On 18/03/2011 19:17, Ben Chobot wrote:
> if we're talking an extra 50MB of memory per cluster, that will start to add up.
Consider this: each such cluster will have:
a) its own database files on the drives (WAL, data - increasing IO)
b) its own postgresql processes (many of them) running in memory
c) its own shared_buffers in memory.
It is highly unlikely that you will manage anything decent with this
type of configuration with a non-trivial number of clusters.