I'm curious as to why you expect there to be less network traffic. It
seems to me that the main traffic would come from query application,
and I don't see any reason for the change you are suggesting to cut
down on that.
Well, the way we've got this set up is table X has copies of itself on
three clusters. So among the three clusters, there's already network
connections open to each of their respective nodes. The new thing we
want to do is to create a "master copy" of table X, so that we can
just update once, and the changes would propagate to the three other
clusters. If each node in each cluster were to create a new network
connection to this "master copy" node, that would increase network
traffic; if we just left it to the providers on each cluster to
propagate to its subscribers, we'd use existing network connections,
and it would possibly be cleaner. Hence, the desire for the table to
subscribe via one setID, and provide via another setID.
The *possible* change would be for there to be less network traffic as
a result of fewer events propagating, but I don't see that changing
network traffic particularly materially.
True. In terms of net bytes transferred, there's no difference. I
suppose it's just the thought of more network connections that
initially raises flags.
--Richard
_______________________________________________
Slony1-general mailing list
[email protected]
http://lists.slony.info/mailman/listinfo/slony1-general