Intel AS/400 RISC Server, 170 Servers, 7xx Servers manual Sizing memory, Disk 1GB Sequential Read

14.6.3.3 Sizing memory

The IBM i operating system Virtual SCSI server supports data read caching on the virtual hosted disk server partition. Thus all I/Os that it services could benefit from effects of caching heavily used data. Read performance can vary depending upon the amount of memory which is assigned to the server partition. Workloads which have a small memory footprint can improve their performance greatly by increasing the amount of memory in the IBM i operating system Virtual SCSI server. Alternatively, a system which works on a large amount of data may not see any benefit from caching. The memory for the IBM i operating system Virtual SCSI server in this case can be set at less then 1 GB.

One method to size this is to begin by looking at your ASP in which your network storage space is located. While the system is running the desired workload, type in the command WRKDSKSTS. Write down the average number of I/O request per second in the ASP which is being used by the network storage space. Now dynamically add memory to the partition. Check the number of I/O requests per second once again (remember to reset the statistics using F10). The number of I/O requests per second should lower and your throughput to the IBM i operating system Virtual SCSI server should increase.

Continue adding memory to the IBM i operating system server until you no longer see the number of I/O requests per second change. If your workload changes at a later date the memory can be readjusted accordingly.

Figure 6 below shows a comparison of measured bandwidth of cached transactions with varying block sizes of operations. The figure includes small (4k-64k) transactions and larger(128k) transactions. A partition which runs completely from memory can experience throughput rates as high as 6GB/sec. If it is memory constrained the systems throughput will be lower.

15 Disk 1GB Sequential Read

	6
	5
Throughput	4
	3
	2
	1
	0
	4	8	16	32	64	128	256	512	1024	2048
					I/ O B lock Size

0% 25% 50% 75% 100%

Figure 6 - The figure above shows a comparison of small, medium, and large transactions affect on memory if cached. The lines represent the amount of data, which is cached in memory. The efficiency of I/O improves with cache hits and larger I/O size. Effectively, there is a fixed latency to start and complete an I/O, with some additional cycle time based on the size of the I/O.

IBM i 6.1 Performance Capabilities Reference - January/April/October 2008
© Copyright IBM Corp. 2008	Chapter 14 DASD Performance	241