Re: Comparative tps question

On 29/11/2012 17:33, Merlin Moncure wrote:
> one thing that immediately jumps out here is that your wal volume
> could be holding you up. so it's possible we may want to move wal to
> the ssd volume. if you can scrounge up a 9.2 pgbench, we can gather
> more evidence for that by running pgbench with the "--unlogged-tables"
> option, which creates the tables unlogged so that they are not wal
> logged (for the record, this causes tables to be truncated when not
> shut down in clean state).
Ok, got myself a 9.2 version of pgbench and run it a few times on
unlogged tables...
changing the number of threads has maybe a 5% change in values which
isn't probably too much to worry about.
-j 25 -c 100 -s 100 gives a tps of around 10.5k
using -N ups that to around 20k
using -S ups that again to around 40k

I'll have to wait until I get to shut the db down again to try the wal
on an ssd. Although unless I did something wrong it didn't seem to make
a huge difference before....

During these tests, iowait dropped to almost 0, user and sys stayed
around the same (60% and 20% respectively). although the disk traffic
was only in the 10s of Mb/s which seems very low - unless there is some
wierd caching going on and it gets dumped at a later date?

John

--
Get the PriceGoblin Browser Addon
www.pricegoblin.co.uk