The biasing algorithm

The biasing algorithm provides a way of balancing workload by specifying that workload distribution should favor particular server groups. For example, if there are two server groups with a bias of 75 and 25, program requests are sent in a ratio of 3:1 to the first server group.

If a server group fails, the internal biasing calculation changes. If two server groups are available, one with a bias of 100 and the other with a bias of 0, all requests are sent to the first server group. If the first server group becomes unavailable, all requests are directed to the second server group. A bias value of 0 is a special case, meaning use only if no other server group is available.

The biasing algorithm works only for ECI calls. If you try to run an EPI application whilst the biasing algorithm is selected, the round-robin algorithm is used instead.

Examples

For each of these examples, assume that workload management is set up with the example configuration described in Information required by Workload Manager, and that workload management is enabled for all programs and server groups.
  • For PROG1, if GROUP1 has a bias of 6, GROUP2 has a bias of 2 and GROUP3 has a bias of 2, approximately 60% of requests for PROG1 go to GROUP1, and approximately 20% to each of GROUP2 and GROUP3. The selection of servers within each group is random, but the Client daemon will only ever use one server from each group.
  • For PROG2, if GROUP2 has a bias of zero, and GROUP3 has a bias of 1, all requests are sent to GROUP3 (and therefore CICS6). If CICS6 becomes unavailable, all requests are sent to GROUP2.

Information Information

Feedback


Timestamp icon Last updated: Tuesday, 19 November 2013


https://ut-ilnx-r4.hursley.ibm.com/tg_latest/help/topic/com.ibm.cics.tg.doc//ctgunx/bias.html