Cell Relay Archive

[Bernd Reuther <reuther@informatik.uni-kl.de>]

Jonathan Turner wrote:
> 
> Like many people, I have been disappointed in recent years at the
> limited success that ATM technology has managed to achieve in
> the commercial market. Recently, I've become convinced that
> one of the reasons for this limited success is that the
> industry chose exactly the wrong strategy for ATM
> deployment into the global communications infrastructure.
...

> Rather than subordinate
> ATM to IP, we should have been using IP as a subnet technology
> for carrying ATM traffic. 


I don't believe that an "ATM over IP" initiative would have changed much
in the development of ATM.

I think, when a "new" network protocol or technology should be established, 
then it is important that there are valuable improvements of the "new" technology 
over the "old" technology, e.g new features (QoS or mobility), faster, cheaper ,...
For ATM (and several other developments) it seems that the QoS feature
is the most valuable improvement. Therefore it is important to make ATMs QoS
available for everyday work in order to establish(=sell) the "new" technologie.

What does this "make QoS available" include:

Infrastructure: 
- availability of QoS capable networks end-to-end

Software:
- QoS support by protocolls and systemsoftware 
- applications must be able to request QoS dynamically or
  administrators must be able to configure static QoS parameters
  
Knowledge (administrators or application programmes or maybe users):
- about the QoS requirements of data flows/applications
  (which QoS is needed or desirable ?)
  
Management:
- accounting (+ billing) and policing 
  (very important for every day usage, but not for experimental usage)

Putting ATM over IP may help to solve the infrastructure problem: 
There are many ATM islands which are not connected to each other. 
(BTW the IETF observed the same problem for IPv6 and defines the 6to4 mechanism. 
http://search.ietf.org/internet-drafts/draft-ietf-ngtrans-6to4-03.txt)

But since ATM over IP addresses only one of many problems mentioned above, 
I don't believe that an ATM over IP initiative in the past would have lead 
to a widespread use of ATM today.

Further I'm convinced that other QoS enabling initatives like IntServ/RSVP or DiffServ
will run into similar problems than ATM.

Nevertheless in our research group we have a project (Beyond-IP) aiming to make
applications indenpendent of concrete transport service providers (TSP).
A special TSP selects an apropiate concrete TSP at runtime.
This way all applications using the special TSP automatically support 
TCP/IP and ATM and others... 

	Bernd