August 16, 2012 at 1:32 pm
Need to test/stress a new SQL Server box to see what she can do. I went to get a copy of SQLIO which is referenced all over Microsofts own website but the actual download is no-where to be found nor can I find any comments or postiing about what happened to SQL IO.
1) What tool(s) would you recomend to test a new SQL Box to see what kind of perfromance once can expect with a heavy load running against it; something that can simulate several hundread users submitting a variety of querys?
2) ANyone know the storyt behind what happened to SQL IO?
Thanks
Kindest Regards,
Just say No to Facebook!August 17, 2012 at 4:51 am
That's interesting. I didn't know SQLIO was gone. Hmmm... Maybe they just renamed it. It never should have been called SQLIO in the first place, just IO. It never tested databases or SQL Server, but primarily the disk sub-system. Measuring the machine is useful to validate what the vendor told you, but not really necessary for establishing a baseline or doing load testing.
As far as load testing, my latest & greatest favorite toy is the Distributed Replay tools in SQL Server 2012. They do a great job either taking an existing trace and replaying it or creating an artificial load and then smacking the server around with it. That's what I use to test things out today.
"The credit belongs to the man who is actually in the arena, whose face is marred by dust and sweat and blood"
- Theodore Roosevelt
Author of:
SQL Server Execution Plans
SQL Server Query Performance Tuning
August 20, 2012 at 10:53 am
Grant Fritchey (8/17/2012)
That's interesting. I didn't know SQLIO was gone. Hmmm... Maybe they just renamed it. It never should have been called SQLIO in the first place, just IO. It never tested databases or SQL Server, but primarily the disk sub-system. Measuring the machine is useful to validate what the vendor told you, but not really necessary for establishing a baseline or doing load testing.As far as load testing, my latest & greatest favorite toy is the Distributed Replay tools in SQL Server 2012. They do a great job either taking an existing trace and replaying it or creating an artificial load and then smacking the server around with it. That's what I use to test things out today.
Thanks for replying Grant. I have to say that I'm shocked that no one else has replied. Considering the question I was certain a number of regulars would have posted a reply about SQLIO having disappeared from Microsofts web site if noting else. There are numeorus postings on the wbe and here at SSC that all reference SQL I/O so its usprising to see no one (other then you) is suprised by SQL I/O's disapearence.
Do you know if the tools in SQL 2012 you mention will work on a SQL Server 2005 instance?
BTW - We are using SQL 2005 because the vendor whose software we use does not yet support a higher version of SQL Server then 2005. They are moving to support of 2008R2 and we will be upgrading to that in a few months but till that happens we have to continue onwards using SQL 2005.
Thanks again for replying.
Kindest Regards,
Just say No to Facebook!August 20, 2012 at 11:20 am
Yes, you can point the distributed playback mechanism to 2005. It just needs a database to connect to. But, the install is 2012, so you have to have that first.
"The credit belongs to the man who is actually in the arena, whose face is marred by dust and sweat and blood"
- Theodore Roosevelt
Author of:
SQL Server Execution Plans
SQL Server Query Performance Tuning
August 20, 2012 at 12:05 pm
Grant Fritchey (8/20/2012)
Yes, you can point the distributed playback mechanism to 2005. It just needs a database to connect to. But, the install is 2012, so you have to have that first.
We do have SQL Server 2012 I just dont have a Server setup to use it yet since none of our SQL Server DB's are for sofwtare that works with 2012 (yet). I found what I believe is the FAQ/REQS page on this tool at MSDN (http://msdn.microsoft.com/en-us/library/ff878239.aspx ). It reads like as if its saying that you can use a SQL Server 2005 DB as your source (what you are tracing ) but that to play it back you have to be on 2008R2 or 2012, is that the way you read it?
Thanks
Kindest Regards,
Just say No to Facebook!August 20, 2012 at 12:31 pm
Actually, yeah, looks like 2005 might be out. That stinks. Never noticed that before.
"The credit belongs to the man who is actually in the arena, whose face is marred by dust and sweat and blood"
- Theodore Roosevelt
Author of:
SQL Server Execution Plans
SQL Server Query Performance Tuning
August 20, 2012 at 1:20 pm
Grant Fritchey (8/20/2012)
Actually, yeah, looks like 2005 might be out. That stinks. Never noticed that before.
Any suggestions on what to do? I need to do some comparison testing aginst the new server to see how it fares against the current DB server, to see if the same workload occurring now will run better on the new system. I was going to run a non-gui trace (server side trace) and then play it back on the new server but I'm thinking here has to be a better way to do this. I'd prefer not to pays lots of moeny to do this but if nothing else I woudl think one of the SQL sofwtare Toools vendors woudl have come up with something like this, that lets you see how the same workload on one server will do on another. Make sense?
BTW - I have and did run SQL Imulator but thats it so far.
Thanks
Kindest Regards,
Just say No to Facebook!August 20, 2012 at 2:07 pm
YSLGuru (8/20/2012)
Grant Fritchey (8/20/2012)
Actually, yeah, looks like 2005 might be out. That stinks. Never noticed that before.Any suggestions on what to do? I need to do some comparison testing aginst the new server to see how it fares against the current DB server, to see if the same workload occurring now will run better on the new system. I was going to run a non-gui trace (server side trace) and then play it back on the new server but I'm thinking here has to be a better way to do this. I'd prefer not to pays lots of moeny to do this but if nothing else I woudl think one of the SQL sofwtare Toools vendors woudl have come up with something like this, that lets you see how the same workload on one server will do on another. Make sense?
BTW - I have and did run SQL Imulator but thats it so far.
Thanks
First, I think you server side trace ides is better because it's going to behave like the actual database engine.
That being said there are tools that will just thrash the disks if that's what you need. As Grant already mentioned, SQLIO basically did just that (a bunch of reads and writes with files of a size you specified).
Take a look at PassMark's BurnInTest suite. It has disk tests built in. The standard and pro editions are both pretty reasonably priced.
August 21, 2012 at 5:04 am
You can use the old Profiler to playback a trace. That will act as a measurable set of tests and will put some load on the system. You just can't scale that load up the way you can with Distributed Replay.
Loath though I may be to suggest other vendors, but Quest used to have a load testing software that I used to use to test out apps. I'm not sure they're still selling it or not. You could check there.
"The credit belongs to the man who is actually in the arena, whose face is marred by dust and sweat and blood"
- Theodore Roosevelt
Author of:
SQL Server Execution Plans
SQL Server Query Performance Tuning
August 21, 2012 at 4:30 pm
Scott D. Jacobson (8/20/2012)
YSLGuru (8/20/2012)
Grant Fritchey (8/20/2012)
Actually, yeah, looks like 2005 might be out. That stinks. Never noticed that before.Any suggestions on what to do? I need to do some comparison testing aginst the new server to see how it fares against the current DB server, to see if the same workload occurring now will run better on the new system. I was going to run a non-gui trace (server side trace) and then play it back on the new server but I'm thinking here has to be a better way to do this. I'd prefer not to pays lots of moeny to do this but if nothing else I woudl think one of the SQL sofwtare Toools vendors woudl have come up with something like this, that lets you see how the same workload on one server will do on another. Make sense?
BTW - I have and did run SQL Imulator but thats it so far.
Thanks
First, I think you server side trace ides is better because it's going to behave like the actual database engine.
That being said there are tools that will just thrash the disks if that's what you need. As Grant already mentioned, SQLIO basically did just that (a bunch of reads and writes with files of a size you specified).
Take a look at PassMark's BurnInTest suite. It has disk tests built in. The standard and pro editions are both pretty reasonably priced.
Maybe I assumed incorrectly that this area (comparison testing of one SQL Server vs another under the same workload) is not as traversed as I thought. I assumed this kind of comparison testing was routine but maybe not.
What my boss wants is to see in advance of our move how the new DB server will fare against the exitsing one under the same workload. This is so as to verify that our current hardware configuration is as it shoudl be. The problem with using profiler/traces is that I have to wait for a good day (when its realy busy) to do my recording (server side trace) and as fate woudl have it we are at the slow period in our month. It will be a week before things pick back up and I can get a trace of a few hours of activity that will properly represnet what its like at our peak use times.
Make sense?
Thanks scott.
Kindest Regards,
Just say No to Facebook!August 21, 2012 at 6:50 pm
Yes, that does make sense. I'd want to see a good heavy load, especially if I was relying on the old single-threaded model of trace replay.
You'd think this would be a fully cooked area of development, but it's very bare. I was quite excited to see Distributed Playback come out in SQL Server 2012. I'm going to be doing a presentation on it at SQL Live in Orlando in December.
"The credit belongs to the man who is actually in the arena, whose face is marred by dust and sweat and blood"
- Theodore Roosevelt
Author of:
SQL Server Execution Plans
SQL Server Query Performance Tuning
August 22, 2012 at 12:24 pm
YSLGuru (8/21/2012)
Scott D. Jacobson (8/20/2012)
YSLGuru (8/20/2012)
Grant Fritchey (8/20/2012)
Actually, yeah, looks like 2005 might be out. That stinks. Never noticed that before.Any suggestions on what to do? I need to do some comparison testing aginst the new server to see how it fares against the current DB server, to see if the same workload occurring now will run better on the new system. I was going to run a non-gui trace (server side trace) and then play it back on the new server but I'm thinking here has to be a better way to do this. I'd prefer not to pays lots of moeny to do this but if nothing else I woudl think one of the SQL sofwtare Toools vendors woudl have come up with something like this, that lets you see how the same workload on one server will do on another. Make sense?
BTW - I have and did run SQL Imulator but thats it so far.
Thanks
First, I think you server side trace ides is better because it's going to behave like the actual database engine.
That being said there are tools that will just thrash the disks if that's what you need. As Grant already mentioned, SQLIO basically did just that (a bunch of reads and writes with files of a size you specified).
Take a look at PassMark's BurnInTest suite. It has disk tests built in. The standard and pro editions are both pretty reasonably priced.
Maybe I assumed incorrectly that this area (comparison testing of one SQL Server vs another under the same workload) is not as traversed as I thought. I assumed this kind of comparison testing was routine but maybe not.
What my boss wants is to see in advance of our move how the new DB server will fare against the exitsing one under the same workload. This is so as to verify that our current hardware configuration is as it shoudl be. The problem with using profiler/traces is that I have to wait for a good day (when its realy busy) to do my recording (server side trace) and as fate woudl have it we are at the slow period in our month. It will be a week before things pick back up and I can get a trace of a few hours of activity that will properly represnet what its like at our peak use times.
Make sense?
Thanks scott.
Totally reasonable. I've found myself in the same boat before. You would think this was well traveled territory but as Grant said, it's not. Server-side traces are still probably the best way to go. It is a shame that you won't be able to get that information sooner but if you can wait until you capture the data, I think you'll be pleased with what you get. The benefit of having a clearly defined idea of what improvements the new hardware offers over the old *before* going into production will probably outweigh the small amount of time lost by waiting to get the trace data.
August 22, 2012 at 5:14 pm
Scott/Grant,
Gotta say I feel a lot better about this after getting your posts.
I still will be able to make use of the Distributed Playback feature when the next phase of this transition comes into play, upgrading our application and the DB it uses to work with SQL 2008R2. Orginally the Admin & CIO wanted to upgrade to 2008R2 while moving to a new server and I convinced them to split that up into 2 phases by moving to a new server will remaining on 2005 and then a few months later after everything is running fine we can upgrade to 2008R2 (the vendor doesn't support 2012 as of yet and probably will not for a year or 2 at least).
I can't wait to start using the 2005/2008R2 features. Even though we are running SQl 2005 the DB is set to a DB Compatability level of 2000 (per the vendors specs because they do not support anything above SQL 2000 without a more recent upgrade to tehir prduct which comes with its own set of gotchas)
Thanks
Kindest Regards,
Just say No to Facebook!August 22, 2012 at 6:18 pm
is this what you were looking for - Date published:8/22/2012
http://www.microsoft.com/en-us/download/details.aspx?id=20163&mnui=5
August 23, 2012 at 10:46 am
prvmine (8/22/2012)
is this what you were looking for - Date published:8/22/2012http://www.microsoft.com/en-us/download/details.aspx?id=20163&mnui=5
Yes.... yes it is. Its interesting how this was published (accoridng to the download pages Publish date) yesterday.
How did you find it? If it was published just yesterday and was unavailable (for whatever reaason) for a while that would explain why you found it yesterday but I had not the days before.
Thanks
Kindest Regards,
Just say No to Facebook!Viewing 15 posts - 1 through 15 (of 21 total)
You must be logged in to reply to this topic. Login to reply