> Hello hackers,
>
> i'm wondering if is possible to somehow spread pretty big db (aprox 50G)
> over few boxes to get more speed ?
> if anyone did that i'd be glad to have some directions in right way,
>
I have done different elements of clusering with PostgreSQL on a per task
basis, but not a fully comprehensive generic distributed cluster. There
are a couple tools you can use if your are an engineering sort of fellow.
Sloney is a replication cluster, all the data is on all the machines.
There is a project that shows promise as a distributed data system:
contrib/dblink. One could segment their database as a number of logical
data managers and use dblink to incorporate the data on one database into
the queries on another. It won't be transparent, but could be fairly
managable if you use views to implement the links.
I guess the real question is what performance do you need to improve? If
it is just read performance, then sloney is probably your best bet. Put a
number of redundant machines behind a load balancer and you are all set.
If you need to increase write performance, well, that can be problematic.
What is it that your want to accomplish?
