Toad World® Forums

TPC-H not doing anything?


#1

I am trying to get a TPC-H benchmark to work, but I am running into problems.
this is a SQL server 2005, SP2
Under Settings, I have set the total length to 30 minutes
I wen through the wizard to create a TPC-H test, scale factor of 1, 1 user (attached is the config file)
When I go to submit the job It generates the tables and populates them.
I then click on the Run Status, and it shows nothing going on, it says pre-sampling, sampling, but under any of the stats, they all show 0. then after about 3 minutes, it stops. If I look at the report, it does show it ran, and completed succsefully, but did not run for 30 minutes.

Ideally, I would like the test to run for 30 minutes, with 10 users.

Version 5.7.1 build 504
InstallBetaPatch.bat (185 Bytes)


#2

Looking at your attached script file the replay test is still running as one user with only one execution. If you want the test to run for 30 minutes with ten virtual users you will need to check the “Execute by time” checkbox, set the time to 30 minutes, and then change the number of users for Stream 0 to 10.

Note that the test will not just stop at 30 minutes as BMF will allow any running transactions to complete before stopping the test, so the test will probably go longer.
Also I recommend that instead of changing the number of users for Stream 0 to 10, that you use the TPC-H Stream script file that can be found in the community library. Here is the link;

http://benchmarkfactory.inside.quest.com/entry.jspa?externalID=2392&categoryID=27

The timing values that are set under the settings menu item are just the defaults for the different types of tests/load scenarios. Since the TPC-H specification for the Power Test (the single user test), is just a single execution, that is what BMF creates and does not use the defaults.


#3

Okay, I think I may understand it.
The amount of time the test runs, has nothing to do with the settings options.
It is a balance of number of users to the number of executions.
So, I figure 1 user doing 1 execution takes just over 1 minute, 2 users at 1 execution would take 2 minutes, and 2 users at 2 executions would take about 4 minutes (more or less on these times), does that seem to be correct thinking?