Re: Oracle vs PostgreSQL in real life

From: Jean-Paul ARGUDO <jean-paul(dot)argudo(at)idealx(dot)com>
To: Hannu Krosing <hannu(at)tm(dot)ee>
Cc: Jean-Paul ARGUDO <jean-paul(dot)argudo(at)idealx(dot)com>, pgsql-hackers(at)postgresql(dot)org, sdinot(at)idealx(dot)com, dbarth(at)idealx(dot)com
Subject: Re: Oracle vs PostgreSQL in real life
Date: 2002-02-27 18:21:46
Message-ID: 20020227192146.A9303@singer.ird.idealx.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

> What was the postgresql.conf set to ?

I put parameters in another mail, please watch for it.

> > The "test" is a big batch that computes stuffs in the database.
> Could you run this batch in smaller chunks to see if PG is slow from the
> start or does it slow down as it goes ?

The batch starts really fast and past 2 minuts, begins to slow down dramatically
and never stops to get slower and slower


> > Linux Red Hat 7.2 with PostgreSQL 7.2 : hours to go (statistically, 45 hours),
> > 80 tps (eighty tps).
> What kind of tps are these ?

Here's what we have in output:

This is the WINDOWS NT4 / Oracle 8.0 ouput when the batch is totally finished:

Time : 00:47:50

Transaction : 25696
Item : 344341
Transaction (in milliseconds) : 111
Item (in milliseconds) : 8

Errors : 0
Warnings : 0
PLU not found : 0
NOM not found : 0
Alloc NOM : 739
Free NOM : 739
Error 1555 : 0

Read : 2093582
Write : 1772364
Read/Write : 3865946

Free memory (RAM) : 117396 Ko / 261548 Ko

PLU SELECT : 344341
NOM SELECT : 1377364
T04 SELECT : 1840
T01 INSERT : 593
T01 UPDATE : 1376771
T02 INSERT : 28810
T02 UPDATE : 315531
T03 INSERT : 41199
T13 INSERT : 9460
RJT INSERT : 0
RJT SELECT : 0

--------------------
Beware "Transaction" does not mean transaction.. a "transaction" here contains one ore
more "item", in the context of the application/database.

What for real DML orders: 3.865.946 queries done in 47 min 50 secs. (the queries
are reparted in many tables, look for detail couting under "Free memory..."
line.. (a table name is 3 letters long)

Thats 1 347 queries per second... -ouch!

This is the Linux Red Hat 7.2 / PostgreSQL 7.2 port of the Pro*C program
producing the output

As you'll understand, it is not the COMPLETE batch, we had to stop it..:

Time : 00:16:26

Transaction : 750
Item : 7391
Transaction (ms) : 1314
Item (ms) : 133

Errors : 1
Warnings : 0
PLU not found : 0
NOM not found : 0
Alloc NOM : 739
Free NOM : 0
Error 1555 : 0

Read : 45127.000
Write : 37849.000
Read/Write : 82976.000

PLU SELECT : 7391
NOM SELECT : 29564
T04 SELECT : 31
T01 INSERT : 378
T01 UPDATE : 29186
T02 INSERT : 3385
T02 UPDATE : 4006
T03 INSERT : 613
T13 INSERT : 281
RJT INSERT : 0
RJT SELECT : 0

---------------- you see

we have 82.976 queries in 16 min 26 seconds thats a

84 queries per second

--

definitely nothing to do with Oracle :-((

Very bad for us since if this customers kicks Oracle to get PG, it can be really
fantastic, this customer has much influence on the business....

Thanks for helping me that much to all of you.
--
Jean-Paul ARGUDO

In response to

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Oleg Bartunov 2002-02-27 18:29:31 Re: single task postgresql
Previous Message Oleg Bartunov 2002-02-27 18:17:29 PocketSQL