Not exactly groundbreaking information, but were you aware that you can run SQLIO with multiple files? Until a couple months ago, I hadn’t really paid attention to that little fact. And maybe I’m the only person that didn’t know, but figured I would share it anyways.
How Multiple Files?
Using SQLIO with multiple files is fairly simple. After you write the SQLIO command-line arguments with the single file, add a space and type the name of the second file. Of course, this is actually documented in the help output for SQLIO, shown in the image below.
In practice, your SQLIO command-line arguments would look like those in Listing 1. In this statement, SQLIO will use the files D:array01testfile.dat and E:array02testfile.dat for testing throughput.
#Listing 1. SQLIO Command-Line Arguement sqlio.exe -kR -t10 -o10 -frandom -b512 -BH -LS D:array01testfile.dat E:array02testfile.dat
Why Multiple Files?
Now… why would someone be interested in running a throughput test against multiple files? There are probably other reasons than the one that I’ve got, but I was doing it to test throughput on multiple flash arrays. I had a server connected to two arrays and wanted to see the combined throughput without monkeying around and running multiple instances of SQLIO. And there isn’t much monkeying around when it comes to a space delimited list of files.
Hopefully this post enlightens someone to a tidbit about SQLIO that they didn’t know. If you are testing throughput across multiple arrays, what do you do to test the throughput? Multiple files or multiple SQLIO instances?
7 thoughts on “Running SQLIO with Multiple Files”
The sqlserverpedia example scripts are good and your tip would another dimension. There is a post here http://paul.dynalias.com/SQL/_layouts/15/start.aspx#/Lists/Posts/Post.aspx?ID=9 about analysing SQLIO results in Analysis Services data mining models. Might have been more interesting adding multiple files to the test scripts.
Interesting outputs, I may have to play with those. Thanks for sharing.
Multiple files listed in a param file. Good for testing fast track data warehouse configurations.
Oh yeah, that would work too. I’ve done that for creating test files but not for running the tests. And FTDW might be exactly what I’m using it for as well.
That brings up another topic 🙂 I’m in the process of setting a POC for a ftdw 30TB solution. I’d be interested in seeing your results. Stage and prod databases are in place and the 2tbs of test data was generated using dbgen. Got a few bugs in my SSIS package to work through. Thanks!
Comments are closed.