Intel AS/400 RISC Server, 170 Servers, 7xx Servers manual Priority Mapping to Delay Cost Curves

of relatively lower delay cost.

yWaiting Time

The waiting time is used to determine the delay cost of a job at a particular time. The waiting time of a job which affects the cost is the time the job has been waiting on the TDQ for execution.

yDelay Cost Curves

The end-user interface for setting job priorities has not changed. However, internally the priority of a job is mapped to a set of delay cost curves (see "Priority Mapping to Delay Cost Curves" below). The delay cost curve is used to determine a job's delay cost based on how long it has been waiting on the TDQ. This delay cost is then used to dynamically adjust the job's priority, and as a result, possibly the position of the job in the TDQ.

On a lightly loaded system, the jobs' cost will basically stay at their initial point. The jobs will not climb the curve. As the workload is increased, the jobs will start to climb their curves, but will have little, if any, effect on dispatching. When the workload gets around 80-90% CPU utilization, some of the jobs on lower slope curves (lower priority), begin to overtake jobs on higher slope curves which have only been on the dispatcher for a short time. This is when the Dynamic Priority Scheduler begins to benefit as it prevents starvation of the lower priority jobs. When the CPU utilization is at a point of saturation, the lower priority jobs are climbing quite a way up the curve and interacting with other curves all the time. This is when the Dynamic Priority Scheduler works the best.

Note that when a job begins to execute, its cost is constant at the value it had when it began executing. This allows other jobs on the same curve to eventually catch-up and get a slice of the CPU. Once the job has executed, it "slides" down the curve it is on, to the start of the curve.

Priority Mapping to Delay Cost Curves

The mapping scheme divides the 99 'user' job priorities into 2 categories:

yUser priorities 0-9

This range of priorities is meant for critical jobs like system jobs. Jobs in this range will NOT be overtaken by user jobs of lower priorities. NOTE: You should generally not assign long-running, resource intensive jobs within this range of priorities.

yUser priorities 10-99

This range of priorities is meant for jobs that will execute in the system with dynamic priorities. In other words, the dispatching priorities of jobs in this range will change depending on waiting time in the TDQ if the QDYNPTYSCD system value is set to '1'.

yThe priorities in this range are divided into groups: y Priority 10-16

y Priority 17-22 y Priority 23-35 y Priority 36-46

IBM i 6.1 Performance Capabilities Reference - January/April/October 2008
© Copyright IBM Corp. 2008	Chapter 19 - Misc Perf Information	304