Take the 2-minute tour ×
Server Fault is a question and answer site for professional system and network administrators. It's 100% free, no registration required.

I did a quick SAN Performance test over the weekend with SQLIO (not SQLIOSim).

The first two tests involved writing random 64k blocks to 1 GB files on H:\ and K:\

Here are the results from SQLIO:

sqlio v1.5.SG 
using system counter for latency timings, -1794967296 counts per second 
1 thread writing for 60 secs to file H:\sqlio_test.dat
using 64KB random IOs
enabling multiple I/Os per thread with 1 outstanding
buffering set to not use file nor disk caches (as is SQL Server) 
using current size: 1000 MB for file: H:\sqlio_test.dat
initialization done
CUMULATIVE DATA:
throughput metrics:
IOs/sec:   220.51
MBs/sec:    13.78
latency metrics: Min_Latency(ms): 2 Avg_Latency(ms): 4 Max_Latency(ms): 30
histogram: 
ms: 0  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24+ 
%:  0  0  6  0 84  8  1  0  0  0  0  0  0  0  0  0  0  0  0  0  0  0  0  0  0

sqlio v1.5.SG
using system counter for latency timings, -1794967296 counts per second
1 thread writing for 60 secs to file K:\sqlio_test.dat
using 64KB random IOs
enabling multiple I/Os per thread with 1 outstanding
buffering set to not use file nor disk caches (as is SQL Server)
using current size: 1000 MB for file: K:\sqlio_test.dat
initialization done
CUMULATIVE DATA:
throughput metrics:
IOs/sec:   331.81
MBs/sec:    20.73
latency metrics: Min_Latency(ms): 2 Avg_Latency(ms): 2 Max_Latency(ms): 107
histogram: 
ms: 0  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24+ 
%:  0  0 59 40  1  0  0  0  0  0  0  0  0  0  0  0  0  0  0  0  0  0  0  0  0

So far, so good. But if I look at the performance counters for these drives (both of which are mirrored "disks" on a Hitachi OpenV SAN), I get 39 MB/s on H: and 55 MB/s on K:.

I'm absolutely sure that nothing else but the SQLIO test happened on these drives at that time. All the other SQLIO tests show a similar pattern. I/O from the PerfCounters is about 2.5 times higher than the SQLIO result.

But which one is right?

Edit: OS is Win 2003 R2

share|improve this question

2 Answers 2

Try iozone, see which one it's closest to.

share|improve this answer
up vote 0 down vote accepted

Stupid me!

As I said, the disks are mirrored. So from SQLIOs point of view its just one disk, but if I add up the perfcounters, I get the writes on both disks!

The rest of the remaining 1.3 factor was just a misinterpretation of the perfcounter chart. I looked at the maximum and didn't take the beginning and end slope into account which reduces the average. These slopes were quite steep so I had to zoom in to maximum detail. Would have been better, if the test ran more than one minute (perfcounter sample size was 10 seconds).

So, both are right and my faith is restored :-)

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.