gordon fecyk

On the network I ranted about in an earlier post, I have each DC running
WINS, and the BDCs at each office run DHCP. Browsing works fine and each
office's client computers are smart enough to log on to their office's BDC
as I've set the WINS node type in DHCP to 0x4 (mixed, meaning broadcast
first) as recommended in KB 142692. But I'm trying to cut down on the WINS
replication traffic and I'd like to know the best scenario I can use for
WINS inter-server replication.

The offices interconnect using Snapgear VPN boxes, and the minimum
inter-office speed seems to be about 128 kbps. The branches have what my
ISP calls "business 1" service, which is 1.5meg down, 160 kb up, though it
feels like ISDN. It does cost 1/3rd the price of ISDN though, not to
mention I don't need three ISDN lines at the main office, so I'm not
complaining. The interconnects work quite well and I do file replication
and have the same set of services on each BDC to cut down on inter-office
traffic. The main office Snapgear doubles as a PPTP server for home office
machines too.

Each BDC runs WINS and each office uses its BDC as the primary WINS server.
The PDC runs WINS to act as a secondary WINS server for everyone. The
servers themselves only use themselves as primary and secondary WINS, as
strongly recommended in KB 150520. Because of this, I need to make sure
each WINS server has a copy of every other WINS server's database, which is
what WINS replication is all about. But what replication strategy to use?
Do I set up push partners, pull partners, or both?

If I was using DNS instead of WINS, since DNS supports push replication
only, I'd have each BDC be the primary DNS server for each of their networks
and have the PDC be the secondary. DNS already knows how to query down the
DNS tree for records outside of its own domain, so it already knows how to
locate and query another branch's DNS server for records. WINS doesn't know
how to do that, not having nodes or domains. So I'd have to have each
branch pull the other branch's records from the same server it pushed its
own records to, or alternately, have the PDC push other office's records out
itself. I prefer push to pull myself, since changes only get copied when
needed, but changes might not be replicating fast enough.

Which is better? Or use both?

I currently do this, all using push partners:

[branch1] -> [pdc]
....repeat for each BDC, and then:

[pdc] -> [branch2]
-> [branch3]
-> [mainbdc]
-> [branch1]

Machines don't move around all that much. Even the home office machines
usually get the same IP address each time they tunnel in. The users don't
move around too much either, and if they did I usually set up an account for
use in each branch they work in (roving profiles don't work over 128k).
Also, all inter-office conenctivity is between the branch and the main
office, except for remote administration where I might want to administer
one branch's server from another branch.