November 23, 2010 at 4:35 pm
Hi,
We recently migrated out MSSQL 2000 servers to another system.
We have both systems working in parallel for the next few weeks, at which point our original system will be shut down.
The original system, system_A, has new data and new requests coming in all the time, as it is live. The new system, system_B, has identical set of data and new data coming in just the same, but no requests are coming into it [e.g. no stored procedure calls].
The goal is to make sure that system_B is at least as good as system_A. By good I mean at least the following:
a. Stored procedure performance is at least as good [all applications just make stored procedure calls]
b. Replication is working fine
c. Jobs take similar amout of time
So far replication [transactional] and job time runs seem fine. My main concern right now are the stored procedure runs, which need to run in sub-second time.
I am guessing that I need to do some kind of stress testing on system_B.
I did some exploration and got some ideas/tools, but not sure which is the best way to go.
For example, I read about Profiler Replay option. I then recorded a trace file using Replay Template on system_A and then ran it on system_B and simultaneously running a standard trace on system_B to see what the results would be. Strangly, while the replay ran, nothing got recorded using the Standard Trace. I need to explore this further. Additionally, I read someone's post saying that Replay option is not real world case as there is no 'think time' added on.
Additionally, we've been using ACT tool to send db requests similar to the ones we have in production. The same requests were sent to one of the servers (which is currently out of production use) on system_A and system_B. Profiler trace was run on both systems and data collected. While this produced good comparison numbers, I feel that we need more systematic approach.
Can someone advise what they have used in such situations for making sure that the new system is as good or better than the old one?
Thank you!
November 24, 2010 at 5:26 am
Hi,
You could consider using a tool from Exact Solutions named iReplay which can capture and replay full scale production workload onto a test system : http://www.exact-solutions.com/products/ireplay
Disclosure : I work for Exact Solutions.
The MSSQL 2000 support for iReplay is coming out next month BTW.
November 24, 2010 at 10:17 am
Thanks Krishna. The tool you suggested looks interesting, but we can't wait till next month.
Has anyone tried this one: http://www.quest.com/benchmark-factory/
It seems they have free trial.
Are there any other free and good tools for stress testing available?
Thank you
Negmat
Viewing 3 posts - 1 through 2 (of 2 total)
You must be logged in to reply to this topic. Login to reply