The IP over ATM Mailing List Archive by date

[onvural@vnet.ibm.com]

*** Resending note of 03/31/96 04:27
 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
 From:
 Subject:
Subject: Re: [Q] Suggest new protocol providing QoS...

Jon,

>>End station SW/HW needs to be upgraded, switched LANs will be
>>needed, and a new backbone needs to emerge?

>right now, we run audio and video on end systems and LANs without RSVP

Are you saying that there is no need to worry about delay, jitter, and
throughput in a shared LAN when voice and video are deployed together
with web and ftp traffic. If this is a dedicated LAN, it is a hardware
change (at the LAN switch). As new commercial applications emerge,
it is very likely that you will need resource reservation and priority
queueing at the workstations as well. (Consider all the related
efforts in IEEE, IETF, and on ATM). If this is an experiment where
you utilize your Ethernet say 20%, it is not a commercial environment
as far as I can say.
I would like to think the new FOO effort that recently started
is a proof of that provisioning for low utilization is not a desired
solution.

>we do NOT need to upgrade anymore than backbone routers to run this
>voice with guaratees (think: teklephones don't need to be upgraded to
>change the backboone of a phone net from analogue to digital - )

Note that ISDN phones are different than POTS phones. New features
require new hardware and software. POTS phone can access ISDN but
can not benefit :-) from all the new features.

>we can use aggregate, or site based reservations, or w can limit the
>_number of vat/vic calls from a site (same as a call blockig model) or
>we can add an authneticated application that does the rsvp and does
>billing - there are a plethora of ways of deploying guarantees and
>rsvp in the internet without changing end system operating systems or
>comms stack code, or even applications....

Absolutely. Then we will spend tons of resources (on pain killers
and headache medicines) in configuring/managing that network.
We can do that today, no need to devise a
new protocol/infrastructure. An IP network provisioned for 20%
utilization will solve all problems. Consider Juha's 4000 site
network. Please note that this is not even close to provisioning
a phone network. The phone backbone is provisioned for fixed bandwidth
channels to be utilized to 40% to guarantee that the down time is
less than 12 minutes per year. God helps us all if we try to provision
Internet with such strict goals in mind. Examples as such forces me
to think the difference between commercial service and a free service
(as the saying goes).
In any case, I believe dynamic protocols have their merits. Otherwise,
what is all this excitement in IETF and other places.

>>Of course I am referring to the availability of quality voice and
>>video services that require advanced scheduling and potentially
>>a lot of hardware support. For these applications, I don't believe
>>a simple software upgrade to current IP backbone will do any good.

>well, on the UK SuperJANET net, we have an existence proof that your
>belief is wrong - we upgraded the backbone, and can run broadcast
>quality video...over IP...

This is great.

>it isn't hard...and in fact, the way it delpys is very nice, since
>there are strong incentives to come up with more scalable solutios to
>admission control and billing than a VC per user per sessio typically
>hints at....

I never defended per VC based mechanisms, less complex solutions should
be (are?) possible.

>one of the problems with VC nets is they encourage you to use the
>lowest level of multiplexing, which gives the least level of traffic
>aggregation...whilsts tennenhouse's paper on "layered multiplexing
>considered harmful" is very well and good , i would counter it with
>"lowest layer multiplexing cosindered injurious to large scale network
>traffic management"...

I can see your point and may even agree. But how am I going to explain
the popularity of FR for data traffic and SONET for voice traffic.
I would like to think they are addressing real problems so these
networks are growing exponentially.

Few cents here, few cents there. Regards, Raif