Kevin Grittner wrote:
> I'm not clear on exactly what you're proposing there, but the thing
> I've considered doing is having threads to try to keep a FIFO queue
> populated with a configurable transaction mix, while a configurable
> number of worker threads pull those transactions off the queue and...
>
This is like the beginning of an advertisement for how Tsung is useful
for simulating complicated workloads. The thought of growing pgbench to
reach that level of capabilities makes my head hurt.
When faced with this same issue, the sysbench team decided to embed Lua
as their scripting language; sample scripts:
http://bazaar.launchpad.net/~sysbench-developers/sysbench/0.5/files/head:/sysbench/tests/db/
This isn't very well known because the whole MySQL community fracturing
has impacted their ability to actually release this overhaul--seems like
they spend all their time just trying to add support for each new engine
of the month. I don't even like Lua, yet this still seems like a much
better idea than trying to build on top of the existing pgbench codebase.
--
Greg Smith 2ndQuadrant US greg@2ndQuadrant.com Baltimore, MD
PostgreSQL Training, Services, and 24x7 Support www.2ndQuadrant.us