WLM Sysplex Workload Distribution

When WLM is called by DNS or sysplex distributor for advice on where to route work in a sysplex, WLM returns a list of eligible servers with a weight assigned to each server. Work is then routed to the server proportionally to its weight so the higher weight servers receive more work. For example, if four servers A, B, C, and D are returned by WLM with weights of 2, 2, 4, and 8 respectively, DNS will route one-eight of the requests to each of servers A and B, one-fourth of the requests to server C, and one-half the requests to server D.

The weights are calculated as follows. WLM first assigns a weight to each system, with a higher weight to systems with the most available capacity. The weights are proportional to the amount of available capacity on each system. Then WLM divides the weight for a system equally among the registered servers on that system. In the case of DNS, there can be multiple servers registered with WLM per system image. For sysplex distributor, however, the WLM server registration is done by the TCP/IP stack so there is only one server per system image. This means the server weight always equals the system weight for sysplex distributor. If all systems are running at or near 100% utilization, then higher weights are given to the systems running less important work. The objective is to send work where there is available capacity or, if there is no available capacity, where the least important work will be displaced. Servers on systems that are in a serious storage shortage (SQA, fixed frame, auxiliary storage) are not recommended unless all systems are in a shortage.