We have the following system design for a rather large distributed system 
[poor ascii diagram below].
We are trying to have 1 master (R/W) and a secondary as hot-spare, then a 
2nd tier of slaves that are data for the systems below them. 

This was to reduce network use between the network components and concerns 
about eventual database load.

Using the slony-setup.pl we have managed to create a working setup, but it 
seems ineffecient as every node talks back to the master (postgres01) and 
seemingly ever other node.

So we are trying to set it up so each tier has no knowledge beyonds it top; 
IE: 3rd tier machines only know about the 2nd tier , so postgres10:dbB 
would only know about postgres11/12 and postgres09:dbB

I was hoping that someone here has done something somewhat similar and can 
offer some adivce/sample configs to help out.

Thanks.
---
[1st teir]

postgres01:master 
postgres02:master 

[2nd Tier - First branch]
--postgres03:slavemaster1-dbA 

[3rd tier - First branch]
	--postgres04:dbA 
	--postgres05:dbA 
	--postgres06:dbA 
	--postgres07:dbA 
	--postgres08:dbA 

[2nd Tier - First branch]
--postgres03:slavemaster1-dbB 

[3rd tier - First branch]
	--postgres04:dbB 
	--postgres05:dbB 
	--postgres06:dbB 
	--postgres07:dbB 
	--postgres08:dbB 



[2nd Tier - Second branch]
--postgres09:slavemaster2-dbA 

[3rd tier -  Secondbranch]
	--postgres10:dbB 
	--postgres11:dbB
	--postgres12:dbB 

[2nd Tier - Second branch]
--postgres09:slavemaster2-dbB

[3rd tier - Second branch]
	--postgres10:dbB 
	--postgres11:dbB 
	--postgres12:dbB 
-----------


---------------------------------------------