[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Fred Goldstein critique of Bell Atlantic ISDN Cost Studies
-----------------------------------------------------------------
Info-Policy-Notes - A newsletter available from listproc@tap.org
-----------------------------------------------------------------
INFORMATION POLICY NOTES
November 11, 1996
This is Fred Goldstein's testimony to the Maryland Public Services
Commission on the ISDN tariff. It contains a devastating critique of the
Bell Atlantic cost study for ISDN pricing. It is rather technical, but I
am posting it because of the importance of the critique of the Bell
Atlantic's statements regarding the Interent and network congestion.
I am leaving for India tonight, and will not be back in the office until
November 20. jamie
BEFORE THE MARYLAND PUBLIC SERVICES COMMISSION
In the Matter of the Residential )
Intellilinq BRI Service Offering ) Case 8730
of Bell Atlantic, Maryland, Inc. )
REBUTTAL TESTIMONY OF FRED GOLDSTEIN
ON BEHALF OF CONSUMER PROJECT ON TECHNOLOGY
November 5, 1996
PREPARED REBUTTAL TESTIMONY OF FRED GOLDSTEIN
Q1. Please state your name and business address, and
summarize your professional qualifications.
A1. My name is Fred Goldstein. I am a Senior Consultant
at BBN Corporation. This testimony does not represent an
official position of BBN Corporation. It is prepared on
behalf of the Center for the Study of Responsive Law's
Consumer Project on Technology. My business address is at BBN
Corp., 50 Moulton St., Cambridge MA 02138. I have worked in
the telecommunications and data network field since 1977. My
experience with ISDN dates back to 1985 when I became Digital
Equipment Corporation's voting member of ANSI-accredited
Technical Subcommittee T1D1 (later T1S1), which was
responsible for producing North American ISDN standards. I am
the author of the book "ISDN In Perspective" (Reading MA:
Addison-Wesley, 1992) and have taught courses on ISDN for
Northeastern University and National Technological University.
I have also been a technical leader in BBN's ISDN acitivities
as an Internet Service Provider. I hold three patents in the
area of Broadband ISDN and Asynchronous Transfer Mode
congestion management and switching.
Q2. What is your opinion of the Bell Atlantic analysis
used to support its residential tariff?
A2. The first testimony is that of Stephen A Reisinger
[REIS]. The gist of this testimony seems to be that Internet
Service Providers place a high load on the network,
particularly when they use analog ports. This is not even on
point to Case 8730 concerning Residential ISDN rates, except
by inference. Reisinger begins [REIS at p. 2/ line 21] by
stating that "far fewer R-ISDN lines can be accommodated on
equivalent pieces of equipment." This would imply that higher
monthly recurring charges (MRCs) should be applied to ISDN.
That ISDN MRCs are higher than analog MRCs are not in dispute;
the application of measured-service charges to R-ISDN is in
dispute.
Reisinger then analyzes the incoming traffic to Internet
Service Providers, none of whom are eligible for R-ISDN
tariffs. He points out congestion that occurs at Central
Offices that receive calls. These lines generate essentially
zero calls. In his attached Report to the FCC calling for
higher charges to be placed on ISPs, he requests that ISPs
bear a large multiple of the cost of the additional traffic
generated by users calling them. Yet the conclusion he is
attempting to draw in Case 8730 is that R-ISDN rates should be
metered at a contributory level in order to pay for the same
thing.
His examples are particularly off point. They mostly
concern analog dial-in ports to ISPs. The first example,
Myersville, had 127 analog lines. The second, Herndon, had
1131 analog lines. Residential ISDN subscribers can not
generally call analog lines! They generally call ISPs who use
ISDN PRI, which does not use any line unit, or sometimes BRI,
which does not use the analog line unit. In fact, his
arguments are for increased migration to ISDN. An analog line
unit on a 5ESS has a maximum capacity of 512 lines and 64 time
slots, or about 3.8 CCS/line if filled [REIS at 11/16]. But
ISDN Line Units on a 5ESS have up to 256 time slots [REIS at
13/footnote 8] to serve up to 512 Basic Rate lines, which can
have up to 1024 B channels.
He states that a load of 8.6 CCS/line will limit an ISLU
to 448 lines. This would not be serious even if it were true,
but it does not add up. A full ISLU with 512 lines at 8.6
CCS/line generates 4403 CCS. From the Poisson table, this can
be served by 149 time slots at P.01 blocking. This is well
below the 256 time slot limit.
Also, the 8.6 CCS/line number is derived from ISDN
experience which is primarily Centrex. Bell Atlantic actively
markets ISDN Centrex as a service that provides unlimited
calling within the Centrex group. This invites users to
"nail" channels. Any tariff that does not provide totally
unlimited usage will discourage nailed usage. Since Centrex
costs substantially less than Bell Atlantic's proposed rate
for ISDN BRI with unlimited use ($249/month), users who plan
to "nail" ISDN lines are likely to use their ISP's Centrex
service, not R-ISDN. Indeed the testimony of Curt Koeppen
[KOEP] seems to contradict Reisinger, where it says, "The
typical customer spends an average of 10-40 hours per month
on-line."[KOEP at 5/13] While this cannot be directly
translated to CCS, it is in line with typical voice usage
patterns, not ISDN Centrex.
Given that the ISDN line unit has such greater capacity
than the analog line unit, it becomes clear that high-traffic
data users should be encouraged to migrate to ISDN. Further,
since analog dial-in traffic costs far more than PRI dial-in
traffic (estimated by Reisinger at $50/month/channel vs.
$75/month/channel), ISDN potentially reduces costs at both
ends of the connection!
Reisinger also argues that ISDN incurs costs by requiring
the installation of 64 kbit/sec clear-channel trunks, compared
to "56 kb/s trunk" facilities used for voice. While much of
the embedded base of trunk facilities is not 64 kbit/sec
"clear", most new facilities are. Also, Bell Atlantic
proposes charging the same usage rates for calls placed on
older facilities, which can use either the speech/audio bearer
capability or the 56 kbps data bearer capability.
A particularly bad scare tactic occurs in Reisinger's
testimony about 911 service [REIS at 16/23]. He supposes that
911 calls would be blocked if 64 analog users simultaneously
nailed up calls on the same line unit. Given a small
percentage of "nailed" users, this is, of course, extremely
improbable! But it would be even more vanishingly improbable
on an ISDN line unit, with its larger number of time slots.
It should be pointed out that the DMS-100 switch has even more
favorable ISDN traffic characteristics. An older-style
analog-only Line Concentration Module (LCM) has 640 one-line
card slots sharing a maximum of 180 time slots, assigned 30 at
a time. An ISDN-capable LCME has 480 card slots sharing a
maximum of 480 time slots. This allows non-blocking use of
one B channel at a time, or approximately 34 CCS/line total.
Robert Terry's testimony is not quantitative; it simply
asserts that ISDN should be regulated as discretionary. But
he notes that "existing demand for ISDN service is less than
1% of BA-MD's total customer base and is forecast to be about
3% by the end of the decade." This is a self-fulfilling
prophecy. Higher penetration rates have already been achieved
in Germany. BA-MD's application of measured rates to R-ISDN
will dramatically suppress demand.
The testimony of John Pehta [PEHT] demonstrates various
reasons why the MRC of R-ISDN is higher than the MRC of POTS.
This is undeniable; however, it does not affect usage costs.
He notes that data calls have an average holding time of 20
minutes [PEHT at 13/15], but this is meaningless: The
average CCS and average monthly hours of use are what consume
traffic capacity. Data calls may be on average longer but
they are fewer in number, and thus incur lower call setup
costs (including processor usage and Signaling System 7
network utilization). He then states, "Because of the
expected additional usage per line for R-ISDN customers, the
average concentration ratio for a switch is expected to be 4
to 1 which allows for a switching module/line unit to provide
service to only 256 customers." This is based on having 64
time slots per line unit, which is the case for analog line
units on a 5ESS. ISDN line units have up to 256 time slots.
And the situation on the DMS-100 is even more favorable.
He then states that concentration ratios will "move closer to
1 to 1" [PETH at 14 /10], which implies an average ISDN usage
of over 500 hours/month. This is more than an order of
magnitude above Bell Atlantic's own estimates [KOEP at 5/13].
He then says that one ISDN switching module/line unit has 64
time slots, contradicting Reisinger's testimony. Again the
problems he ascribes to ISDN users are in fact problems caused
by analog data users.
The net cost per minute of an ISDN data call, which I did
not see because it is expurgated, is perhaps higher than the
cost per minute of an ISDN voice call, due to the use of newer
trunks. However, it is probably lower than the cost of an
analog voice or voiceband data call, because it avoids the
analog line units. The Reisinger-written Report to the FCC
[BELL] states that an analog ISP port costs $75/month for its
traffic load, versus $50/month for a PRI channel, with an
average load of 608 minutes per day. This translates to
$.0041 per minute for analog usage vs. $.0028 for PRI audio-
bearer usage.
Curt Koeppen's testimony[KOEP] seemingly seeks to
contradict the other testimonies. He tries to show that Bell
Atlantic's proposed tariff is really quite low. This follows
from his assertion that the typical user's volume is quite
low. This low volume usage is in fact consistent with other
telco's experiences. Pacific Bell, for instance, cites in its
current ISDN filing before the California PUC (docket A.95-12-
043) requesting a cap on off-peak hour unlimited usage) an
average usage of 47 hours/month, which includes some
percentage of "nailed" users and a majority of users
originating under 20 hours/month. This is the result of a
tariff that has unlimited usage except on Monday to Friday
from 8 AM to 5 PM. He also cites average Internet and on-line
service usage studies in the range of 9.9 to 47.3 hours per
month, and says Bell Atlantic's own studies "show that typical
residential Internet/On-line usage is only 10 to 40 hours per
month" [KOEP at 27]. If this is the case, then why all the
fuss about high traffic levels?
Koeppen cites the recent Bell South filing [KOEP at 38/11]
which caps "flat" rate at 200 hours per month, with 1 cent per
overtime minute, as a means to discourage "nailed" users.
Bell Atlantic certainly could use a similar mechanism to
achieve nearly-flat rates for R-ISDN. This would be more
palatable to the residential market. Indeed the Maryland
public is so displeased with the possibility of mandatory
measured service that legislation was passed to outlaw it.
Under Bell Atlantic's proposal, then, users wanting flat rate
or even nearly-flat-rate service would be encouraged to use
analog lines.
Koeppen also uses unsound reasoning in developing
contribution margins. He states, "in deriving our margin
above direct costs, our cost calculation for usage included
only 90% utilization of the package usage allotment."[KOEP at
37/6] Thus someone who subscribes to the 60-hour package is
expected to use 54 hours. This is clearly not valid.
Subscribers generally do not know ahead of time their exact
calling patterns, and Bell Atlantic's proposed rate structure
gives positive incentives to buy a larger than necessary
Callpack, because the overtime rate does not decline with
usage. Thus a 35-hour user might be better off with a 60-hour
Callpack, as 15 hours at 2c/minute costs $18/month, versus a
$14 marginal cost for Callpack 60 over Callpack 20. A 90-
hour user will be better off with Callpack 140 than with
Callpack 60, even if all usage is off peak, as the former
would cost $60 and the latter ($45+60*.01*30=) $63. Thus the
average Callpack will be much less than 90% utilized.
Experience also shows that residential users favor
predictable, "flat" rates over measured rates, even if the
latter are slightly lower. This will cause residential users
to favor analog over ISDN, Centrex ISDN over R-ISDN, and
larger Callpacks over "optimal" Callpacks that have a risk of
overtime.
Q3. Does this conclude your testimony?
A3. Yes.
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
INFORMATION POLICY NOTES is a free Internet newsletter sponsored
by the Taxpayer Assets Project (TAP) and the Consumer Project on
Technology (CPT). Both groups are projects of the Center for
Study of Responsive Law, which is run by Ralph Nader. The
LISTPROC services are provide by Essential Information. Archives
of Info-Policy-Notes are available from
http://www.essential.org/listproc/info-policy-notes/
TAP and CPT both have Internet Web pages.
http://www.tap.org
http://www.essential.org/cpt
Subscription requests to info-policy-notes to listproc@tap.org with
the message: subscribe info-policy-notes Jane Doe
TAP and CPT can both be reached off the net at P.O. Box 19367,
Washington, DC 20036, Voice: 202/387-8030; Fax: 202/234-5176
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++