On 5/1/2012 8:06 AM, Merlin Moncure wrote:
On Tue, May 1, 2012 at 7:51 AM, Walker, James Les<JAWalker@xxxxxxxxxx> wrote:
Exactly, if turning off fsync gives me 100 commits/sec then I know where my bottleneck is and I can attack it. Keep in mind though that I already turned off synchronous commit -- *really* dangerous -- and it didn't have any effect.
well synchronous commit is not as dangerous:
fsync off + power failure = corrupt database
synchronous commit off + power failure = some lost transactions
still waiting on the ssd model #. worst case scenario is that you tps
rate is in fact sync bound and you have a ssd without capacitor backed
buffers (for example, the intel 320 has them); the probable workaround
would be to set the drive cache from write through to write back but
it would unsafe in that case. in other words, tps rates in the triple
digits would be physically impossible.
another less likely scenario is you are having network issues
(assuming you are connecting to the database through tcp/ip). 20
years in, microsoft is still figuring out how to properly configure a
network socket.
merlin
Even if its all local, windows doesnt have domain sockets (correct?), so
all that traffic still has to go thru some bit of network stack, yes?
-Andy
--
Sent via pgsql-performance mailing list (pgsql-performance@xxxxxxxxxxxxxx)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-performance
