Databases Reference
In-Depth Information
Table 12-9: Four-Disk Test E Results (continued)
Read/Write
Sequential/Random
Block Size
IOPS
MB/Sec
Avg Latency
W
sequential
128
754.4
94.3
1
W
sequential
256
392.2
98.05
2
W
sequential
512
194.2
97.1
4
W
sequential
1024
100.6
100.6
9
W
sequential
2048
50.51
101.03
19
W
sequential
4096
25.4
101.6
38
W
sequential
8192
12.8
102.4
77
W
random
2048
39.5
79
24
W
random
4096
22.23
88.92
44
Starting from the top, you got a best result for Random Write of the 95.9MB/Sec with 95.9 IOPS for
1024KB blocks at 9 mSec latency. That's pretty close to the predicted results.
For Random Read you got 12.71MB/Sec with 101.7 IOPS at 128KB block size and 9 mSec latency. Again
this is exactly what you predicted.
For Sequential Read you got 110.8MB/Sec and 110.8 IOPS for 1024KB blocks at 8 mSec latency. That's
nowhere near the 200MB/Sec that was predicted.
For Sequential Write you got 100.6MB/Sec and 100.6 IOPS for 1024KB blocks at 9 mSec latency. Again
this is nowhere near the numbers you predicted, so now you need to go figure out why you aren't getting
what you expected.
Before looking for the bottleneck, do one final test run using a different stripe size to see if this has
any impact.
Testing Multiple Configurations
There are a lot of other configurations you could run tests against. For best results, test as many different
configurations as possible. Process the results, write them up, and store the results somewhere safe so
that you can refer back to them in the future when you are investigating a possible storage performance
issue or when you are considering a storage configuration change and need to compare the results for
the new configuration against the results from this set of tests.
InvestigatingStorageBottlenecks
It's beyond the scope of this topic to go into the details of how to troubleshoot storage bottlenecks.
However, it can show you how you may want to approach this particular problem. Read Chapter 6
Search WWH ::
Custom Search