Here's my question about "multi-master" replication whether it's Oracle
or not. How in the world does it work over high latency, low bandwidth
connections w/o getting pummelled in performance? I mean it's fine when
you have whole bunch of servers in the same building. But I've got
servers across the country and waiting for 2-phase commit across a
100ms+ connection just sounds ugly like hell. I've tried some simple
tests -- a client running some simple transactions across a VPN from San
Francisco/CA to Sterling/VA and it's 10000 times slower than on a local
100mbit LAN.
Andrew Sullivan wrote:
> On Mon, Mar 14, 2005 at 09:47:24AM -0800, Chris Travers wrote:
>
>>Oracle and DB2 also offer an ability to parallelize queries across
>>nodes, so that you can query extremely large (multi-TB) data sets quickly.
>>
>>They also all market multimaster replication.
>
>
> "Market" is the operative word in the IBM case, note. It's a fine
> product, no question, but I've never been able to see any significant
> way in which it's really multimaster. And you can actually do the
> same sort of thing using the right combination of hardware and
> PostgreSQL. In fact, my colleagues and I (ok, mostly my colleagues
> -- I'm relegated to moving the furniture about) doing it right now.
>
> ORAC is in another league entirely. I hope we can do something about
> that soon, too.
>
> A
>
