Tiers, shares, and limits in Workload Manager
Using the data gathered from running WLM in passive mode and your business goals, decide which tier number will be given to every superclass and what share of each resource should be given to the various classes.
For some classes, you might want to define minimum or maximum limits. Adjust the shares and tier numbers to reach your resource-allocation goals. Reserve limits for cases that cannot be solved only with shares. Also, you might decide you need to add subclasses.
- Use minimum limits for applications that typically have low resource usage but need a quick response time when activated by an external event. One of the problems faced by interactive jobs in situations where memory becomes tight is that their pages get stolen during the periods of inactivity. A memory minimum limit can be used to protect some of the pages of interactive jobs if the class is in tier 0.
- Use maximum limits to contain some resource-intensive, low-priority jobs. Unless you partition your system resources for other reasons, a hard maximum will make sense mostly for a nonrenewable resource such as memory. This is because of the time it takes to write data out to the paging space if a higher priority class needs pages that the first class would have used. For CPU usage, you can use tiers or soft maximum to make sure that if a higher priority class is immediately assigned CPU time.
When creating and adjusting the parameters of subclasses, you can refresh WLM only for the subclasses of a given superclass that do not affect users and applications in the other super classes, until you are satisfied with the system behavior.
You can also define other configurations with different parameters, according to the needs of the business. When doing so, you can save time by copying and modifying existing configurations.