Josh,
* Josh Berkus (josh@agliodbs.com) wrote:
> I'd also love to hear from the PostgresXC folks on whether this solution
> works for them. Postgres-R too. If it works for all three of those
> tools, it's liable to work for any potential new tool.
AIUI, PG-XC and PG-R need an order, so they both use an independent
system (eg, the PG-XC GTM) to provide that ordering.
Again, AIUI, Simon's proposal would not guarantee any ordering but
instead would only guarantee non-overlap. Since the proposal being
pushed appeared to involve all the complexity of dealing with something
like a GTM, by having to have some third system which manages the
allocations, figure out what to do if it isn't available, etc, perhaps
the requirement to provide an ordering should be added on to it and
then the PG-XC GTM simply used for it.
If we're not going to have an ordering requirement then I'm not
convinced that the pre-allocation approach (where you break the space
up into many more blocks than you would ever expect to use and then
double-up those blocks on to the same physical system) is a bad
solution. It's certainly also been done a number of times, typically
quite successfully.
Thanks,
Stephen