14.6.2.1 Native vs. Virtual Performance

Figure 1 shows a comparison of measured bandwidth using virtual SCSI and local attached DASD for reads with varying block sizes of operations. The difference in the reads between virtual I/O and native I/O in these tests is attributable to the increased latency using virtual I/O. The difference in writes is caused by misalignment, which causes a read for every write. A write alignment change is planned for a future IBM i operating system release which will make virtual and native writes similar in speed.

Native vs. Virtual

 

100

MB/s

80

60

40

 

20

 

0

Small Transactions

(4k-16k)

Medium Transactions (32k-64k)

Large Transactions (128k+)

Native Read

Virtual Read

Native Write

Virtual Write

Figure 1 - The figure above shows a comparison of native vs virtual. This experiment shows that virtual write performance is significantly less then native. Read performance performs similar or better then native depending on the read-cache performance.

14.6.2.2 Virtual SCSI Bandwidth-Multiple Network Storage Spaces

Figure 2 shows a comparison of measured bandwidth while scaling network storage spaces with varying block sizes of operations. The difference in the scaling of these tests is attributable to the performance gain, which can be achieved by adding multiple network storage spaces. This experiment shows that in order to achieve better performance from the hard disk, multiple network storage spaces can be used.

15Disk 5GB Sequential Read

 

200

 

180

 

160

 

140

MBps

120

100

 

 

80

60

 

 

 

 

 

 

40

 

 

 

 

 

 

20

 

 

 

 

 

 

0

 

 

 

 

 

 

4

8

16

32

64

128

256

 

 

 

I/O Block Size

 

 

 

1NWSS

 

2 NWSS

 

4NWSS

 

8 NWSS

 

 

 

15 Disk 5GB Sequential Write

50

 

 

 

 

 

 

40

 

 

 

 

 

 

30

 

 

 

 

 

 

MBps20

 

 

 

 

 

 

10

 

 

 

 

 

 

0

 

 

 

 

 

 

4

8

16

32

64

128

256

 

 

I/O Block Size

 

 

1 NWSS

 

2 NWSS

 

4 NWSS

 

8 NWSS

Figure 2- The figures above show performance while scaling network storage spaces. This experiment shows that adding NWSS increases the throughput for read/write performance. The best performance is achieved by using 8 NWSS.

IBM i 6.1 Performance Capabilities Reference - January/April/October 2008

 

© Copyright IBM Corp. 2008

Chapter 14 DASD Performance

235

Page 235
Image 235
Intel AS/400 RISC Server manual Native vs. Virtual Performance, Virtual Scsi Bandwidth-Multiple Network Storage Spaces