How Fast and Robust Is Your Storage? - Professional SQL Server 2005 Performance Tuning - page 447

Table 12-7: Two-Disk Test E Results (continued)

Read/Write Sequential/Random Block Size IOPS MB/Sec Avg Latency

R sequential 2048 51.6 103.2 18

R sequential 4096 26 104 37

R sequential 8192 13.05 104.46 76

W sequential 8 6898.2 53.89 0

W sequential 16 4398.6 68.72 0

W sequential 32 2722.7 85.08 0

W sequential 64 1494.1 93.38 0

W sequential 128 771.9 96.48 1

W sequential 256 394.4 98.6 2

W sequential 512 202.1 101.05 4

W sequential 1024 99.6 99.6 9

W sequential 2048 50.61 101.23 19

W sequential 4096 25.45 101.83 38

W sequential 8192 12.76 102.08 77

Let's do the same process of finding the max throughput for each I/O type again. First off let's look for
the maximum throughput. This time you got 102.3MB/Sec and 102.3 IOPS for 1024KB Sequential Reads
at 9mSec latency. In comparison Sequential Write managed 99.6MB/Sec and 99.6 IOPS again at 1024KB
and 9mSec latency.

Random access performance was proportionally similar to the 1 disk results in that the best random
write results were at 51.35MB/Sec and 102.7 IOPS for 512KB blocks and 9 mSec latency. Random Read
performance peaked at 12.6MB/Sec and 100.8 IOPS for 128KB blocks and 9 mSec latency.

4DiskTests32KBStripe
The next step was to reconfigure the array into a 4-disk stripe set. The smallest stripe size of 32KB was
chosen again.

Before you look at the results, let's take a quick guess at what you might think you should see this time.
Let's go back and summarize the changes between the 1 disk and 2 disk results, and from there extrapo-
late what you should get for the 4 disk tests.