PostgreSQL for a mission critical system

Greetings everyone, I'm new to this list (No, I'm not about to ask how to
install pgsql on Windows 3.11!)

We are involved with a financial project, similar to Paypal (Web Payments
Standard/Payment Buttons) but the scope and functionality is much narrower. This
project will be released in a market niche that will not exceed thirty thousand
(30000) users in the very best (and very optimistic) of scenarios and will
surely have a much lower transaction/user/day ratio than Paypal.

Our development department has to take all the technical decisions (what RDBMS
to use, what additional personnel is required and what equipment to
acquire/rent), and at the head of those decisions is, unfortunately, me. We have
vast experience in MySQL, though not in large-scale or mission critical database
setups. We have reviewed some MySQL specific use cases and whitepapers and have
done intensive research (Paggo is a highlighted use case here, since it's
similar though much bigger).

However, I refuse to take the bait and just follow the propaganda. I've been
working with PostgreSQL a few (very few) times before and I believe it's
superior, TCO is much better (since you hire expertise and not licenses), but I
don't know how easy it can be to scale up/out and if it can handle the most
optimistic scenario (optimistic from the business point of view, for me it will
be a pain in the ...).

In addition, we have to design this application in such a way that it can be
scaled up/out to the presumed limit (30000 users) by just upgrading hardware,
acquiring new hardware, or modifying the hardware setup. We don't want to
redesign the application or modify it heavily once launched, though some tune up
or bug fix will surely happen. We have our way to deal with this, we've planned
for some front-end web applications (written in PHP, with caching and
acceleration) and a back-end which is responsible for all the database
management (probably written in Java and C/C++, we haven't decided yet). The
front-end and back-end would communicate by means of a web services API (say
SOAP, or the more lightweight XML-RPC), this will allow us to scale the
front-end and back-end parts separately.

This system could grow to more than 30000 users (say, to a million) in the very
far future, but for that to happen, this "pilot test" should be a complete
success (which means our client pocket would be full of money) and the market
niche should have to be extended. Meaning, if we need to jump higher, we can
redesign our application and/or switch to another database/programming
technology, such as Oracle (which I strongly hate :@). However, it's more likely
that if we succeed, you find me again on this list looking for arguments to
adopt PostgreSQL :D.

What I would like to know (if you are so kind ;) ) is:

1) Do you really think that we can start low (say, a master/master or
master/slave replication system on commodity servers) and jump high (say,
upgrading to Opteron 8300 servers)? That is... how much can we scale up, without
the need to scale out?

2) Do you know of any Hosting/Collocation service that can provide PostgreSQL
installation/configuration support at an affordable price? (This is very
important in the early stages). I've seen the hosting list at postgresql.org,
but if possible, please share your experiences.

3) Any articles, benchmarking, white papers, ebooks, seminars, personal
experience, use cases that you can recommend about this? (Yes, I've downloaded
the postgresql.org use cases and I can use Google, but your voice is louder,
specially if it speaks for your experience)

4) Any other comments are welcome.

Thanks in advance (and sorry if I have bothered you in any way),

Andrés Robinet | Lead Developer | BESTPLACE CORPORATION 
5100 Bayview Drive 206, Royal Lauderdale Landings, Fort Lauderdale, FL 33308 |
TEL 954-607-4296 | FAX 954-337-2695 |
Email: info(at)bestplace(dot)net  | MSN Chat: best(at)bestplace(dot)net  |  SKYPE: bestplace |
 Web: bestplace.biz  | Web: seo-diy.com