Posted by: Colin Smith
Disk Performance, IO, SAN, SQL Server
Just the other day I was asked to prove to a user that our SAN disk would be faster than the servers Local Internal Disk. Now I am no SAN Admin but I know that in most cases the Fiber attached SAN is going to be faster than the internal local 10,000 RPM drives in my servers. Well thanks to Brent Ozar, once again, I was able to use SQLIO to run some tests and prove that the Fiber Attached SAN outperforms the local disks, at least in my case.
I watched and read the article that I have linked above and I ran SQLIO on both the internal and the SAN disks. I used the same parameters for each test. Both ran for about 18 hours since I did what is suggested and tested with multiple parameter values.
After the tests were complete I had about 90 pages of output for each test and was asked to come up with a summary of results to present by the next day. I was a bit worried about this. In Brent’s post he talks about pulling the Data into SQL Server and I am sure that is a great way to handle it. I do not have a server set up at this time to put this data into and, I am sorry to say, I am not comfortable enough with writitng SQL to do this and slice and dice the data in a short amount of time and I was up against a deadline.
I decided to use Powershell to sort the data for me and pull out only what I needed. I had Powershell put this into a csv file so that I could then manipulate the raw data in Excel. After I had it in Excel I was able to make some conclusions, Reads on the SAN were almost 80% faster and the Writes were almost 30% faster. Now SQLIO does not simulate SQL Server behavior, but this proves to me that the SAN is faster at the most basic level. If you would like to simulate SQL Server than you can use SQLSIM. I have looked at it but I have not dug into it much so I can not really say much about it at this time. Just something else to learn.