cr> Bradley re: isochronous techhie talk

1996-01-24

Richard Moore

@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
Date: Tue, 23 Jan 1996
From: •••@••.••• (Allan Bradley)
Subject: Re: cr> ...building a parallel isochronous Internet

Richard Moore wrote:

>        I'd welcome _a bit_ of techhie talk on the list about the likely
>new net architectures, and how Internet might co-exist or inter-operate
>with them -- just so we can have an informed model of the likely scenarios.

F.Y.I

I wrote this some months ago. Maybe it will stir some interest.


First,  I breakdown the new "Communications" market into three (3)
categories:  1. Technology (applications, hardware/software, etc - of all
sorts) 2. Bandwidth (the size of the pipes to support the Technology) and
3. Distribution (the layout allocation of the pipes, e.g., fiber,
satellite, cable, etc.).

ATM (Asynchronous Transfer Mode) Network Technology

ATM takes data information (voice, video and data) and chops it up into
small chunks which are called "cells".  These cells are essentially
segmented in the hardware of LAN/WAN Switch Hubs into "virtual circuits".
Within these hubs are virtual switching engines (in hardware) which forward
the cells from their source to their destination.  The problem is that
massive amounts of virtual circuits can be immense.  For example, lets say
you need to connect 200 buildings - that turns into a 200 by 200 possible
switch matrix which means 40,000 potential switched circuits.  This tends
to mean that you have to be selective on what sites get what resources and
management of switching fabrics can be expensive.

Although the switching is transparent to the user and the virtual circuits
are basically self defining, the issue of really controlling the switching
fabric is paramount.  A switching fabric (hardware based) is called
connection oriented because a virtual end-to-end circuit has been
established (required for voice, video and data streams) and is load
independent (more users performance integrity remains the same).  The
Internet is based on connectionless data or datagram schemes.  The main
difference is that the Internet is software based (TCP/IP)  and therefore
highly programmable - the drawback is that it is slow and unpredictable
(jerky video) and load dependent (more users worse performance) - the
advantage is that it self allocating and managing based on user addresses
- not circuits.  I think a key difference is that the Internet is user
controlable by addresses, ATM will be more vendor controlled by circuits.

Enter the standards bodies (ATM Forum) basically made up of network
hardware, software vendors and phone companies.  Best of all worlds is to
combine the connection data streaming performance with the managing
software of connectionless control.  The definition of what are called
Adaptation Layers were established.  There is still huge debates on what
these Adaptation Layers will really provide and to what level they will be
a true industry open standard.

For last few years there has been a big battle on the standards committees
on how to price switched circuits.  The phone companies wanted a header to
be put on the cell so that Bandwidth on Demand services could be defined
(essentially the more bandwidth you use the more you pay) the network
vendors did not want a cell header because it would slow performance and
create more switching overhead.  I believe now there will be to two
categories of switched connectivity and will be allocated on a platform by
platform basis.   Which is fine because both are needed for different
markets.

The real difference is that as a communications company (based on the phone
model) I could run a line and price it as a dedicated path (pipe) with only
a singular associated bandwidth payload - which is the way it is today.
With new technologies I can take that same path and create multiple
aggregate payloads and the physical rate structure is the same to the
communications company.

What does this mean.  Well for example, lets take two different public
sites: a residential community and  a business park.   And for the sake of
universal access both have been wired with the same amount of fiber,
copper, etc. to support high-end communications.  Both from a
communications company perspective have cost exactly the same amount as far
as distribution.  But the technology levels given the business park will
require a thousand times more bandwidth than that of the residential
community.  Where do you think the profit margin and the emphasis is going
to be? It will be about circuit control and pricing.  The thing to remember
is that with ATM your getting more capability given the existing physical
infrastructure - is it a coincidence that many Public Utilities Commissions
tie usage rates to the cost of distribution, but with the advent of
technologies like ATM, deregulation and unbundling of rate structures are
now a strong part of the Bill.

The Telecom Bill transport fairness rules are based on old vertical market
allocation, when the real money will be in horizontal market bandwidth
distribution.  That is the monopoly danger.  That is why I harp on defining
the models.  If you have the money for the technology, you have the huge
advantage in the market.

My concern is that schools, public services, city communities may get the
short end of the stick, because it is a distribution bandwidth market and
they need cheap bandwidth and that could lower market margins.  The
investment cost in these technologies today is very high.  The problem is
that regulations need to include bandwidth distribution and not exclusively
physical distribution as the basis of universal access/distribution.


Allan Bradley

ConsulMetrix, Inc.
Setting the Standards in Technology Consulting

@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@



 ~=-=-=-=-=-=-=-=~=-=-=-=-=-=-=-=-=~--~=-=-=-=-=-=-=-=-=~=-=-=-=-=-=-=-=-=~
 Posted by Richard K. Moore (•••@••.•••) Wexford, Ireland
   Cyber-Rights:   http://www.cpsr.org/cpsr/nii/cyber-rights/
   CyberLib:       http://www.internet-eireann.ie/cyberlib
 Materials may be reposted in their entirety for non-commercial use.
 ~=-=-=-=-=-=-=-=~=-=-=-=-=-=-=-=-=~--~=-=-=-=-=-=-=-=-=~=-=-=-=-=-=-=-=-=~