Re: PostgreSQL performance on 64 bit as compared to 32 bit

From: Claudio Freire <klaussfreire(at)gmail(dot)com>
To: Umesh Kirdat <umesh(dot)kirdat(at)yahoo(dot)com>
Cc: "pgsql-performance(at)postgresql(dot)org" <pgsql-performance(at)postgresql(dot)org>
Subject: Re: PostgreSQL performance on 64 bit as compared to 32 bit
Date: 2012-09-22 02:43:49
Message-ID: CAGTBQpawtxo+u0BArtXX_HBq4qQ6dbuarXT8Uy4FVbv0-xZL9g@mail.gmail.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-performance

On Sun, Sep 16, 2012 at 12:48 AM, Umesh Kirdat <umesh(dot)kirdat(at)yahoo(dot)com> wrote:
> The issue we have noticed is the 9.0.4 (64 bit) version of PostgreSQL has
> slower performance as compared to 8.2.2 (32 bit) version on an identical
> hardware.

First of all, that's comparing apples and oranges. Compare the same
version in 32-vs-64, and different versions on same-arch.

> To investigate further we tried monitoring the PostgreSQL process using
> strace and found that the earlier version of PostgreSQL was using _llseek()
> system call whereas the later version is using lseek() system call.

Second, I doubt that's the problem. It's most likely increase memory
footprint due to 64-bit pointers, a known overhead of the 64-bit arch,
but a price you have to pay if you want access to more than 3-4GB of
RAM. You'll be better off using a profiler, like oprofile, and compare
the profile between the two arches.

In response to

Browse pgsql-performance by date

  From Date Subject
Next Message Claudio Freire 2012-09-22 03:08:15 Re: Cost of opening and closing an empty transaction
Previous Message Brandon 2012-09-21 17:08:28 Query Planner Optimization?