Distrubted NT/Win2k environment problem?

[jacob200]

I need some advice. We originally had one big flat network(single subnet). A bunch of switches daisy-chained to each other. Recently I have broken the network up into several smaller segments using Cisco 6506 switchs(each switch is a separate subnet). I did this with all the enviroments to make them easier to manage and troubleshoot. It also gives us more of a distributed services configuration. The VMS and Unix guys loved this. On the NT/Win2k side there is some debate. Actually with just one guy. It is our Directory Services administrator. We have approximately 265 servers on our internal network. The DS admin would prefer that he just keep his boxes on two different switches that are daisy-chanined together instead of two separate switche/subnet. The cons to this are that it will still be hard to troubleshoot when there are problems and it doesn't seem scalable. We seem to just keep growing and growing. This will continue to be an enormus NT/Win2k server segment/lan. Also, what effects one switch will affect the other since it is the same wire. I wanted to divide up my services, for example, one DNS server on one switch and one DNS server on another. The same would go for other services, web, peoplesoft, etc. Now I know there can only be one PDC emulator and there is not an automatic failover for that service. Also our High Availability clusters will have to stay on one switch/subnet until we get a .DOT-NET solution implemented. The PDC emulator, Exchange 5.5 and the clusters still need Netbios I believe. So while there are a few exceptions to every rule I still believe the separate switch/subnet scenario far out weighs the daisy-chained single subnet/switch scenario. The DS Admin insists that the separate subnet/switch scenario buys him nothing since there is only one PDC emulator and the current cluster exceptions. He also insists that the extra but minimal administration to have multiple Wins Servers, DNS servers, etc. is too much trouble.
Can someone give me some advice on which would be better in their opinion? Also, isn't the MS DOT-NET direction more a distributed configuration anyway that would embrace the separate switch/subnet configuration? Please help me to see the pros/cons and clear up any misconceptions I may have.

Thank you,

 

[GiaBetiu]

Well,...

The DCs have to serve the clients! Not the Administrator (DS Admin...).
I would use in every subnet a DC. The DC must be closer to the clients.
His job is to create subnets, sites. And configure them.
Indeed require work, but that's our job.
I don't know what is the componence of your OSes. I suppose that you have all kind (here the need of the PDC emulator).
But, anyway as a rule I would add also a Global Catalog and DNS to each of those subnets. Of course, it depends on traffic, and is a matter of sizing.
What your DS admin has to do he should know very well.
And your answer to him should be: Domain Controllers must serve the clients, they should be close to them, you will not have authentication traffic between switches, it will assure fault tolerance (if some switches will go down), and having defined sites then your computers will be part of that site (they will know to authenticate to the DCs from same site. For this don't forget your W9x clients need DS Client installed).
I hope that will help you. If not, we can make a nice document with pros and cons (but I can see just pros for your proposal). Gia Betiu
giabetiu@chello.nl
Computer Eng. CNE 4, CNE 5, soon MCSE2k