Does Cisco consider ATM only suitable for LAN, not WANs

The practice exam was written by an idiot. ATM was never intended to be a LAN protocol (it uses a small, constant cell size for very fast switching, to facilitate aggregating thousands of PVCs on a circuit) and LANE was the biggest nightmare of Cisco's, 3com and Lucent's existence.

ATM is the primary layer-2 protocol for the internet backbone. It is the primary layer-2 protocol for all WAN connections (that aren't PTP HDLC). In fact, if you have a frame-relay connection, odds are it is ATM as soon as it leaves the first switch.

LANE is dead (thank whatever higher power you believe in).

Jonathan

I'm interested, why was it so bad?

Dave.

Then why was ATM just announced to be removed from the CCIE written ?

ATM has seen its days and will start going away. The trend is to go towards GigE, even in the long haul/backbone.

I don't know if I'd agree with this being a 'stupid' trend. The high speed ether solutions are much more elegant, cleaner, simpler, and are more bandwidth efficient than ATM. Some commercial DSLAMs are now going GigE / dot1q instead of ATM on the backhaul side. I beleive that very soon the only place ATM will be found is on DSL subscriber loops. Many people feel there's no need any more for the hocus pocus, SAR/AAL that ATM incorporates; there are more BW efficient means now to offer QoS, and carriers like the idea of making more BW available for billable purposes.

It didn't work.

Firt off, you have to replace the ethernet NICs in your PCs (at about $400 per), then you have to run fiber everywhere (the base connection was an OC-3), then you had to start buying ATM servers for registration (BUS, LECS), and let us not forget the complexities involved here. Insted of a MAC address you use an NSAP (an example:

47.0091.8100.0000.1111.1111.1111.2222.2222.2222.00) Then, on top of all of this crap, you are now running IP.

Of course, your infrastructure costs just spiked through infinity as you started buying ATM switches and then wondering how you were going to make it all work together.

The real problem is just beginning, because none of the vendors knew how to make it work either.

Jonathan

Probably the same idiotic reason they removed ISDN.

Jonathan

ROFLMAO

Really?

Except for the ONS stuff, most provider's backbones are pure ATM running over OC-3s, 48s and 192s.

And what happens when you start aggregating a few hundred thousand customers on ethernet?

Jonathan

The biggest stupidity in this is that companies like Cisco are putting so much effor to create "ATM service like" form GigE that it becomes funny...

This statement was about WAN connections (read MetroE) and not about LAN. In LAN, ethernet was and is the top.

Hope the floor is clean. ;-)

I did not mean to imply that carrier networks, as they are today, do not deploy ATM. I meant to make a case that many people, including carriers, are considering the future of their networks, and the possibility that ATM may not be the fundamental building block of these future networks that it is today.

Perhaps you've read me wrong, or I could have been clearer in what I stated. Maybe the word 'trend' should read "future." Many believe that soon all information will be IP based. Or at least a majority of it. Some of these people also believe that when this happens, placing packets directly on fiber with DWDM (and related/similar transports) will reduce and/or eliminate the need for ATM. There are many papers that talk about this. There are many standards working groups discussing these issues. When discussing the all "optical internet," this idea is in the forefront.

I have installed DSLAMs that have already moved away from the tradditional ATM backside ('backside' being facing away from the DSL subscriber) to a GigE/dot1q model.

The fact remains that ATM solutions include more overhead than GigE solutions, and many organizations have recognized this.

So really, you can get up off the floor now.

Back to the original question, if I may.......

So why DID Cisco remove ATM from the exam? Surely not because they are "stupid." There must be some rationale for doing so, and I simply took a stab at offering up one possibility for consideration.

Be well.

-ja

PS

Here's an interesting article sort of related to the topic. Lots more out there, too in cyberspace....

Scroll down to white papers, and click on the Ethernet in the WAN pdf.

Cheers, again.

Jonathan, Thanks for letting me know.

Dave.

While I do not think ATM at and to the prem is going to be around forever, as a long-haul, mass-aggregator of PVCs, ATM is the bomb.

Since each cell is a constant size, my switches don't need to look at every single bit to determine when one chunk ends and another begins, so I can just chop at every 53-bytes.

Also, ATM is the only layer-2 (read that again) protocol that has inherent QoS built into from an initial design perspective (instead of Ethernet where they have been modifying it constantly to keep it alive since the 1970s).

So do I.

What does that have to do with ATM?

ATM is specifically a Layer-2 protocol, it doesn't care what layer-3 protocol is running on it.

Except for ATM machines (which still run SDLC (because they are bloody masochists), I see very little that will not be IP-based.

GigE is packet over Sonet... in a way...

But from the prem to the DSLAM, it is pure, good-old-fashioned ATM.

Yes, but if you are a service provider, your primary concern is aggregating as many PVCs as possible.

Doing this on Ethernet is nuts... Even if you do VLAN tunneling, you are still restricted to about 4000 Customers (that is very on the high side).

We have OC-192s carrying a hundred thousand PVCs (all DS0s).

Well, they are.

ISDN BRI is a core technology for backing up frame-relay circuits. Why did they remove that? Cuz they are dumb.

Jonathan

tradditional

aggregating

Maybe they see a "trend" to migrate to DSL for backup?

;-)