Comparing transaction rates with HammerDB

Follow / 16 Sep 2014 at 9:26pm


I am comparing the results I get from HammerDB to the results I get with Benchmark Factory.  The numbers I get from HammerDB are much higher than I get from Benchmark Factory.  I also notice that the cpu consumed by the Oracle user processes is much lower when I used HammerDB than when I run Benchmark Factory.  The target DB is exactly the same in both cases, and I create a new schema for each tool.

I am using TPC-C in both cases.  I am curious to know why there is such a high discrepancy (up to 2X) when both tools claim to do TPC-C like workloads.




  • Follow / 16 Sep 2014 at 10:51pm

    I use HammerDB quite a lot. It does NOT implement "keying & think" time by default - you have to either check that option or set it manually in the config file. So HammerDB gets much higher numbers. Once you check this option - the numbers will be much more in line - in fact almost identical.

  • Follow / 18 Sep 2014 at 6:55pm

    Thanks Bert.  I have key, and think time set to zero on both HammerDB and Benchmark Factory.  

    Interestingly I get higher numbers on HammerDB for both SQL server and Oracle.  I am simply multiplying the TPM number generated by Hammer and comparing it to the TPS*60 value in BMF.  

  • Follow / 18 Sep 2014 at 7:50pm

    when I get some free time I'll try to reproduce this - I've done so in past and got same approx numbers. But I'll recheck when time permits .....