Site Cost Links Question

genXrcist · 07-09-2009, 11:39 PM

Let's say you have two sites, with two connections between the two. One connection is T1 and the other is 128KB. You setup your two site-links between the two, one with a cost of 100 for the T1 and the other with a cost of 200 for the 128KB connection.

WHen it comes to the actual passing of data, how do the DC's know that the site-link of 100 is attached to the T1? In other words, when they're replicating data accross the 100 cost site-link, how does it know to use the T1 connection?

astorrs · 07-09-2009, 11:59 PM

Quote:

Originally Posted by genXrcist

Let's say you have two sites, with two connections between the two. One connection is T1 and the other is 128KB. You setup your two site-links between the two, one with a cost of 100 for the T1 and the other with a cost of 200 for the 128KB connection.

WHen it comes to the actual passing of data, how do the DC's know that the site-link of 100 is attached to the T1? In other words, when they're replicating data accross the 100 cost site-link, how does it know to use the T1 connection?

The system uses it to calculate the "cost" of the link. Therefore in your example the link with a cost of 100 is less than the link with a cost of 200. Or think of it this way - it costs $200 to send traffic down the 128K link, but only $100 to send it down the T1.

genXrcist · 07-10-2009, 02:03 AM

Right, but what tells the DC that the $100 T1 is the pipe it's actually using? I mean, when it sends out the packets on the wire, how does it know it's actually using the T1?

ElwoodBlues · 07-10-2009, 02:41 AM

I think that you typically would not have 2 site links between only 2 sites unless 1 is a demand dial. Keep in mind that these are only logical objects.

Psoasman · 07-10-2009, 06:04 AM

AD creates a topology for the forest to determine what the best path to replicate data. The KCC - Knowledge consistency checker- runs on all DC's and it will build the topology automatically for you.
AD will figure out the best path, but you can set the cost yourself. The higher the cost for the path, the lower the preference that KCC has for that site link.
For example, if you have 3 domains and 2 are connected by T1's and the other by a 128 KB connection, replication would probably eat up the bandwidth of the 128 connection. You can manually increase the cost of the 128 site link, so the KCC will only use it last. I think the cost must be higher than the other connections.
This will force the replication over the site links with the lower cost.
You also have the option of scheduling the replication. You could schedule the replication for the 128 site link during slow times, when users aren't there.

I have been studying this material this week and I think I'm getting a handle on it...finally

genXrcist · 07-10-2009, 01:17 PM

Quote:

Originally Posted by ElwoodBlues

I think that you typically would not have 2 site links between only 2 sites unless 1 is a demand dial. Keep in mind that these are only logical objects.

This is my point, and what I'm struggling with. How do the DC's know which logical object corresponds to which physical line connection?

Quote:

Originally Posted by Psoasman

AD creates a topology for the forest to determine what the best path to replicate data. The KCC - Knowledge consistency checker- runs on all DC's and it will build the topology automatically for you.
AD will figure out the best path, but you can set the cost yourself. The higher the cost for the path, the lower the preference that KCC has for that site link.
For example, if you have 3 domains and 2 are connected by T1's and the other by a 128 KB connection, replication would probably eat up the bandwidth of the 128 connection. You can manually increase the cost of the 128 site link, so the KCC will only use it last. I think the cost must be higher than the other connections.
This will force the replication over the site links with the lower cost.
You also have the option of scheduling the replication. You could schedule the replication for the 128 site link during slow times, when users aren't there.

I have been studying this material this week and I think I'm getting a handle on it...finally

I agree with this 100% and I believe I understand what the function of the site-link and their costs are, as well as how the KCC utilizes them for replication topology, but I still don't get how it knows which physical line to place the replication packets on? I mean, how does it know that one site-link w/cost of 100 is the T1?

I don't see anything in Sites & Services that allows you to designate the speed of the site-link, just the cost.

Let me put it like this, say you have T1 access from Qwest and 56KB access from a local ISP. You create the 2 site links, one with a higher cost than the other. When the DC goes to replicate, how does it know that the higher cost site-link means 'send the packets on the local ISP line'?

NetAdmin2436 · 07-10-2009, 01:59 PM

I think I understand your question....and here's what I think (someone correct me if I'm wrong).

I don't think Windows can tell the difference in the two physical lines, if both site links are assigned to the same site (subnet) and/or if both of the Internet lines are assigned to the same subnet for their LAN side. I think it's more going to depend on how you have your firewall/router and connections configured. Typically you will have a firewall/router that can handle the two Internet connections and make any fail over decisions. I think the domain controller just sends the replication traffic to the router and says "here, send this to <ip address>". Then the router would make the decision which physical line to use. So in essence, I don't think creating multiple site links in Active Directory Sites and Services from site A to site B really does anything in this case if both internet connections are on the same subnet.

dynamik · 07-10-2009, 02:36 PM

Yea, I was thinking about this since we talked about it, Gen, and I think the underlying network infrastructure is transparent to Windows.

genXrcist · 07-10-2009, 05:53 PM

Quote:

Originally Posted by NetAdmin2436

I think I understand your question....and here's what I think (someone correct me if I'm wrong).

I don't think Windows can tell the difference in the two physical lines, if both site links are assigned to the same site (subnet) and/or if both of the Internet lines are assigned to the same subnet for their LAN side. I think it's more going to depend on how you have your firewall/router and connections configured. Typically you will have a firewall/router that can handle the two Internet connections and make any fail over decisions. I think the domain controller just sends the replication traffic to the router and says "here, send this to <ip address>". Then the router would make the decision which physical line to use. So in essence, I don't think creating multiple site links in Active Directory Sites and Services from site A to site B really does anything in this case if both internet connections are on the same subnet.

Ahhhh... now this makes sense. It bases it on the Subnet the P2P sitelinks are on...ok, now I understand how the App layer is getting down to the Network layer. Thanks NetAdmin, you're a genius!!!

Quote:

Originally Posted by dynamik

Yea, I was thinking about this since we talked about it, Gen, and I think the underlying network infrastructure is transparent to Windows.

Yep, and it was this underlying transparency that was throwing me off. Now that I understand it's based on the site subnet, it makes total sense.

Thanks Everyone!

NetAdmin2436 · 07-10-2009, 08:21 PM

Quote:

Originally Posted by genXrcist

Thanks NetAdmin, you're a genius!!!

That's what you think. I can show you some of my Microsoft exam scores that say otherwise