Re: BIG installations of PostgresQL?

-----Original Message-----
From: pgsql-general-owner@postgresql.org on behalf of Logan Bowers
Sent: Thu 9/15/2005 10:25 PM
To: pgsql-general@postgresql.org
Subject: [GENERAL] BIG installations of PostgresQL?

Hello,



My company is looking at two very large DB vendors to bear most of data
burden of our company; however I'd like to propose PostgresQL to handle
some of the tasks, specifically a large number read-only search DBs.
Postgres already has the win in terms of features and it's now down to
risk. 



Many of the other folks at the company feel an open-source DB is more
risky because it is less well tested compared to commercial
counterparts.  I'm looking for examples of large installations of
Postgres with huge data sets, high traffic volumes, high update rates,
etc, particularly large, recognizable names.  I know you guys get this
question a lot, but can anyone share experiences with using Postgres in
large settings with huge query rates, replication, etc?  Any stories you
guys can share (in public or private) would be greatly appreciated.
Thanks in advance! 

------------- END OF ORIGINAL MESSAGE --------------------

(I apologize if this thing is going out as html mail, I'm stuck on a web - Exchange client at the moment...)

If they're read only databases, your risks are minimized, because you can always keep backup machines ready, and spread the load if needed.  Downtime can be minimized

There are many large to VERY large sites running on PostgreSQL, I'll let those folks on the list that are on large / high traffic pgsql sites list themselves for ya.  But it's especially popular with folks who need to do GIS stuff since it understands spatial types and data, and can index them.  also, it's ability to have functional and / or partial indexes is quite useful too.

The only test that counts is how it works for you.  If Oracle works for 1,000,000 other sites but can't handle your particular load, then it's a bad choice, not because it was or wasn't tested thoroughly by others, but because it doesn't work for you.

PostgreSQL gets a LOT of testing.  Look up OSDL, who use postgresql running a variety of loads to test and tune linux based servers.  Plus, being free, there are literally thousands of thousands of small installations that use it for things like back end databases for ticketing / tracking and reporting systems.  So, it's got lots of testing in the small to middle sized database category.

In the last couple of years it has started receiving the kind of testing under large and heavy loads, and the features and improvements to go with it, that allow it to start encroaching on more and more of the enterprise territory that Oracle and DB2 and a few of the big boys hold.  Is it equal to Oracle or DB2? 

That said, one problem I have never had with PostgreSQL is unreliability.  Even when beaten firmly into the ground under load, it just keeps working.  While certain loads present problems (high update rates that outrun vacuum can be an issue, but much less now than a few years ago) most (nearly all really) are handled with no hiccups or interruptions.

That said, if you want a "commercial" database, there are companies that are more than willing to sell you a commercialized version of PostgreSQL with their support and testing to stand behind it.  If your company is bound and determined to spend money to buy a database, they can do that with PostgreSQL too.  And sometimes, it's not a bad idea, especially when you're just starting out.