Re: COPY v. java performance comparison - Mailing list pgsql-general

From Rob Sargent
Subject Re: COPY v. java performance comparison
Date
Msg-id 533D75B6.5080705@gmail.com
Whole thread Raw
In response to Re: COPY v. java performance comparison  (Adrian Klaver <adrian.klaver@aklaver.com>)
List pgsql-general
On 04/02/2014 08:40 PM, Adrian Klaver wrote:
On 04/02/2014 05:30 PM, Rob Sargent wrote:
On 04/02/2014 06:06 PM, Adrian Klaver wrote:
On 04/02/2014 02:27 PM, Rob Sargent wrote:
On 04/02/2014 03:11 PM, Adrian Klaver wrote:
On 04/02/2014 02:04 PM, Rob Sargent wrote:
On 04/02/2014 02:36 PM, Adrian Klaver wrote:
On 04/02/2014 01:14 PM, Rob Sargent wrote:
On 04/02/2014 01:56 PM, Steve Atkins wrote:


Have you tried moving the input file to the same disk as the server,
to factor out the controller?

I labour under the delusion that it is through the controller one reads
and writes and that there might be some slight advantage to not doing
both against one drive if avoidable. Wrong again?

Well there is one way to find out:)

Might try with something less then the whole file to get come up an
approximate row/sec rate.



Well things slow down over time, and lots of "too frequent"s:

    Have done 500 batches in 24219 ms
    Have done 1000 batches in 52362 ms
    Have done 1500 batches in 82256 ms
    Have done 2000 batches in 113754 ms
    Have done 2500 batches in 149637 ms
    Have done 3000 batches in 211314 ms
    Have done 3500 batches in 301989 ms
    Have done 4000 batches in 430817 ms
    Have done 4500 batches in 596043 ms
    Have done 5000 batches in 804250 ms

where a batch is 500,000 lines.  This on the java side of course.

This is drive to drive or on single drive?







Same as first go round, drive to drive.

pgsql-general by date:

Previous
From: Tom Lane
Date:
Subject: Re: size of interval type
Next
From: Adrian Klaver
Date:
Subject: Re: Any way to insert rows with ID used in another column