WAN clarification.... Part II

[ScottMac]

In this context, no, it really doesn't relate. At the source and destination device, the information passes layer-to-layer for whatever processing that layer is to accomplish. When in transit from one place to another, Telnet doesn't care if it's on Ethernet or Token-Ring...it can go over either or both (in turn)...it's an application...so deeply encapsulated that whatever's happening on the outside of the onion is of no concern.

An intermediate router would be both a destination and source device, it does fiddle with the MAC address between the gozinta port and the gozoutta port. But that same router just knows it's putting the data into some layer-one media (as Garion says " It could be a carrier pigeon" (assuming a high enough time-out...see the RFC)...to Layer two, it's hits the PHY and it's gone...if the layer one media has some amplifiers, repeaters, or layer-one-style switch...it's all just wire/media and and the layer two processes don't care.

A layer two bridge (or switch) doesn't look at or care about anything at layer three or above....nada. It doesn't care about layer one..there are a number of commercial-grade switches that easily translate from one media to another (Xylan/Alcatel, Cisco, NorTel...etc). Layer two only looks at layer two stuff. Layer three only looks at layer three stuff, layer one only looks at layer one stuff. Stuff will move up & down the stack, but none of the layers asks " What do you think you're gonna do with this stuff when I give it to you?" The stuff gets passed, and the processes at the next layer do with it whatever they're s'posed to do with it.

That's the way I remember it....

FWIW

Scott

 

[Nutz]

I was looking at more of from an encapsulation and framing standpoint--such that DL Layer would be touching the info coming up from the PHY or down from the NET.

 

[xyyz]

okay i've read these over and over

and I regret to say that I'm even more confused.



I don't understand how you can have a p-2-p network and do a traceroute and get multiple hops... to me this still seems contradictory to the concept of p-2-p

 

[ScottMac]

well, let's try it the other way around. Post your understanding of how it works, and we can take it from that point.

Point-to-point is from your place to the ISP (or another company location). That will show up as two hops. If you were talking from one company location to another, and your company had the most basic LAN/WAN setup (one switch at each place, one router at each place that ties to a P2P T1), and you did a traceroute from one pc at loc#1 to a PC in loc#2, your traceroute would show two hops, one for each router (PC->switch->router->T1->router->switch->PC).

If you logged on to the near router and did a traceroute, it would show one hop (router->T1->router)...each router demarcs a hop. Hops are bracketed by routers. There's a router at each end of a hop. All of these "hops" are Layer Three only...traceroute only looks at layer three.

Any other hardware at layer one for increasing the distance or managing the traffic doesn't count as anything except "WIRE."...It's only media.

Any other hardware at layer two (bridges for the most part) only exist to limit the collision domain, and (as far as traceroute is concerned) is still just "WIRE with latency." L2 Switches are essentially just fast, multi-port bridges...no effect on the traffic (except some latency).

Anything messing with the data at layer three is likely to be a router of some sort, and counts as one end of the hop. If you have a router with two Ethernet ports, and you ping from one Ethernet to the other, you'll get a traceroute report of one hop.


Once it hits the ISP, the ISP is gonna send it to wherever it's going through a system of routers.

I wish Anand had a White Board in here somewhere....picture make it a little easier......

Anyway..YOU post it the way you understand it...maybe we can work it out from there.....


Happy New Year

Scott

 

[Santa]

xyyz do you have access to a router that has Frame Relay's setup on them?

looking through the config file should help you build a picture in your mind better.

DLCI routing is the way I try to envision frame relay. It is routing but at a differnt layer.

As Scotty has reitterated that you should only see hops when a router actually gets ahold of the packet to "route" it somewhere else. You should only see many hops if you do not have a point to point or frame relay.

Frame relays create virtual point to points through the cloud.

What perhaps is confusing you is how someone brought up that you will be going through multiple "hops/medium" even though you are point to point yet you do not see these in your trace.

The trace is not picking up the extra media or "hops" because traceroute is not built to do so. It is only built to report back when a decrementation of the TTL and this is done by a router to router communication not a router to (DLCI to DLCI) to router.

The (DLCI to DLCI) does not register on the trace and everything within it does not register because it is frame relay and or point to point.

BTW: the (DLCI to DLCI) can also be depicted as (Cloud)

Your trace is of the internet and it is not point to point or frame so it must go through as many routers as it takes to reach the final destination. Since you do not know the exact location of the final destination you will get bounced from one gateway to the next till you end up at your destination.

 

[xyyz]

okay... sorry for not posting...

actually, I bet many of you are relieved...

Santa,

I have done a frame-relay lab... but it didn't help too much... maybe if I play with the advanced lab setup...

anyways... i'm still in vacation mode... so once that passes i'll post my understanding like ScottMac suggested.

 

[xyyz]

alright...

a few things before I explain how I see it...

can someone explain a gateway to me? from what I gather a gateway is an exit from your network... that lets you communicate with other networks... but I would like alot of details if you don't mind.

 

[Garion]

Actually, the concept of a gateway is a very simple one, and you have it pretty much exactly right - It's an exit point for your network where traffic is directed on towards it's ultimate destination.

- G

 

[xyyz]

<< Point-to-point is from your place to the ISP (or another company location). That will show up as two hops. If you were talking from one company location to another, and your company had the most basic LAN/WAN setup (one switch at each place, one router at each place that ties to a P2P T1), and you did a traceroute from one pc at loc#1 to a PC in loc#2, your traceroute would show two hops, one for each router (PC->switch->router->T1->router->switch->PC). >>



Okay, this much I understand... because a traceroute will show layer-3 devices along the path.

now you mentioned the following in an earlier post:



<< LA - T1,HDLC (User)--> LA (POP) -->router (decides which path to send the data)-->DACS (aggregates the T1 into a DS3,OC3, OC48,OC192,whatever)-->Long Lines (as part of an OC3,48,192)-->NYC (POP)-->DACS (splits the T1 back out) -->router (T1, HDLC)--->end user router (T1, HDLC to Ethernet)-->server. >>



When you say long line... do you litterally mean a wire that runs from a DACS (is the cisco 12000 router a DACS) in LA to a DACS in NY?

Is there a limit on the distance for a P-2-P link?



<< Anyway..YOU post it the way you understand it...maybe we can work it out from there..... >>



I'm working on how I see it and i'll put out a long explanation on how I see it, i've sorta been reading up on some stuff so hopefully that will allow me to be more clear in my explanation.

 

[Santa]

Theoretically the "Long Line" could be as long as it need be. Now the further you travel across fiber the longer the Latency or time it takes to get a response back. Networking is a two way street. A Request needs a Reply.

It would be in both yours and your Provider's best interests to go the shortest distance between points but sometimes this isn't possible. To get there at slowly is better than not getting there at all sometimes. Notice how I said "sometimes" If the latency is too bad then some applications complain. Don't get too hung up on the distance. (unless you want to be a provider architect worry more about what your total throughput is and the latency that you are getting and this will allow you to sleep better at night.

As long as throughput is fine and latency is within limits I could care less how they get packet from point A to B. I think Garion said it best with the..

It could be carrier pidgeons, just as long as it works. *grin*

I sometimes think that when our lines are slow that they forgot to feed our pidgeons personally