Re: large database - Mailing list pgsql-general

From Chris Angelico
Subject Re: large database
Date
Msg-id CAPTjJmqzGBcnorBMQ3UU6YTNf5Mw6kHPkLhkJBi9NQxhy0R8pw@mail.gmail.com
Whole thread Raw
In response to large database  ("Mihai Popa" <mihai@lattica.com>)
Responses Re: large database  (Gavin Flower <GavinFlower@archidevsys.co.nz>)
Re: large database  (Ben Chobot <bench@silentmedia.com>)
List pgsql-general
On Tue, Dec 11, 2012 at 7:26 AM, Mihai Popa <mihai@lattica.com> wrote:
> Second, where should I deploy it? The cloud or a dedicated box?

Forget cloud. For similar money, you can get dedicated hosting with
much more reliable performance. We've been looking at places to deploy
a new service, and to that end, we booked a few cloud instances and
started playing. Bang for buck, even the lower-end dedicated servers
(eg about $35/month) majorly outdo Amazon cloud instances.

But don't take someone's word for it. Amazon let you trial their
system for a year, up to (I think) ~750 computation hours per month,
of their basic instance type. You can find out for yourself exactly
how unsuitable it is! :)

The fact is that cloud platforms offer flexibility, and that
flexibility comes at a significant cost. I don't think PostgreSQL can
adequately exploit X nodes with 600MB RAM each, while it _can_ make
excellent use of a single computer with gobs (that's a new SI unit,
you know) of memory.

Incidentally, I've heard tell that cloud instances can vary enormously
in performance through the day or week, but we did some cursory
testing and didn't experience that. That doesn't prove you won't have
problems, of course, but it's one of the purported downsides of
clouding that clearly isn't as universal as I've heard said.

ChrisA


pgsql-general by date:

Previous
From: Jan Kesten
Date:
Subject: Re: large database
Next
From: Gavin Flower
Date:
Subject: Re: large database