Using the same DLCI throughout the network

totts

I'm having a bit of difficulty unravelling this one, could someone please try and provide a better explaination. I'm reading ICND Examination Guide.

From what I understand, ISPs can assign different DLCIs to different Virtual Circuits, which would make perfect sence so that they've all got a unique address, OR the provider can assign the same DLCI to all VCs on the network.

If all the Virtual Circuits on a network are using the same DLCI, how does a Frame Relay switch know where to forward frames to and when the frame arrives at the destination, how does the router know who sent the frame... everyone's got the same DLCI right??

Netstudent

I think what the author meant was you can use the same DLCI's in the network, but the DLCI's must still maintain local significance. If you look at the frame-relay techlab here, you will see that a total of 6 DLCI's are being used for 3 PVC's. Each PVC uses a set of completely different DLCI's . You could also do the same thing with only 3 DLCI's.

From techlab:

---

Say 2501A had a global DLCI of 100

2501B had a global DLCI of 200

2501C had a global of 300.

---

Then 2501A would have 2 local DLCI's and they would be 200 and 300. Both defining a separate PVC.

Then on 2501B you would have local DLCI;s 100 and 300

2501C would have local DLCI's 100 and 200.

So you used the same 3 DLCI's but they are still locally significant.


Thats my interpretation of what the author is trying to say. I could be wrong though.
If you don't see what i'm getting at, let me know and I will modify the techlab frame routes to what they would be if you were conserving DLCI's using the method I have described.

jediknight

The key thing to remember is that DLCIs are Locally Significant. This is due to the fact that the DLCI is not advertised and the router would never know that the router on the other end is using the same DLCI.

Basically by mapping the DLCI to a IP Address (which is unique), the router can get the frame to it's destination. The Frame Relay swithes use their own lookup tables to get the frame to it's destination once in the ISPs network.

This is how I understood it...

Netstudent

also local DLCI's are the actual DLCI's you see when you do a show frame pvc. Those are the DLCI's used to get to the other side.

Global DLCI's are the DLCI"s that other routers use to get to you. So if you are sitting at 2501A, your global DLCI for 2501A is the local DLCI at the other two sites.

totts

Thanks Netstudent. This has been doing my head in a little bit. I've been reading the same few pages all day!

Ok, so each router is assigned a unique 'global' DLCI which is 'locally' mapped on a remote router against that routers IP address, right?

So Router F for example may have a 'global' DLCI of 40 and an IP address of 172.16.16.2 so all the other routers on the network would 'locally' map 172.16.16.2 against DLCI 40?

totts

jediknight wrote:

The Frame Relay switches use their own lookup tables to get the frame to it's destination once in the ISPs network.

That makes all the difference knowing that the provider use their own lookup tables. Thanks again, netstudent for explaining the difference between global and local DLCIs. Hopefully I can put this one to bed, unless someone comes along and drops another bombshell!!!

Netstudent

totts wrote:

Thanks Netstudent. This has been doing my head in a little bit. I've been reading the same few pages all day!

Ok, so each router is assigned a unique 'global' DLCI which is 'locally' mapped on a remote router against that routers IP address, right?

So Router F for example may have a 'global' DLCI of 40 and an IP address of 172.16.16.2 so all the other routers on the network would 'locally' map 172.16.16.2 against DLCI 40?

yes theoretically that is correct.

totts

Thanks, I'm happy again, until the next page eh

dtlokee

The concept that the DLCI is "locally siginificant" comes down to the fact it must be unique on the access-link, that is unique on the interface. If you had a router with 2 serial interfaces on it both using frame-relay encapsulation, they could both be using the same DLCIs on both interfaces but the PVC's could have totally different remote ends. When frame is sent, it has a DLCI on it to identify the PVC, when the frame switch receives the frame it then converts the local DLCI to a global DLCI and sends it on it's way (this connection is typically an ATM connection to other ATM switches) At the receiving end the frame switch receives the frame with the global DLCI, then converts it to the local DLCI and sends it out the interface to the customer.

totts

So has every serial interface/access link been assigned a unique 'global' DLCI, for example DLCI 70 on Router Bs serial interface and every other router on the network terms DLCI 70 as the 'local address' for Router Bs access link?

rjbarlow

totts wrote:

So has every serial interface/access link been assigned a unique 'global' DLCI, for example DLCI 70 on Router Bs serial interface and every other router on the network terms DLCI 70 as the 'local address' for Router Bs access link?

Well, think to "global DLCI" like an address, that is unique onto the whole network for all the DTEs who want to refer to a same DTE.
For example: routerB, C, D, E and F refer to their own PVC to reach routerA, by means of DLCI 75.
75 is the Global DLCI for routerA. But this is not a rule, is only a "style" that the provider could choose or not,, because routerB, C, D, E, F, could all refer to routerA through different DLCIs.
Have been clear I?

totts

Yes you have been clear rjbarlow and thanks for the explanation. Its coming together now, but I do have another question...

If global addressing is an optional extension of LMI and a Frame Relay network works fine without it, then why would we implement global addressing at all?

totts

I think I've found the answer to my own question... The advantages of using a global addressing scheme is two-fold... Firstly, if access links are assigned their own unique DLCIs then the network is more readily understood. Secondly, ARP can be used on a network with unique DLCIs or L2 addressing.

rjbarlow

OK, however I-ARP is not related to the global addressing scheme, it works even if You aren't in a global F.R. addressing scheme.

totts

Would you not need a unique L2 address in order for inverse ARP to work, ie a unique DLCI and therefore global addressing scheme?

rjbarlow

totts wrote:

Would you not need a unique L2 address in order for inverse ARP to work, ie a unique DLCI and therefore global addressing scheme?

Not, You could also make a manual IP address/DLCI mapping and all would work, even if not there is a global addressing scheme for DLCIs. F.R. switches supply to deliver frames to the correct destination by means of their internal map.
Is a fact that different DTEs in F.R. can reach the same DTE, everyone using a different DLCI.

totts

OK, so I think we have established now that the real benefit of a global addressing scheme is that its intuitive to look at right?

So in fact, every VC can be assigned the same DLCI by the ISP
and as long as we've configured an IP address on a physical or subinterface on our routers, the Frame Relay switches, due to their own mapping and our IP addresses will deliver the frames to the correct DLCIs right?

rjbarlow

totts wrote:

OK, so I think we have established now that the real benefit of a global addressing scheme is that its intuitive to look at right?

Yes.

totts wrote:

So in fact, every VC can be assigned the same DLCI by the ISP
and as long as we've configured an IP address on a physical or subinterface on our routers, the Frame Relay switches, due to their own mapping and our IP addresses will deliver the frames to the correct DLCIs right?

So, the main rule is to perform the following goal by the provider when it assign DLCIs to PVCs:
that DLCIs known by LMI messages from the attached switch F.R. must not to be replied on the same DTE (read as: "one DTE does not must refer to differnt PVCs by means of the same DLCI");
I guess Your trouble is to understand that DLCIs are not parameters belonging directly to the DTEs (like an IP address or a MAC address), but they refer mainly to the PVCs in order to make F.R. switches able to deliver frames correctly within the provider network.
Switches F.R. ever worry about IP addresses because they are L2 devices.
F.R. addressing is a topic not at all straightforward, but if You read well the Odom's book or Sybex, it's covered pretty good; also using a simulator helped much me to understand how F.R. addressing works, even if it was much bugged and limited.