Trying to Understand Layer 2

Yes, that is true.

It depends, but not sure which is more common, but lots use ATM and FR, and I think some are using SMDS now. Local links use PPP. Not sure what the big boys are using on the backbones these days.

This is by far the most complete and best layed out protocol reference I have ever seen:

Your best reference (IMHO) for network stuff is Cisco:

#1 A good link was provided by another poster, so I'll skip this.

#2 Every layer on the OSI model requires that all the layers below it exist. Sometimes layers are combined based on protocol, but they still need to be there. So, to answer your question... IP runs on layers 3 and up. That means that for this communication to exist, layers 1 and 2 have to be present. This should help you understand a little better...

Let's say that PC1 has ip 192.168.0.1 and PC2 has IP 192.168.0.2. When PC1 wants to talk with PC2, it will first make an arp request - this is a request for the layer 2, mac address burned into the nic. No response - communication fails. Traffic is now sent to the destination mac address and destination IP address (both are needed for IP communications).

If PC1 and PC2 were on separate subnets, then PC1 would send an arp request for its default gateway. It would send the packet out with destination mac of the router and destination ip of the end workstation. The router on PC2 network would send an arp request for PC2s mac address and send the packet to it.

This is why a switch (layer 2 device) doesn't really care about IP address. It just discovers what mac addresses are on what ports. A router (layer 3 device) will need to forward traffic based on what the IP address is.

From your pc, ping an address that is on your subnet. Then, type in the command 'arp -a'. You will see the mac address of the other device. You should also see the mac address of all the devices you have communicated with recently, including your default gateway.

Hope that helps,

Jim

most of your protocols are in wide use, but not for core ISP networks since they are not cost effective at high speed.

SMDS is pretty dead - no one makes the telco switches any more - and it topped out at 25 to 50 Mbps

most of the core ISP network at work uses packet over sonet across fibre / lambdas or SDH (STM-16 @ 2.4 Gbps, or STM-64 @ 9.8 G). also seeing Ethernet @ 10G as well.

a lot of the newer feed networks are moving to Ethernet on lambda or over SDH.

First of all, let me thank each and everyone one of you for your responses they were very helpful and I think I am finally understanding it. Let me also say that I believe I understand Local Area Network running Ethernet and most of my interest was how a company would connect into a private network (using a T1) in order to receive some type of service.

I was hoping I could reiterate what I have learned and someone could confirm that I have it correct.

When someone says they are running an ATM backbone as opposed to an IP Backbone the differentiating factor is the physical equipment within the core. For example, someone running an ATM backbone would have just ATM switches in the core as opposed to someone running an IP Backbone they would be using layer 3 switches or routers to move traffic around. With the IP backbone, they layer 2 protocol could be Frame Relay or ATM, but the router is really making the decision about where the traffic should be routed. Does that sound correct? If the above is true, I have two questions and then I promise I will go away forever:

#1- The internet would be considered an IP backbone, correct?

#2- If you are strictly an ATM backbone, you route traffic around based on something in the ATM header and once it arrives and the edge device on the backbone it would strip off the ATM header and read the underlying protocol. But as it packet travels around the backbone, it would only look at the ATM header to know where to be routed.

Thanks again all!!!

Generally speaking, backbone would refer to the layer 2 infrastructure. In this case, ATM. IP would be the protocol that they are using over the backbone. You can have an ATM backbone and have IP be your primary protocol. To say that a network has an IP backbone, would mean that they are routing IP traffic regardless of layer 2 techonology.

That would be correct.

Wrong. Routers make decisions based on layer 3 information and using their routing tables. That said, it will ultimately pass it based on layer 2 information as was given in my arp example for ethernet.

Let's say that I see a packet destined for device 192.168.0.10. I need to figure out how to forward that packet. My routing tables tell me to send it to 192.168.50.1. At that point, I need to find the means of connecting via layer 2 to that device. In ethernet, it is arp, in Frame-Relay it is lmi, in point to point connections it is PPP or HDLC. I have not worked with (or studied) ATM, so I can't speak for the process of discovering devices at layer 2 there. But, the idea should be the same.

So, it will forward the frame based on layer 2 headers, but forward the packet based on layer 3 headers.

Hope that helps,

Jim

Sort of - but real networks tend not to be such a "pure" design.

my company has an underlying network with a mix of SDH and lambda / fibre.

then we have a 2.4 Gbps ATM network (which supports Frame relay) in parallel with that is an IP / MPLS backbone. and various other bits, like TDM voice.

but the transport get "mixed" - we still have a lot of Frame / ATM based router networks And the ATM gets used as a access link for high speed Internet access, consumer broadband "plumbing" and for corporate private DSL connections.

you need to remember that most companies that provide telecomms are not a "pure" voice, data or internet etc supplier - they do a bunch of different services.

And most have built up infrastructure over years or decades, so there is inertia built into the services depending on the installed base. new networks may go on IP / MPLS, some older ones based on Frame may migrate, but others operate as a hybrid, and some keep adding to the existing Frame and ignore IP.

If the above is

Yes - although inside a telco it may run across other things - most popular right now is MPLS

backbone is just a word to contrast against access - the bits connected to customers. No-one has a single backbone type - even a big telco will have links to other countries where they rent plumbing rather than use their own fibre etc.

so we have a DWDM "backbone" with 5 or more type of DWDM system / kit, (but with some rented dark fibre, lambdas and other telco SDH glueing bits onto it) with an SDH backbone on top, and an MPLS backbone partly directly on fibre, lambdas and on SDH, and an ATM network in a similar way, and vision circuit network"backbone" over SDH etc.

And despite all the stuff from manufacturers there isnt one dominant all singing best protocol stack or technology - so you end up with a mix to suit what you want to do.