If a workload is running fewer requests than the workload throttle allows, that workload is underused. With individual workload throttles, no other workload can benefit from this underused capacity. However, workload group throttles allow an underused workload and a busy workload running the same application to share capacity. In this situation, TASM divides requests between the workloads, and each workload has a workload throttle. The following examples describe the effect of an underused workload with different throttle types.
No Capacity Sharing with Workload Throttles
With individual workload throttles, if one workload is underutilized, no other workload can benefit from the unused capacity, even if both workloads run the same application. In this example, all requests from an application classify to one of two different workloads. Both workloads have throttles. If one of the workloads is not running all the requests it can, the entire application does not reach potential.
| Workload |
Request Count at Throttle Limit |
Request Count If Only 3 Medium Requests Are Active |
|---|
| Medium |
6 |
3 |
| Long |
3 |
3 |
| Combined |
9 |
6 |
Capacity Sharing with Workload Group Throttles
Group throttles allow a combination of throttled workloads to be treated as a group by a single, higher-level throttle. In this example, the application has a group throttle limit of 9 requests. To allow capacity sharing between the two workloads, each workload has a slightly larger throttle limit than the intended level. For example, Medium has a limit of 8 requests instead of 6. If one workload is not busy, the other workload can start more requests, up to the group throttle limit or the workload throttle limit.
| Workload |
Request Count at Throttle Limit |
Request Count If Only 3 Medium Requests Are Active |
|---|
| Medium |
8 |
3 |
| Long |
6 |
6 |
| Application Workload Group Throttle |
9 |
9 |
| Combined |
9 |
9 |