Re: postgresql roadmap for horizontal scalability?

Timasmith,

> This database has exceeded all expectations and perhaps I am getting
> ahead of myself but is there anything in the roadmap to match Oracles
> ability to cluster databases across multiple servers for concurrent
> access?

There are several efforts to do *better than* RAC, which is actually not very 
effective (RAC scales very poorly, and not at all for some applications).  
These include pgPool, Skytools, pgCluster, Replicator, Postgres-R and 
Sequoia.

> The reason I ask is for the architecture of an open source database
> project I am working on.  It is tempting to ignore caching altogether
> and push the onus on to the database to perform more database queries.
> This is makes life easier and allows the focus to be on the business
> logic.

Yes, except even on a clustered system proper use of caching will *always* at 
least double database peformance.  There's simply no way to make an ACID 
RDBMS anywhere near as fast as a simple thing like a cache.

> However if Postgresql is the datababase and the only means to scale is
> to buy a bigger single server then that could become a problem of cost
> to the community.  It would be cheaper to scale across multiple smaller
> servers (I think).

Yes, it would.  However, scalable OLTP clustering is a problem not yet solved 
by anyone in the industry.  It may not be solvable at all.

> The application I am working on is OLTP with say 90% reads to writes
> ratio.

Hmmm ... that makes it more like web than OLTP.  You may be able to use a 
query replication system like pgPool or Sequoia.

-- 
Josh Berkus
PostgreSQL @ Sun
San Francisco