> Definately - that 10% number was on the old-first hardware (the core 2
> E6600). After reading my post and the 185MBps with 18500 reads/s number
> I was a bit suspicious whether I did the tests on the new hardware with
> 4K, because 185MBps / 18500 reads/s is ~10KB / read, so I thought thats
> a lot closer to 8KB than 4KB. I checked with show block_size and it was
> 4K. Then I redid the tests on the new server with the default 8KB
> blocksize and got about 4700 tps (TPC-B/300)... 67/47 = 1.47. So it
> seems that on newer hardware, the difference is larger than 10%.
That doesn't make much sense unless there's some special advantage to a
4K blocksize with the hardware itself. Can you just do a basic
filesystem test (like Bonnie++) with a 4K vs. 8K blocksize?
Also, are you running your pgbench tests more than once, just to account
for randomizing?
--
-- Josh Berkus
PostgreSQL Experts Inc.
http://www.pgexperts.com