Thanks Huxton & Gerd; I think pgbench is not suitable for those who want to test application on desired tables. pgbench is highly hard coded and tests only self generated tables. I used Mercury LoadRunner but it is not freeware. The link http://bristlecone.continuent.org/HomePage only has details about this software but not the download link. Secondly it is for Unix not for Windows. And I found AppLoader freeware but it does not include PostgreSQL in its database list. My research is ongoiung and will prompt you as I get something fruitful and expecting the same from you.
Regards, Abdul Rehman. ________________________________ From: Richard Huxton <d...@archonet.com> To: Abdul Rahman <abr_...@yahoo.com> Cc: pgsql-general@postgresql.org; ashis...@synechron.com Sent: Wednesday, February 18, 2009 2:13:27 PM Subject: Re: [GENERAL] Download link for 'Bristlecone' Abdul Rahman wrote: > Dear All, > > I searched a lot but failed to find any download link for Bristlecone. Kindly > help me. Or let me know about any open source Load Tester for PostgreSQL > based application. > Ashish, according to my search Bristlecone is open source. Thanks in advance. Sorry - never heard of Bristlecone. You might want to start with "pgbench" which is in the contrib/ directory of the source distribution: http://www.postgresql.org/docs/current/static/pgbench.html Make sure you read the manual page above, and also search the mailing-list archives for tips on how to get useful information from it. You might also be interested in the sourceforge DBT project http://sourceforge.net/projects/osdldbt/ Read this msg first though, there is more up-to-date code available than the sf.net download page. http://archives.postgresql.org/pdxpug/2008-12/msg00010.php Really though, unless you're doing this as part of an academic paper you'll want to write your own test cases. It doesn't matter whether PostgreSQL is fast or slow on the tests above, what matters is how it performs *for your purposes*. The only way to test that is to generate the sort of data and workload that you plan to have. Oh, and it's very easy to end up with data that tells you nothing useful. Make sure your tests are long enough, and that they're not too sensitive to minor changes in configuration or load. HTH -- Richard Huxton Archonet Ltd
