Timasmith wrote:
>
> That is awesome. I am not worried about the time to replicate - this
> is supposed to be static, very infrequently changing data (like form
> controls, security etc.) so there is no expectation changes are real
> time.
Be very sure that this is covered in the specifications or you may end
up with PHB types asking you why the changes aren't propagated in real
time.
> I am envisioning a powerful,.highly available central server and then a
> slew of cheap servers.
If you intend to use the slony based approach above, then I suggest
that your cheap servers need to be reliable. In the event of a single
node's failure, cruft will begin accumulating on all the nodes until
that node is either dropped or brought back online and catches back up.
Drew
