Resources
This section discusses WLM resources in more detail.
Processor
Hard limits on processors ensure that applications do not monopolize resources. Although allocating applications to specific processor resource sets is also effective, this approach is somewhat inflexible and may inhibit performance for multi-threading applications like WebSphere.
Having specified limits for maximal workloads and shares that reflect each application's expected workload provides fair performance at lower levels, assuming constant incoming workloads.
Memory
Paging WebSphere Application Server JVMs significantly damages performance. You can reduce this by specifying minimum memory percentages or memory resource sets.
You must ensure the allocation allows non-WebSphere Application Server applications sufficient memory. If memory is overcommitted, unmanaged applications or even the system bear the brunt of paging, which can trigger thrashing.
Disk
WebSphere applications rarely monopolize disk I/O bandwidth. Most data access uses enterprise services such as databases and J2C connectors, which generate network I/O. AIX WLM does not control network bandwidth.
If one application monopolizes I/O bandwidth (for example, excessive tracing to System.out), then I/O shares or limits might prove useful, but their effect is to make it look like a slow I/O subsystem.