Re: Clustered index to preserve data locality in a multitenant application?

From: Kenneth Marshall <ktm(at)rice(dot)edu>
To: Nicolas Grilly <nicolas(at)gardentechno(dot)com>
Cc: Vick Khera <vivek(at)khera(dot)org>, pgsql-general <pgsql-general(at)postgresql(dot)org>
Subject: Re: Clustered index to preserve data locality in a multitenant application?
Date: 2016-08-31 16:22:15
Message-ID: 20160831162215.GG4950@aart.rice.edu
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-general

On Wed, Aug 31, 2016 at 06:06:54PM +0200, Nicolas Grilly wrote:
> On Wed, Aug 31, 2016 at 6:05 PM, Kenneth Marshall <ktm(at)rice(dot)edu> wrote:
>
> > We just run it via cron. In our case, we run it once a day, but depending
> > on
> > your churn, it could be run once a week or more.
> >
>
> Could you provide some numbers: what is the size of the tables or tables
> that are repacked? how long does it take? is there a performance impact
> during the repack?
Hi,

The key is that there must be enough I/O capacity to handle your routine
DB access needs plus the additional usage by the repack operation. Running
the repack during a slow period does not tangibly impact DB performance and
completes in a reasonable period of time. Your best best would be to set
up a test environment. By way of example, running our repack during peak
usage can take over 5 hours and less than 30m outside of peak when there
is excess I/O capacity. You should be able to use your I/O monitoring
system to get an idea about your I/O headroom.

Regards,
Ken

In response to

Browse pgsql-general by date

  From Date Subject
Next Message Nicolas Grilly 2016-08-31 21:55:47 Re: Clustered index to preserve data locality in a multitenant application?
Previous Message Nicolas Grilly 2016-08-31 16:06:54 Re: Clustered index to preserve data locality in a multitenant application?