At 11:11 AM 12/13/2006, Cosimo Streppone wrote:
Interesting, eh? Cosimo
What I find interesting is that so far Guido's C2D Mac laptop has gotten the highest values by far in this set of experiments, and no one else is even close. The slowest results, Michael's, are on the system with what appears to be the slowest CPU of the bunch; and the ranking of the rest of the results seem to similarly depend on relative CPU performance. This is not what one would naively expect when benching a IO intensive app like a DBMS.
Given that the typical laptop usually has 1 HD, and a relatively modest one at that (the fastest available are SATA 7200rpm or Seagate's perpendicular recording 5400rpm) in terms of performance, this feels very much like other factors are bottlenecking the experiments to the point where Daniel's results regarding compiler options are not actually being tested.
Anyone got a 2.33 GHz C2D box with a decent HD IO subsystem more representative of a typical DB server hooked up to it?
Again, the best way to confirm/deny Daniel's results is to duplicate the environment he obtained those results with as closely as possible (preferably exactly) and then have someone else try to duplicate his results.
Also, I think the warnings regarding proper configuration of pgbench and which version of pgbench to use are worthy of note. Do we have guidance yet as to what checkpoint_segments should be set to? Should we be considering using something other than pgbench for such experiments?
Ron Peacetree
