Intel 7xx Servers, 170 Servers manual CPU milliseconds to process virtual Scsi I/O transaction

To calculate IBM i operating system Virtual SCSI CPU requirements the following formula is provided. The number of transactions per second could be collected by the IBM i operating system command WRKDSKSTS. Based on the average transaction size in WRKDSKSTS, select a number from the table.

Size of IO

		4	8	16	32	64	128	256
Type of Transaction	Read	16	22	34	57	92	163	314
	Write	21	26	36	54	82	148	282

Figure 5- CPU milliseconds to process virtual SCSI I/O transaction

The table above shows the time in milliseconds per transaction that Virtual SCSI takes to process one transaction. This value can be used in the formula below to estimate the amount of CPU required per a partition.

(# of Transactions per second * Time in Milliseconds per transaction) = CPU Utilization

1,000,000

For example.. If your workload performed 10,000 16k read transactions the equation would look like this (34 was selected from the table above):

(10,000 * 34) = .34(34% of a total CPU)

1,000,000

The total CPU required for a workload, which performs 10,000 16k read transactions per second, would be 34% of a 2.2Ghz POWER5 processor. If a different size processor is used adjust these numbers accordingly. Remember the number chosen in WRKDSKSTS is an average of all I/O’s. Your workload could be a mixture of very large transactions and very small transactions. This is to provide a guideline of how to size your CPU correctly, and your results might vary.

Using Dedicated processor partitions may require more CPU then necessary that could be used by other partitions, but will guarantee peak performance. It is most effective if the average I/O size can be estimated so that peak bandwidth does not have to be assumed. Most Virtual SCSI servers will not run at maximum I/O rates all the time, so the use of surplus processor time is potentially wasted by using dedicated processor partitions.

IBM i 6.1 Performance Capabilities Reference - January/April/October 2008
© Copyright IBM Corp. 2008	Chapter 14 DASD Performance	239