October 12, 2010 at 1:20 pm
Wouldn't you have to arrange an asynchronous test to truly duplicate what is going on in production?
October 12, 2010 at 1:29 pm
robbemorris (10/12/2010)
Wouldn't you have to arrange an asynchronous test to truly duplicate what is going on in production?
Not sure what you mean.
I had SSMS opened and I had 5 connections running a loop for 2 minutes nonstop.
You could change that to had ms delays to simulate "real" use. Also you could increase from 5 connections to 100 from multiple machines if you wanted to.
October 12, 2010 at 1:52 pm
You may want to check your wait stats during a test to see what the delay is.
October 12, 2010 at 1:54 pm
My gut tells me to convert his process to inserting a record for each request and then running a scheduled job to get the counts and update tableA with a single update statement for each record.
You should do this. There are very few reasons where you would need to have a record updated several times per second in real-time. There will also be significantly less stress on the server.
October 12, 2010 at 2:45 pm
I'm talking about thousands of simultaneous connections and requests. A simple loop with 5 or even a 100 simulated requests won't raise the same kinds of issues with this much attention to the same record all at the exact same time.
I'm already moving forward with a different strategy. Thanks for everyone's input. If the moderator can close the thread, please do so.
Viewing 5 posts - 16 through 19 (of 19 total)
You must be logged in to reply to this topic. Login to reply