Intel 7xx Servers, 170 Servers Performance Testing Results, Qdynptyscd = ‘1’ on Qdynptyscd = ‘0’

yPriority 47-51

yPriority 52-89

yPriority 90-99

Jobs in the same group will have the same resource (CPU seconds and Disk I/O requests) usage limits. Internally, each group will be associated with one set of delay cost curves. This would give some preferential treatment to jobs of higher user priorities at low system utilization.

With this mapping scheme, and using the default priorities of 20 for interactive jobs and 50 for batch jobs, users will generally see that the relative performance for interactive jobs will be better than that of batch jobs, without CPU starvation.

Performance Testing Results

Following are the detailed results of two specific measurements to show the effects of the Dynamic Priority Scheduler:

In Table 19.1, the environment consists of the RAMP-C interactive workload running at approximately 70% CPU utilization with 120 workstations and a CPU intensive interactive job running at priority 20.

In Table 19.2 below, the environment consists of the RAMP-C interactive workload running at approximately 70% CPU utilization with 120 workstations and a CPU intensive batch job running at priority 50.

Table 19.1. Effect of Dynamic Priority Scheduling: Interactive Only

	QDYNPTYSCD = ‘1’ (ON)	QDYNPTYSCD = ‘0’
Total CPU Utilization	93.9%	97.8%
Interactive CPU Utilization	77.6%	82.2%
RAMP-C Transactions per Hour	60845	56951
RAMP-C Average Response Time	0.32	0.75
Priority 20 CPU Intensive Job CPU	21.9%	28.9%
Table 19.2. Effect of Dynamic Priority Scheduling: Interactive and Batch
	QDYNPTYSCD = ‘1’ (ON)	QDYNPTYSCD = ‘0’
Total CPU Utilization	89.7%	90.0%
Interactive CPU Utilization	56.3%	57.2%
RAMP-C Transactions per Hour	61083	61692
RAMP-C Average Response Time	0.30	0.21
Batch Priority 50 Job CPU	15.0%	14.5%
Batch Priority 50 Job Run Time	01:06:52	01:07:40

Conclusions/Recommendations

yWhen you have many jobs running on the system and want to ensure that no one CPU intensive job 'takes over' (see Table 19.1 above), Dynamic Priority Scheduling will give you the desired result. In this case, the RAMP-C jobs have higher transaction rates and faster response times, and the priority 20 CPU intensive job consumes less CPU.

yDynamic Priority Scheduling will ensure your batch jobs get some of the CPU resources without significantly impacting your interactive jobs (see Table 96). In this case, the RAMP-C workload gets

IBM i 6.1 Performance Capabilities Reference - January/April/October 2008
© Copyright IBM Corp. 2008	Chapter 19 - Misc Perf Information	305