Hi Group,
I think labelling the ATM technology as dead would be an "overkill". ATM has
its advantages as well as disadvantages. Mere opinions don't change facts.
Following are some facts about ATM:

ATM has evolved as a stable connection oriented transport that currently
operates, ATM switch to ATM switch at up to OC-48 line rates. It also lends
itself ably to traffic engineering (prior to MPLS it was the only technology
that offered traffic engineering features). It delivers many advanced
features such as PVC creation from any ingress to any egress in a given ATM
backbone, sophisticated ( but complicated ) signaling to simplify path
creation and re-routing around failures, and QoS features for bandwidth
reservation, constant bit rate, variable bit rate, and unspecified bit rate
services, applied to the cell.

However, with these advantages, ATM also has certain drawbacks. The first
and foremost being that of "overhead".  ATM consumes nearly 10% of available
bandwidth with a 5 byte cell header for each 48 byte payload cell, plus an
additional 5% is needed for the adaption layer for IP over ATM as per RFC
1483. For example, an ATM OC-48 link requires 494Mbit/sec for overhead.
Compounding the bandwidth issues is ATM's limited scalability at higher link
rates. ATM switches have only recently delivered OC-48 interface rates and
it is questionable whether OC-192 is feasible considering the overhead
associated with segmentation and reassembly, wasted bandwidth, and other
inefficiencies of pushing 53bytes across 10Gbit/sec links. Today the fastest
IP router ATM interface is OC-12, which creates a bottleneck with the advent
of OC-192 capable transport systems.

When ATM is used as the transport for delivering IP in the Internet core we
face a different set of issues. ATM requires its own administrative domain
distinct from IP at Layer 3. The ATM network elements must be interconnected
in such a way to provide redundancy. The entire ATM topology is transparent
to the IP Layer 3 topology. Therefore a second topology at Layer 3 must be
overlaid atop the ATM fabric. This is achieved by establishing PVC's between
layer 3 routers. This creates another set of problems:

1.      Two separate modalities are required for element management adding
complexity and cost to network management.

2.      IP route exchange with an IGP requires direct peering/adjacency with all
neighbors, therefore the number of PVC's required grow by a factor of
n-to-the-power-2; where n is the number of internal IGP routers. For
example, for 300 routers: 44,850 PVC's would be necessary to establish a
complete mesh. If 4 more routers are added the PVC count jumps to 46,056 (an
increase of 1206 PVC's). This represents a substantial network-provisioning
problem. In the event of a router failure in this scenario, the surviving
routers will issue IGP routing updates on the order of n-to-the-power-3 (300
routers would issue 27 Million updates). This effect can be reduced by
configuring route-reflectors/confederations, however, it still adds to the
complexity and becomes a provisioning nightmare.

3.      ATM uses its own signaling protocol (PNNI) to establish PVC's. IP uses
OSPF, IS-IS, and BGP as its signaling protocols. The two signaling layers
operate independently and therefore complicate interworking between the
layers. To gain advantage of ATM traffic engineering features IP signaling
protocols must run within the ATM PVCs.

The question boils down to Howard's C. Berkowitz's often-quoted saying,
"What is the problem that you are trying to solve ? Knowing the advantages
and disadvantages that ATM offers, educated choices regarding its usage can
be made depending upon one's application.

Aziz S. Islam
marchFIRST Inc.; http://www.marchFIRST.com
55 York Street, Ste. 1500
Toronto/ON M5J 1R7/CANADA
ph:(416)368-2222 Ext. 211
fx:(416)366-6667
pg:(416)563-7355
mailto:[EMAIL PROTECTED]
mailto:[EMAIL PROTECTED]


___________________________________
UPDATED Posting Guidelines: http://www.groupstudy.com/list/guide.html
FAQ, list archives, and subscription info: http://www.groupstudy.com
Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]

Reply via email to