David Parker wrote:

>I have 3 nodes A,B,C. 
>
>A and B are "clustered" such that B is just a backup for A, and needs to get all changes made on A replicated. A in its turn is a subscriber to a set of tables from C, and also has some tables for which it is master. 
>
>Now, for B to be a full backup to A, it needs to be subscribed to the A-mastered tables, and to the set from C, as well.
>
>My question: is it better/more efficient for B to just subscribe directly to C, or for it to subscribe to C's set via A, with A having forwarding turned on?
>
>TIA for any information/opinions/corrections....
>  
>
You need to define metrics for "better" and/or "more efficient" in order 
to determine which subscription is preferable.

- If B subscribes directly to C, then it will get those updates more 
quickly than if it waits for them to be provided by node A.

- If B subscribes to A, for the updates originating on C, then it won't 
get them as quickly as if they came direct from C, but this reduces the 
load on node C.

There lies the trade-off:  Do you want B to be "more up to date," or do 
you want to take load off of node C?