|
Home > Archive > 70-217 > September 2002 > Need help with Site Links
You are viewing an archived Text-only version of the thread.
To view this thread in it's original format and/or if you want to reply to
this thread please [click here]
| Author |
Need help with Site Links
|
|
| Tech Ranger 2002-09-19, 8:27 pm |
| I have read in 2 text books that different site links can be set up for different WAN links and relative costs can be configured to establish a preference for replication over less costly links. In theory this sounds great. What I am having difficulty with is the actual way that this is implemented. Neither book goes into that. Win2K doesn't ask you which link uses dial up and which uses a leased line. The OS does not know which link is which. Self-study can be difficult at times because there is no teacher to ask questions. Nonetheless, I am guessing that you accomplish this objective by fooling around with the subnets or the bridgehead servers, or I am just stupidly failing to see something obvious here. Can anybody clarify?
I hope I am asking my question clearly.
We have 2 sites which are connected physically by 2 WAN links with different costs. We configure 2 site links between the 2 sites. One is given a cost of 10 and the other 1. How does the OS know which site link object represents which physical WAN link? | |
|
| I did ask this question but noone really satified it! But, fortunately, I found it in 70-219 Sybex, Yes, MS dont have a clue which is which, they dont have a mechanism to recognize the links!
But Costs are actually used in mutilple sites environment (>3), eg
A - B - C
A - D - B,
Sorry, I cant express this in a more visualize fashion, but just imagine A-B-D are connected in a ring modern!
So, in order for C connect to D, they will determine the cheapest cost to get to D,
say,
BC=200
BA=70
BD=30
AD=50
So with path KCC would choose to use???
C-B-D=200+50
instead of
C-B-A-D=200+70+30
That's it!
But I keep wondering why MS keep talking about choosing Broadban and Dialup link using site cost?Can Any1 help?
Hope this helps! | |
| dwatts 2002-09-20, 4:40 am |
| Im not sure what led you down the track of thinking AD knows much about your network. But just to clear this up once and for all.
Sites exist so that AD can map services efficiently on a network. Sites are groups of IP subnets with fast connectivity. Site links connect sites together, and a cost is assigned to those links in order to persuade AD to take specific routes through that network. This is a manual process - meaning, you have to set up sites yourself, and you have to create costs associated with links (sure there is a default, but every link has the SAME default, rendering it useless.)
That is the extent of knowledge that AD has about the underlying physical network. It does not know nor care whether a link is a LAN connection or a dial-up. All it cares about is Sites and Link Costs. It might appear it knows more, but that will happen only because of the costs associated with links. If the costs have been assigned correctly, then AD will make intelligent decisions. If they have been assigned poorly, then AD will make dumb decisions. | |
| Tech Ranger 2002-09-20, 6:10 am |
| I understand your response. However, it doesn't answer the question. What I am saying is that in 2 seperate text books, one of which is an MS Press book, it is stated that you can have 2 sites connected to each other by more than one physical connection, and that you can establish site links asssociated with those physical connections. By assigning costs, you can encourage AD to choose the less costly path when it is available. My question boils down to this. If you tried to do what the books say, you would find yourself in a quandary because there doesn't seem to be a way of carrying that out. If I create 2 connections between 2 sites and 2 logical site link AD objects so that I can differentiate between the 2 methods of connectivity, AD has absolutely no idea about my intentions. How would it choose the physical connection to use? | |
| Zaraspook 2002-09-20, 6:59 am |
| quote:
How would it choose the physical connection to use?
The WAN links are the physical media used to transfer communication. The site links are a LOGICAL method of manipulating Windows to use the correct physical connections.
Think of in this way, a print-device is the actual physical printer. (The plastic, metal, and ink cartridge.) The printer is the logical device you configure in Windows. By using the Add Printer wizard, you can set up a printer numerous times corresponding to the same print-device in Windows for different purposes. You would use the logical printer's properties to configure each printer for each specific purpose. You use the same physical/logical relationship for site links. You use the properties of the site links you set up to configure cost, which manipulates Windows into using the preferred physical links. Hope this helps!
Here's the previous Quyda discussion:
site config... | |
| dwatts 2002-09-20, 7:52 am |
| --If you tried to do what the books say, you would find yourself in a quandary beause there doesn't seem to be a way of carrying that out. If I create 2 connections between 2 sites and 2 logical site link AD objects so that I can differentiate between the 2 methods of connectivity, AD has absolutely no idea about my intentions. How would it choose the physical connection to use?
Actually, I did answer your question. You are simply over-complicating it. Perhaps the wording in your books is poor.
I repeat: AD actually knows nothing about the topology of the network. It could not care less whether a link is via a WAN or via a dialup. It makes it decisions based purely on site link costs.
If a book you have suggests there is some other way other than by costs of telling AD which link to use then its wrong. The link it uses is arbitrary, based on assigned cost. | |
| jeff_j_black 2002-09-20, 10:10 am |
| The scenario using two sites with two links is a little confusing, it would be better if they illustrated it using three or more links. like 'spook and 'watts are saying, the site links you create are logical replication links that don't neccessarily represent physical connections, using two sites and two links makes it more difficult to understand.
code:
a
/ \
t-1 / \t-1
/ \
/ \
c---------b
56k
Say we want c to replicate with b. Well the connection between the two is a 56k dialup, not as efficient as the t-1 connections between c and a or a and b. If we make the costs of c-a and a-b add up to less than c-b, than those links will be selected. So say c-a has a cost of 40 and a-b has a cost of 40 and c-b has a cost of 100 then c-a (40) + a-b (40) = 80, which is less than the link c-b (100). AD replication does not know about the physical connections and what media they are on, just that you have created logical connections and by assigning a cost to those logical connections, given AD a means to choose the best site links for replication.
Good to see dwatts making it around to this side of the internet! | |
| dwatts 2002-09-20, 10:23 am |
| Oh, I get around --LOL-- | |
|
|
| Tech Ranger 2002-09-20, 6:26 pm |
| What I am understanding from the reponses is that 2 sites wouldn't have 2 site links configured between them because there would be no rhyme nor reason to it. They would essentially be redundant and couldn't be associated with any physical WAN hookups so as to prefer the use of one over the other by virtue of cost. It seems that site link costs have value in the context of multiple sites connected by multiple site links. The paths from any site to any other site through site links can be optimized by the KCC comparing the total costs of various paths.
I would propose the following, however. Let's assume that I indeed set up 2 sites connected by 2 WAN links of varying bandwidth. I want AD to prefer the connection with higher bandwidth for replication. But, I have a dial up for fault tolerance. Couldn't I configure 2 site links and use Bridgehead Server designations combined with cost configurations to force AD into associating a physical WAN connection with a logical site link? | |
| Zaraspook 2002-09-20, 7:16 pm |
| Lets take two sites - New York and Los Angeles. We have two T1 WAN links that connect the two sites. We must configure two site links in AD Sites and Services to represent these physical T1 WAN links. We name the first site link A and the second site link B. We double click site link A and access its properties. We assign it a cost of 200. We then double click site link B and access its properties. We assign it a cost of 500. Which replication path would Active Directory choose?
Active Directory would choose site link A because in this scenario it has a lower cost!
Again, lets take the same two sites - New York and Los Angeles. And again, we have two WAN links that connect the two sites. This time though, one is a 56k WAN link and the other is a T1 WAN link. We configure two site links in AD Sites and Services to represent both of these physical WAN links. We name the first site link, the 56k WAN link, A. We name the second site link, the T1 WAN link, B. We double click site link A and access its properties. We assign it a cost of 200. We then double click site link B and access its properties. We assign it a cost of 500. Which replication path would Active Directory choose?
Active Directory would choose site link A, even though it is a 56K WAN link and site link B is a T1 (much faster) WAN link.
The reason is because we configured site link A with a lower cost. This would be stupid to do under most normal circumstances, but illustrates the purpose of site links and how they can be configured and manipulated. The only prerequisite is that a site link for each WAN connection to be used be configured. The fact that it is a Fractional T1, full T1, T3, or plain old 56k connection makes no difference. The cost you apply, and only the cost you apply, determines the preferred replication paths and is, therefore, the method used to administer and control replication.
quote:
I want AD to prefer the connection with higher bandwidth for replication. But, I have a dial up for fault tolerance. Couldn't I configure 2 site links and use Bridgehead Server designations combined with cost configurations to force AD into associating a physical WAN connection with a logical site link?
Exactly! The whole purpose of cost is to give administrators the ability to manipulate AD to use the fastest WAN connections for replications. By manipulating cost, administrators are able to optimize replication. | |
| Tech Ranger 2002-09-20, 7:22 pm |
| quote:
We name the first site link, the 56k WAN link, A. We name the second site link, the T1 WAN link, B. We double click site link A and access its properties. We assign it a cost of 200. We then double click site link B and access its properties. We assign it a cost of 500. Which replication path would Active Directory choose?
Just by virtue of your naming convention, AD is not going to recognize your intention that 1 of these site links represents 1 of the speific WAN links, and the same for the other pairing. You have great intentions, but isn't AD clueless? | |
|
| OK let me demon the prob. in a more visualize fashion:-)
/------L1-56k----------\
A B
\------L2-T1----------/
OK, say we r configuring Sitelink in a smart way
L1=100
L2=50
OK, you say W2k would choose L2 right! That's nature but when you setup sitelink, MS doesn't have a Bind method,
eg, L1 -bind to-> Dial up
L2 -Bind to -> T1
So how can W2k choose 56 or T1?? I got really confused about it and wondering why I got MCP 217? Am I paper??? | |
| jeff_j_black 2002-09-20, 10:53 pm |
| You would have to use the fallback feature of the routers, as you are correct, there is no way to bind one link to a physical connection. That is why the best illustration is three or more connected sites. | |
| Zaraspook 2002-09-21, 9:58 am |
| quote:
Originally posted by Tech Ranger
Just by virtue of your naming convention, AD is not going to recognize your intention that 1 of these site links represents 1 of the speific WAN links, and the same for the other pairing. You have great intentions, but isn't AD clueless?
Right, AD is totally clueless, it is a tool just like the keyboard I used to type this sentence. Naming conventions are only for the benefit of we humans, computers understand numbers that's all. Just as I manipulate keys to form words, I can manipulate cost to control AD and optimize replication. |
|
|
|
|
