affinity” to these.

The first server region that is started tends to be overloaded when temporal affinities exist and the load of the other servers decreases. Thus server regions with low or zero utilization exist. This is because, when new work requests arrive, WLM selects the first server that it finds with inactive server tasks. The rationale behind this is that there is faster selection for work requests that have no temporal affinities.

WebSphere now allows you to specify the selection algorithms for static or dynamic application environments, as follows:

• Option 1: First available server (default):

  • Selects the first server region with inactive server tasks
  • Work tends to be clustered on fewer servers and these are more effectively managed
     
• Option 2: Round robin:

  • Selects server by lowest number of affinities
  • In case of a tie, selects server by lowest number of active server tasks
  • Work is spread across all servers

This is exploited with WebSphere V5.