roundtable: Re: feast-or-famine, FYI from Edupage

roundtable: Re: feast-or-famine, FYI from Edupage

Re: feast-or-famine, FYI from Edupage

Curtiss Priest (cpriest@juno.com)
Fri, 01 Aug 1997 09:37:45 EDT 

To: ROUNDTABLE@CNI.ORG
Subject: Re: feast-or-famine, FYI from Edupage
Message-Id: <19970801.093606.7775.1.cpriest@juno.com>
From: cpriest@juno.com (Curtiss Priest)
Date: Fri, 01 Aug 1997 09:37:45 EDT


--------- Begin forwarded message ----------
From: "Kirt Olson (CITS)" <kolson@HELIOS.ACOMP.USF.EDU>
To: Multiple recipients of list
CYBER-SOC,<CYBER-SOC@LISTSERV.READADP.COM>

On Thu, 31 Jul 1997, Curtiss Priest wrote:
> 
> RESEARCHERS ADVOCATE INTERNET USE CHARGE
> 
> ...when users are encouraged by fast response times, they ramp up
> their Internet activities, thus creating the "storms," or bursts of
> congestion, that continually plague the Net.  When the response time
> slows to a crawl, users back off, and eventually things get back to
> normal.
> 
> To avoid the feast-or-famine scenario, the researchers advocate
> charging all users according to the amount of bandwidth they use.

Strikes me as like paving the bumpy dirt road and adding speedbumps to
the paving.

This proposed solution adds costs, creates classes of users,
subdivides an existing public utility, benefits network owners at the
expense of most users, may reduce the incentive to build more
bandwidth, and does not address the delay in expanding the network.

Bandwidth is not a new problem, nor are network storms. Nor is
bandwidth the most expensive, or even the key, ingredient.

Added costs come from the need to process each packet and assess its
priority and thus its cost, and to formulate and deploy a billing
scheme. None of this is free, and we can be sure that the providers
will want the billing services to be profitable in themselves.

Users are segmented into the permanently underserved and various
stages of luxury service. Once your packet is labelled, there is no
competition for transit services, your packet is permanently put in
first class or cattle car based on your ability to pay. While it may
be technically possible to create variable-cost packets, I personally
believe they are essentially unsalable to most users.

We do not discuss here the concept of building faster luxury networks.
Such networks already exist--private networks that provide security,
speed, and reliability for those who need such services. Nope, we are
talking about dividing up the existing network and creating the high,
medium, and low classes of service by worsening the medium and low so
as to reserve bandwidth for the high.

Network owners stand to gain revenues out of proportion to any
investment. I assert that the costs of billing will be recovered, and
the costs of service will rise. This proposal makes no sense to a
network owner if the scheme is revenue-neutral since, I assert, it
creates a greater total number of dissatisfied customers, greater
expense, and no offsetting benefits.

In an environment where ISP's are currently waiting months for new T1
or better lines, why would the networks build more bandwidth? Seems to
me they have plenty of customers to whom they cannot deliver
installation services now and they would get more revenue on existing
bandwidth under this proposal. Certainly the telcos have shown little
interest in providing bandwidth as they have dragged their feet on
ISDN and are busily withdrawing the bare copper line tariffs that let
clever folk build their own high bandwidth solutions.

Then there's the issue of time lag in building to meet increased
demand. One does not ramp up production of bandwidth in some short
time. It's rather like farming, one plants the cable, makes the
connections, installs the switches, and then tries to sell the crop.
Demand can rise quickly and the response cannot unless the bandwidth
is already available and unused. Is repricing likely to affect supply
in such an instance? Will new entrants leap in? I think not.

Some people act as though the Internet snuck up on them overnight.
Telcos, for example, suddenly notice that calls are longer and their
switching strategy is outmoded. But this trend is two decades old. We
did not install 100 million PC's last night, or even last Thursday.
Large, sophisticated companies that manufacture modems and telephone
switches and run telephone networks might have noticed modem sales,
don't you think? Could not some communications guru in a well-funded
corporate laboratory have noticed that the total connected leaf speed
was growing faster than the backbone speed in the network?

Network storms were noted on small ethernets almost right away. We
pretty much know that an ethernet needs a bandwidth about three times
its loading to function well. This is old news. Usually the cure is
segmentation--keep the local traffic local. Of course, if one decides
to exchange messages among nets only in DC, IL, and CA, it may be hard
to deal with the intrastate traffic in GA, but why is rate
discrimination a relevant cure?

Bandwidth is cheap, and getting cheaper. High bandwidth fiber on new
routes is not proportionately costlier, and new modulation schemes on
older routes permit much more bandwidth per dollar invested. Just as
new fast modems are cheaper than the slower ones of 20 years ago, so
the backbone capabilities are cheaper as well.

I believe those who propose pricing by bandwidth live in the past and
act as we once did when memory, hard drive size, and processor speed
were to be conserved compared with the convenience of people. These
things grew ever cheaper, and convenience ever more valuable.  In the
same way, bandwidth is growing cheaper and the gain is precious time.

--Kirt

--------- End forwarded message ----------

[CNI Home Page]