Re: Any examples of companies using PG in high volume and/or large scale environs...

From: Andrew Sullivan <ajs(at)crankycanuck(dot)ca>
To: pgsql-admin(at)postgresql(dot)org
Subject: Re: Any examples of companies using PG in high volume and/or large scale environs...
Date: 2005-03-23 23:36:23
Message-ID: 20050323233623.GD27286@phlogiston.dyndns.org
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-admin

On Wed, Mar 23, 2005 at 02:07:03PM -0800, David B wrote:
> .info administrator might be dealing with 100 transactions per
> hour...or per second. Not seen anything on that. And a few hundred

The short answer to this is "both" (I am one of those .info people.
Also .org, .in, and a bunch of others).

Our databases are several gig on disk (for comparison, < 10G after
pg_dump is done). They are by no means very large databases these
days -- the entire top-level domain system of the Internet isn't even
a very large database these days, given that .com is under 40 million
name records. Also, our transaction volume is by no means huge as
compared to, say, eBay or Amazon.

What we do have is extremely "spikey" load, with some hours of
practically nothing followed by occasional periods of hundreds of
transactions per second. We have definitely had 10 million
transaction days, but I'm pretty sure we've never had a 100 million
transaction day in production (we've done tests to those levels,
though). We also have not-five-nines SLAs. I will tell you that the
users _expectation_ does not conform to the SLAs, and is instead
rather closer to five nines than I am happy with. I think this will
come as no surprise to anyone who's worked on the operations side of
databases.

You can find out more about our PostgreSQL use by looking for Afilias
on the ICANN web site (icann.org) -- look particularly for bid
materials for TLD delegations. We successfully backed the ISOC
proposal to take over .org from VeriSign in 2002. If you need more
detail, it'd be good to contact me at work (I'm andrew at ca dot
afilias dot info there).

> statements such as "we have a hugh database". One persons hugh is 1GB
> while another's is 100GB while another's is 1TB+. And as one who has
> dealt with TB+ in Oracle the issues are way different.

Yes. We also have some rather unpleasantly large off-line databases,
but nothing yet in the TB-in-one-db range. We are scraping that
range, however. In case it helps, we do have one system with
absolutely critical financial data in it, which we anticipate will be
a TB in not very long. Postgres has performed well for us so far,
thanks partly to the team of great DBAs I've been able to assemble.

A

--
Andrew Sullivan | ajs(at)crankycanuck(dot)ca
This work was visionary and imaginative, and goes to show that visionary
and imaginative work need not end up well.
--Dennis Ritchie

In response to

Browse pgsql-admin by date

  From Date Subject
Next Message Michael Fuhr 2005-03-24 01:34:39 Re: Resizing images contained in oid fields
Previous Message Andrew Sullivan 2005-03-23 23:15:53 Re: Clustering