Help with conceptual question on MPLS vs. ATM

Unanswered Question
May 15th, 2009

A network event speaker mentioned that she always saw MPLS as a "replacement to ATM" technology.

Does that make sense?

I know how MPLS can be used, but so far I haven't seen that related to replacement of ATM.

If the statement makes sense can elaborate? Thanks.

I have this problem too.
0 votes
  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 5 (1 ratings)
cisco_lad2004 Sat, 05/16/2009 - 02:06

The only similarity that I can think of is that with ATM the IP packet is encapsulated and cells are switched from one ATM box to the other based on VPI/VCI parameters. and MPLS switches packets based on both do not look at the IP packet while its being switched (not including PHP).


Joseph W. Doherty Sat, 05/16/2009 - 04:43

It could be seen as a replacement in the sense that often WAN cloud providers, if they offer both, seem to push MPLS.

It could also be seen as a replacement for its differences when dealing especially with data. For example, multipoint vs. VCs, no "cell tax", QoS across a path, etc.

Having used both on a global scale, the only weakness I've seen with MPLS vs. ATM (or frame-relay) is often the inflexible QoS models provided by vendors for cloud egress (which isn't a technical limitation).

One advanage vendors seem to push is MPLS will cost less than ATM (similar to ATM will cost less than frame-relay). My experience has been, this might be true for "raw" bandwidth cost, but due to the common limitations of provider MPLS QoS, you'll need to buy more bandwidth if you have finely tuned QoS for ATM (or frame-relay).

marikakis Sat, 05/16/2009 - 05:39

We once had a methematician who told us "mathematics are all the same"! We had algebra, analysis, matrix theory, complex variables etc. What did he mean?

Knowledge goes from the specific to the abstract/general and vice versa. You study specifics and then sit back and think about what they all have in common. This is a mind procedure that classifies things in order to understand things better. For example, when you find a "new" specific, you sit back and think what it has in common with other "general" theories and be able to predict part of its behavior with less effort. Classification is a procedure that aids research, understanding and memory.

In this case, MPLS is not exactly the same thing as ATM, but it has similarities. For those who knew ATM, it was easy to understand how MPLS lookup and forwarding works. I think this is the "biggest" similarity, i.e. exact lookups versus IP lookups, "labels" with local significance and a "label" mapping mechanism. In some sense you could say MPLS looks more like FR because of the lack of requirement for chopping packets into cells. The ATM segmentation of packets into small cells is a controversial issue. QoS can be better when possibility for interleaving exists, but poses challenges for packet reassembly and causes big issues with the bandwidth cost (header overheds and padding overheads especially for "small" packets) as it has already been mentioned.

Network device enginnering and design face trade-offs all the time. There exist techniques to optimally handle sub-cases in different manners, which makes hard to decide what to do and the more techiques incorporated, the more complicated the design becomes, which means error-prone or even suboptimal average performance. The story from FR to ATM to MPLS sometimes looks to me as if going back in circles. I guess one day the whole design might stabilize on commonly agreed design principles. But then again, as speeds grow and new challenges have to be faced, reconsiderations might take place to find the optimal under the new circumstances. Everything changes and in some ways stays the same in the sense that revolutionary techniques are rare and the optimal combination of known techniques at each point in time is usually the question to be answered by manufacturers.

Giuseppe Larosa Sat, 05/16/2009 - 11:32

Hello Marlon,

from an historical point of view MPLS has been introduced first in ATM networks, (using cell mode) and at the beginning was an attempt to have a better use of ATM networks to carry IP services.

Other vendors had proposed alternate frameworks like newbridge MPOA.

Then MPLS has become a sort of "logical ATM" that can run on different technologies.

And the economic aspect is also important:

a 2,5 Gbps POS interface used for MPLS and IP is far less expensive then a 2,5 Gbps ATM interface.

From a technical point of view ATM can support more features: if you compare Call admission control on ATM PVCs with CAC on MPLS TE tunnels you will see that ATM PVCs cannot send traffic much more than contracted SCR, and instead MPLS TE CAC works only as an administrative parameter used during setup but then if you put 40 Mbps of traffic over a tunnel with 2 Mbps of declared BW none stops this!

So yes MPLS can be seen as a replacement for ATM inside WAN providers and ISPs.

Hope to help



This Discussion