Re: external sort performance - Mailing list pgsql-performance

From Jon Nelson
Subject Re: external sort performance
Date
Msg-id CAKuK5J2Z5d=ADthwGvaxu2OVJUHMwsHQSJvpmPtskRsZAkHKzQ@mail.gmail.com
Whole thread Raw
In response to Re: external sort performance  (Jeremy Harris <jgh@wizmail.org>)
Responses Re: external sort performance
List pgsql-performance
On Sun, Nov 20, 2011 at 7:56 AM, Jeremy Harris <jgh@wizmail.org> wrote:
> On 2011-11-17 17:10, Jon Nelson wrote:
>>
>> external sort ended, 7708696 disk blocks used: CPU 359.84s/57504.66u
>> sec elapsed 58966.76 sec
>>
>> Am I to understand that the CPU portion of the sorting only took 6
>> minutes but the sort itself took almost 16.5 hours and used approx
>> 60GB of disk space?
>
>
> I realise you've had helpful answers by now, but.... that reads
> as 16 hours of cpu time to me; mostly user-mode but with 6 minute
> of system-mode.  98% cpu usage for the 16 hours elapsed.

Thank you very much!
I was going to post a followup asking for help interpreting the log
line, but now I don't have to. Do you happen to recall if disk I/O is
counted as user or system time? If it's counted as system time, then I
have more questions, namely:

If using a hash table (hash aggregation) shows that the GROUPing can
take place in 35 minutes, but a Group Aggregation takes 16 hours, how
much of that is CPU and how much is waiting for I/O?


--
Jon

pgsql-performance by date:

Previous
From: Jeremy Harris
Date:
Subject: Re: external sort performance
Next
From: Jeremy Harris
Date:
Subject: Re: external sort performance