<p dir="ltr">Hi,<p dir="ltr">Le 9 août 2014 05:57, "Ramirez, Danilo" <<a
href="mailto:Danilo.Ramirez@hmhco.com">Danilo.Ramirez@hmhco.com</a>>a écrit :<br /> ><br /> > Thanks to all
forthe great info. We are new to postgresql and this discussion has both instructed us and increased our respect for
thedatabase and the community.<br /> ><br /> > I am seeing a behavior that I don’t understand and hopefully you
guyscan clear it up.<br /> ><br /> > I am using AWS postgresql db.m3.2xlarge and using pgadmin III 1.18 comparing
againstAWS oracle on db.m3.2xlarge using sql developer and TOAD.<br /> ><br /> > I am running a query with 30
tablesin the from clause, getting 137 columns back (this is our most basic query, they get a lot more more complex).
Itreturns back 4800 rows.<br /> ><br /> > In oracle 1st run takes 3.92 seconds, 2nd .38 seconds. Scrolling to
endtakes and extra 1.5 seconds for total of 5.5.<br /> ><br /> > Using pgadmin, I run the query. Looking at the
lowerright hand I can see the time going up. It stops at 8200 ms or close to it every time, then it takes an extra 6
secondsbefore it displays the rows on the screen. 2nd, 3rd, etc. runs all take about same amount of time 8 sec plus 6
sec<br/> ><br /> > I then changed it to return only 1 column back. In oracle/sqldeveloper identical behavior as
before,same time. In postgresql it now goes down to 1.8 seconds for 1st, 2nd, etc. runs.<br /> ><br /> > I then
changeit so that I am asking for the sum of 1 column. In oracle time goes down to .2 seconds and postgresql now goes
downto .2 seconds also.<br /> ><br /> > I then change it back to get the full result set and behavior goes back
tooriginal, oracle .38 since its cached, postgresql 8 seconds.<br /> ><p dir="ltr">Are you sure this is postgresql 8
seconds?I'd believe this is more something like postgresql something really low and PgAdmin around 8 seconds displaying
it.What I mean is, PgAdmin uses really slow UI components and the time it shows is the time to execute the query and
displaythe data. IOW, you shouldn't use it to benchmark. You should better use psql. Or, much better, you should set
log_min_duration_statementto 0 and see exactly how much time postgresql needs to execute it.
