Scenes from life on the data superhighway -- from IEEE Computer Society's membership magazine, COMPU

[David Farber <>]

From: MElliott () computer org
Date: Tue, 26 Apr 94 15:58:00 EST
To: farber () central cis upenn edu
Cc: lewis () cs nps navy mil


Dave - IP may be interested in the following column by the editor of
the IEEE Computer Society's membership magazine, COMPUTER.  Ted Lewis
is at the US Naval Postgraduate School.  Comments may be sent to Ted
at t.lewis () computer org.   Michael




Editor-in-Chief's Message:
April 1994 COMPUTER


Scenes from life on the data superhighway


Main Street through Cyberspaceville is a path most often traveled
by writers of fiction, as well as a favorite avenue for cruising
by vice presidents; but Main Street may turn out to be little more
than a dirt trail for covered wagons rather than a freeway for
electric cars. Why? Let's get real for a moment.


What kind of home computer is going to recycle a gigabit of multimedia
consumer trash during every second of prime time? What kind of cable
TV company is going to serve 100,000 distributed interactive game
players in real time? What kind of computer is going to provide
full-time knowledge services to the inert cyberpunk Simpsons? Here are
a few scenarios to contemplate.


Scenario 1. The 1995 multimedia desktop machine will be capable of 1
million 8-bit pixels per frame multiplied by 30 frames per second.
Assuming we can compress this signal by a factor of 100 (lossy
compression is acceptable in the consumer market), we need a data
driveway with a bandwidth of approximately 2.4 megabits per second. This
is comparable to a typical Ethernet LAN, so it is not beyond current
technology, unless you want it sustained.  Current workstation- class
machines are capable of sustaining little more than one tenth of this
bandwidth for the length of a movie or an interactive 3D simulation.
Unix, for example, must buffer, swap, interrupt, and go through a number
of other contortions just to handle text-only e-mail. So what makes us
pioneers of the electronic frontier think we can provide on/off ramps
for the superhighway?


One idea is to bypass the operating system or get a new one. While we're
at it, we must redesign the TCP/IP protocol to work on really fast
networks.  And what if we want to keep some of this data? Then we simply
invent low-cost terabyte data storage vaults. Finally, what if we want to
take a snapshot? All right, then we invent a low-cost VCR printer. these
four simple requirements could be the basis for several industries. Bill
Gates, where are you?


Scenario 2. Ma and Pa Bell own the local cable TV system with 100,000
paying customers. Their any-time-of-the-day pay-per-view movie department
must be able to handle up to 100,000 requests for a single movie at once.
They need a 10,000-transaction-per-second database machine in their garage
and an unbelievably wide driveway connecting them to the superhighway just
to handle Friday night.


Idea number one is to boost contemporary database machines from 1,000 tps
to 10,000 tps. To solve this annoying little problem, we need a tenfold
increase in transaction processing. This should keep the database machine
vendors busy for at least a decade. Now, suppose each of the 100,000
customers orders the standard 2.4-Mbps service. The 1-gigabit superhighway
starts to look like the Ventura Freeway at rush hour.  Idea number two is
for cable TV companies like Ma and Pa Bell to buy a telephone company because
telephone companies are the only data truckers in business today with
sufficient switching capacity.


Scenario 3. The superhighway is so much fun that 100 million people get on
it every night to play interactive games like "Presidential Election 1996"
and "Stock Market IPO." They not only send/receive 2.4 Mbits of data each
second, but they also do some actual computing. For example, they might play
against an expert system with more political tricks than Ross Perot, or they
might exercise a best-selling 500-variable world-economic model while waiting
for the microwave to defrost dinner: simple stuff like that. Suppose each
player needs 100 Mflops of raw compute power. That adds up to a potential of
10 billion Mflops of compute power!  So let's convert Rhode Island into a
massively parallel processor with silicon circuits etched on every square acre.
Alternately, we could use up the idle cycles on all those unused Apple Newtons
connected to each other by wireless networks. RISC or not, future computer
architects have job security.


Scenario 4. The renaissance era of super libraries arrives. We no longer have
to lower our blood pressure by biking to the local library because everything
we ever wanted to know is just a keyboard and telephone (cable) away. In fact,
2.4 Mbits of raw data is streaming past our doorway each second. If only we
could convert this raw data into information.


Idea number one is to create a whole new industry of knowledge workers who
filter raw data and convert it into information. This reduces the flow rate
to a few thousand facts per second. Idea number two is to create the
twenty-first-century electronic publishing industry. Packaging a few thousand
facts per second into meaningful amusement for the Simpson family should keep
all of us busy for a few decades. And creating entirely new industries
stimulates the economy.


The data superhighway is only the tip of the integrated circuit. Sure,
computers need more I/O capability, but they also need more storage, more
processing power, and more connectivity. Instead of solving one problem,
the superhighway may well create so many technical challenges that the skills
of Computer's readers will remain in great demand until the year 2094!


Ted Lewis, editor-in-chief


Distributed with permission from the April 1994 issue of COMPUTER, magazine,
a publicaton of the IEEE Computer Society.




=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
 Dr. T. Michael Elliott               TEL   +1-202-371-0101
 Executive Director                   FAX   +1-202-296-2187
 IEEE Computer Society
 1730 Massachusetts Ave., N.W.        EMAIL  melliott () computer org
 Washington,  DC  20036-1992
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=