Minimum speed to replicate from site to site

  • Thread starter Thread starter Guest
  • Start date Start date
G

Guest

I was curious as to what the minimum speed would be to enable site to site
replication of a rather small ADS installation (under 300 objects)?

-Dan
 
In
snafu-ed said:
I was curious as to what the minimum speed would be to enable site to
site replication of a rather small ADS installation (under 300
objects)?

-Dan

I believe greater than 256 is recommended, but you can use a slower link. I
think 512 would be a better minimum, but honestly, you have 300 users and
there will be other types of traffic and would probably suggest at minimum a
T1. If it's an async line, such as dialup, it is suggested to create a
separate domain and use an SMTP link because of the line not being able to
handle the NTFRS and the DomainNC container.

If a client is logging on to a DC across a WAN, the minimum bandwidth for
normal client logon features is minimum 500kb by default. Below that many
GPO and other startup features don't run, other than security
implementations. I assume you will setup AD Sites and subnet objects in AD
to control logon traffic and replication traffic.

--
Ace

This posting is provided "AS-IS" with no warranties or guarantees and
confers no rights.

If you are having difficulty in reading or finding responses to your post,
instead of the website you are using, if I may suggest to use OEx (Outlook
Express or any other newsreader of your choosing), and configure a newsgroup
account, pointing to news.microsoft.com. This is a direct link into the
Microsoft Public Newsgroups, and it is FREE and DOES NOT require a Usenet
account with your ISP. With OEx, you can easily find your post, track
threads, cross-post, and sort by date, poster's name, watched threads or
subject.

Not sure how? It's easy:
How to Configure OEx for Internet News
http://support.microsoft.com/?id=171164

Ace Fekay, MCSE 2003 & 2000, MCSA 2003 & 2000, MCSE+I, MCT, MVP
Microsoft MVP - Windows Server Directory Services
Microsoft Certified Trainer
Assimilation Imminent. Resistance is Futile.
Infinite Diversities in Infinite Combinations.
=================================
 
We currently have a 256k/128k Frame Relay PVC line going into the new site.
The site itself only has 12 people there and 7 systems. I will be placing a
Domain Controller onsite so that the only thing that we should have to worry
about is cross site replication (at least from what I'm thinking). We aren't
using GPO's for anything more than security control and account managment.
So I'm not pushing packages and updates out through it.

I was under the impression that I could set cost and scheduling in a way
that would allow us to minimize the cross site replication of the entire AD
database. Also we don't have 300 users, I was counting the users, groups,
shared printers, etc....

So if I'm correctly understanding your response then the above situation
would not be too overburdened if I can sync the replication during non-peak
times?

-Dan
 
In
snafu-ed said:
We currently have a 256k/128k Frame Relay PVC line going into the new
site. The site itself only has 12 people there and 7 systems. I will
be placing a Domain Controller onsite so that the only thing that we
should have to worry about is cross site replication (at least from
what I'm thinking). We aren't using GPO's for anything more than
security control and account managment. So I'm not pushing packages
and updates out through it.

I was under the impression that I could set cost and scheduling in a
way that would allow us to minimize the cross site replication of the
entire AD database. Also we don't have 300 users, I was counting the
users, groups, shared printers, etc....

So if I'm correctly understanding your response then the above
situation would not be too overburdened if I can sync the replication
during non-peak times?

-Dan

Hi Dan,

I've never been fond of Frame Relay because of it's inherit latency. RPC is
not too tolerant when it comes to latency. If I understand correctly,
(correct me if I'm wrong), that 256/128 means 256 down and 128 up, then that
makes it an async line. Not too sure how RPC will handle that and if my
guess is right, you may see replication issues during peak (or other) times.
To control replication traffic, (and control logons to only their respective
DC), and not sure if you were aware of it, you *must *create AD Sites, and
Subnet objects that you would associate with the appropriate site. Then on
the IP link, you would schedule when replication is allowed and for how
long, if you want to allow replication traffic only at night.

If you have Exchange, that magnifies site to site traffic *considerably*
because of increased Outlook/Exchange traffic which can't be governed by
Site link replication scheduling.

Ace
 
No, the 256k/128k means that there is a 256k burst speed and a 128k dedicated
PVC circuit. Believe it or not until just recently we operated many of our
sites at 64k/32k connections. They were running stand alone domains but
still the WAN traffic was minimal. We do use exchange at the main site and
we have the existing stand alone sites setup to run in cached mode and
workstations attach to it using synced passwords at both the local sites and
the main site.

As far as the sites and services go I have already laid things out a little
on a test server and it is as follows.
192.168.x.x subnet for main location
192.168.x.y subnet for second site

Site A is assocated with the .x subnet
Site B is associated with the .y subnet

There is a site link from A to B that I will configure depending on traffic
patterns that I still need to determine. I will set replication costs as
well. My plan will be to schedule replication for off hours.


-Dan
 
In
snafu-ed said:
No, the 256k/128k means that there is a 256k burst speed and a 128k
dedicated PVC circuit.

I see.
Believe it or not until just recently we
operated many of our sites at 64k/32k connections. They were running
stand alone domains but still the WAN traffic was minimal.

Is this one forest?

Standalone domains (separate domains) that are either externally trusted or
part of the existing forest only has minimal text based traffic. Once you've
incorporated all the domains into one domain in your forest, it will consume
bandwidth, hence the higher bandwidth requirements.
We do use
exchange at the main site and we have the existing stand alone sites
setup to run in cached mode and workstations attach to it using
synced passwords at both the local sites and the main site.

Cached mode in Outlook 2003 is a nice feature, but I have some reservese on
it. With 2003 DCs and Sites, you can use Universal Group Membership Caching
on the GCs at each Site once you've implemented your Sites to lessen logon
authentication.
As far as the sites and services go I have already laid things out a
little on a test server and it is as follows.
192.168.x.x subnet for main location
192.168.x.y subnet for second site

You mean:
192.168.x.0
and
192.168.y.0
Site A is assocated with the .x subnet
Site B is associated with the .y subnet

There is a site link from A to B that I will configure depending on
traffic patterns that I still need to determine. I will set
replication costs as well. My plan will be to schedule replication
for off hours.

-Dan


Are you planning on having one domain or keeping multiple domains, one at
each site?

Ace
 
It depends entirely on how much churn you have (changes to be replicated) and
how much latency you are willing to accept. I have replicated domains with 100k
users across 64k lines before. It just eats it all up and takes a while.

joe
 
Back
Top