SMART Letter #73
Buy as Many Nines as You Need
June 26, 2002
!@#$%^&*()!@#$%^&*()!@#$%^&*()!@#$%^&*()!@#$%^&*()!@#$%^&*()
------------------------------------------------------------
SMART Letter #73 -- June 26, 2002
Copyright 2002 by David S. Isenberg
isen.com -- "single point of success"
isen@isen.com -- http://isen.com/ -- 1-888-isen-com
------------------------------------------------------------
!@#$%^&*()!@#$%^&*()!@#$%^&*()!@#$%^&*()!@#$%^&*()!@#$%^&*()
CONTENTS
> Buy as Many Nines as You Need
> More on Cisco and Corporate Culture
> SMART People save $2000 at WTN Summit
> Motorola Launches Canopy
> Quote of Note: Billy Tauzin
> Copyright Notice, Administrivia
-------
Buy as Many Nines as You Need
by David S. Isenberg
Two recent articles in Business Communications Review (BCR)
on the concept of five-nines (99.999 percent reliability)
in circuit and packet networks miss one key fact: To the
extent that true competition is available, customers can
buy as much network reliability as they are willing to pay
for (see David Yedwab and Chris Talbott's article in BCR,
June 2002, pp. 46-52, and Gary Audin's article in May 2002,
pp. 22-27).
The only certain path to reliability is physical redundancy
-- the use of multiple independent networks with dissimilar
infrastructures in different places. A network with the
same brand of router everywhere is subject to single-point
failures of hardware or software. So are facilities that
share the same route or the same building. Design for
reliability is design for multiple, heterogeneous networks.
The definitive tome, Engineering and Operations in the Bell
System, (R.F. Rey, ed., Bell Labs, Murray Hill NJ, 1977),
which sat on the shelf of every Bell Labs engineer in the
Bell System's heyday, devoted but two paragraphs to
physical redundancy. It declared that network designers
faced, "two fundamentally conflicting goals: to minimize
the total cost of the network and to provide redundancy as
a protection against major service interruptions." It
continued, "Redundancy is provided by multiple, physically
separate long-haul routes, although at increased cost."
That was the last word. To Engineering and Operations,
reliability was more an economic issue than a technical
one.
A Buyer's Market
In the bad old days, the extra costs of redundancy were
folded into the monopoly's rate base, and the customer had
no choice. Any color, as long as it's black.
Today, competition exists. Customers, especially business
customers, can buy connectivity in a variety of forms --
copper, fiber, wireless, etc. -- from several different
providers. Each provider wants their service to be
reliable, but the survival imperative is low cost --
everybody wants to be the low-cost provider.
This creates a buyer's market for reliability. If one
network is not reliable enough, buy connectivity from two.
After all, competition makes network services inexpensive,
and customers can buy boxes that will fail-over gracefully.
If the probability of an outage in either network is 1%,
the probability that a customer with two one-percent
networks will be without service is 0.01% That's four
nines of up-time. Still not enough nines? Buy service
from a third provider.
Sure extra reliability costs more -- it always has, and it
always will. But today the trade-off is explicit, and it
is in the hands of the customer.
Speaking for myself, the fact that I have cable modem
service, dial-up Internet access and "wireless web" on my
cell phone practically means that I'll *never* be without
an Internet connection. Of course I pay for each increment
of reliability, but I'd rather have the choice than have to
buy my reliability as part of a hairball of hidden features
sold by a single provider.
A Broader View
Both the BCR articles referenced above take too narrow a
view of reliability differences between IP and TDM
networks. It is like trying to compare the flying
abilities of bumble bees and 747s. Not only are there vast
differences (and a single similarity), but it is much
easier to prove that a TDM network is reliable and that a
747 can fly.
A TDM circuit is a definite entity. It is made or broken.
Its latency is built into its structure. If it is up for
99 minutes and down for one, you can measure the
probability of an outage.
But IP fudges the distinction between "working" and
"broken". An IP network is designed to perform over
unreliable physical media in the first place. When a
packet does not arrive intact the first time, and then TCP
does its job just fine and the packet arrives on the second
attempt, is the network broken? When packet jitter begins
to rise, when does the nine-counter begin to decrement?
When can it be said that the effort to deliver a packet is
less than "best effort"?
Conclusion
Admittedly, sometimes the Internet breaks. Sometimes I
can't dial in. Sometimes I can't get mail from the POP
server. Sometimes my cable modem won't bring me the New
York Times. But I can stash my mail in two places. I can
dial in when the cable fails. And I can always read the
Washington Post. I have never yet seen a failure that a
smart customer can't overcome. There are material costs of
administering reliability like this, but these costs, too,
are explicit and under customer control.
Just as the Internet pushes value creation to the edge of
the network by treating any feature that is integral to a
specific set of facilities as just another network-specific
difference to route around, so does the Internet push the
ultimate responsibility for reliability into the lap of the
customer.
[This article originally appeared in Business
Communications Review, June 2002, p. 53, reprinted by
permission of Key3Media, the publisher of BCR.]
-------
More on Cisco and Corporate Culture
I was gratified to see that "The New Cisco Kid -- SMART
Letter #73" stirred up the beehive at Cisco. I got at
least 15 new Cisco-employed subscribers.
One of them, who requested that his or her name not be
used, nailed the patrimonial bureaucracy syndrome when he
or she wrote:
"I see a couple of things that have happened in the last
four years or so compared to how we were in the early
90's:
"1. people are more concerned about the past, present,
and potential emotional states of VP's instead of what's
right. A kind strange kind of "VP potential emotional
state" poker gets played in meetings:
"John is really excited about plan A"
"Larry has a lot of heartburn about doing plan B"
"I don't think Bill would like to see plan A happen"
"I don't want to be the one to tell Kevin we can't
make A happen"
[There seems to be little consideration of] the relative
risks and merits of the options. [Instead] bargaining
is based on various VP's emotional reaction to the
results of the meeting.
"2. Individuals cannot take action or make suggestions:
you have to be following orders from a VP or relaying
his desires."
Now consider this passage from Moral Mazes (Robert Jackall,
Oxford, New York, 1988), the best objective study of
corporate culture and the origin of the term "patrimonial
bureaucracy":
"When managers describe their work to an outsider, they
almost always first say: 'I work for [Bill James]' or 'I
report to [Harry Mills].' and only then proceed to
describe their actual work functions . . . The key
interlocking mechanism of [modern corporate culture] is
its reporting system. Each manager . . . formulates his
commitments to his boss; this boss takes these
commitments and those of his other subordinates, and in
turn makes a commitment to his boss . . . This
'management-by-objective' system, as it is usually
called, creates a chain of commitments from the CEO down
to the lowliest product manager or account executive.
In practice, it also shapes a patrimonial authority
arrangement that is crucial to defining both the
immediate experiences and the long-run career chances of
individual managers. In this world, a subordinate owes
fealty principally to his immediate boss."
Moral Mazes goes on to describe how bosses use ambiguity
with their subordinates (and other more-or-less unconscious
subterfuges) in order to preserve the power to claim credit
and deflect blame, which tends to perpetuate the
personalization of authority. Unlike a straight, Max Weber
style bureaucracy, which is procedure-bound and rule-
driven, a patrimonial bureaucracy is a set of hierarchical
fiefdoms defined by personal power and patronage.
You can read my review of Moral Mazes in "Corporate
Culture's Twisty Passages -- SMART Letter #22" at
http://isen.com/archives/990601.html
-------
SMART People Save $2000 at World Technology Network Summit
The WTN Summit and Awards will be held July 21 and 22 in
New York City. SMART People who REGISTER BEFORE JULY 1
will pay $1950 (normal fee is $3950). After July 1, the
SMART People's rate is $2750 (versus $6750). Last year's
WTN meetings in London were excellent!
For the special SMART Person's rate, see:
http://www.wtn.net/registration/smart/intro7735.html
-------
Motorola's Canopy comes out from under
Motorola has just launched a wireless, unlicensed (5GHz)
last-mile product line named Canopy that delivers up to 10
megabits per second per subscriber, for up to 1200
subscribers per access point, for a one-time cost that
asymptotes at around $500 list. Note that Canopy is not
Cognitive Radio -- it is just good solid wireless
connectivity with a keep-it-simple-stupid proprietary
modulation method. I used an early version last year. We
set up a 1/3 mile link with ease and it worked flawlessly
for the three days we needed it. My friends at Motorola
are aiming at the Wireless ISP market, and it looks like
they might have hit it -- see http://motorola.com/canopy
-------
Quote of Note: Billy Tauzin
"I don't want our military operations to be interfered
with, and I don't want planes to fall out of the sky.
But I want real-world evidence that tells us whether
ultrawideband devices, on a stand-alone or cumulative
basis, would cause these things to occur."
U.S. Rep. Billy Tauzin (D-La.),
http://www.infoworld.com/articles/hn/xml/02/06/06/020606hnf
cc.xml?0607frpm
-------
CONFERENCE ON MY CALENDAR
July 21-22, 2002, New York. World Technology Network Summit
and Awards. For a SMART Person's discount, see
http://www.wtn.net/registration/smart/intro7735.html
October 8-10, 2002, Atlanta GA. Fall VON. I'll be giving
an Industry Perspective talk. Time and date to be
determined. See http://www.von.com/
October 15-17, 2002, New Orleans LA. Fiber to the Home
Council Annual Conference. I'll be giving a keynote (on why
neither telco nor cable TV co will bring us fiber to the
home). Nothing on the website yet, but keep checking
http://www.ftthcouncil.org for information.
-------
COPYRIGHT NOTICE: Redistribution of this document, or any
part of it, is permitted for non-commercial purposes,
provided that the two lines below are reproduced with it:
Copyright 2002 by David S. Isenberg
isen@isen.com -- http://isen.com/ -- 1-888-isen-com
-------