Hello everybody,
I'm experiencing a performance related issue during some validation
measurements.
Let's first of all clarify what kind of problem I am facing.
I've set up a plain tpc-h database without any additional indexes or
something like that.
To get some performance impressions I wrote a little script that uses 1)
pg_ctl to start postgres, 2) executes the same query several times and
finally 3) stops the postgres instance using pg_ctl again.
These 3 steps are executed several times resulting in varying performance
results of which I think that they should be smaller (taking runtime
differences I don't have any influence on into account) .
Run1:
real 0m15.005s
user 0m0.000s
sys 0m0.000s
Run2:
real 0m14.012s
user 0m0.000s
sys 0m0.000s
Please be aware that I provided numbers taken from the middle of each run
(to exclude any io-related effects), so the buffers are warm, I disabled
intels' turbo boost feature and that I have exclusive access to the machine.
I used explain (analyze, buffers) to make sure no data has to be loaded from
disks during runtime (top and vmstat did not show any movements as well).
I flushed the os buffers (no effect regarding performance difference), I
disabled autovacuum, I adapted my buffer sizes to hold the whole data
required, but I was not able to sort the performance difference out yet.
The query I'm currently using is quite easy:
select
min(l_extendedprice * l_discount * l_tax)
from lineitem;
Note that I'm aware that my shared_buffer settings has to be bigger than 1/4
of the table to get past the ring buffer.
Would be great if anyone could give me some advise where to look at or
provide me with some information about postgres performance behaviour.
Thank you very much
--
View this message in context:
http://postgresql.1045698.n5.nabble.com/Varying-performacne-results-after-instance-restart-tp5801717.html
Sent from the PostgreSQL - performance mailing list archive at Nabble.com.
