Hi kc!  For those of you who don't know her, she is one of the
pioneers of the 'Use _accurate_ data, stupid' movement [1].  She and
CAIDA have produced some of the best real-world analysis of IP
networks that I have seen (and I've spent considerable time looking.)
She is also a serial presenter at NANOG and imho one of the brighter
members of the networking community.  I think that she can help us
more accurately look at our deployments in a way that can make them
comparable to other types of wide-area network infrastructure.  This
will both allow us to present a compelling case to regulators and make
more informed decisions about our own operations.

>From my view, the primary advantage that WISPs possess lies in our
ability to deploy low-cost, high-bandwidth, high-reliability systems
using a minimum of (costly [2]) resources.  We need to be able to
demonstrate this to the regulators in a way in which it can be easily
compared to other providers of similar services.  We then need to
demonstrate how the regulators can enhance WISP efficiency by
providing us with resources which they control -- spectrum, better
tower regulations, more favorable interconnection laws (PSTN and IP,)
and restrictions on lower-government interference.  To communicate
effectively with the regulators we need to have accurate (and
reproducible) data.


I see the data as having two basic components:

1. Information about the industry which provides an accurate picture
of the capabilities of "properly implemented [3]" WISP networks.  This
data would need to be in a form that could be directly compared to
other methods of providing network access.

2. Real-world case studies of successful deployments that illustrate
the data in (1.) so that any layman can make the association between
the specific case study and the overall data.

I'm not sure what the best formats/units are. Hopefully, k, you can
help us figure that out.  These are my 2am thoughts for data points
which could begin the discussion [4]:

For point-to-multipoint networks:
* Resources required to deploy a network serving "a given area"
  Time
  Money
  Expertise (scarce knowledge resources)
  Intangibles (spectrum, etc)

Ultimately expressed in:
  Total Infrastructure cost per Mbit/sec/MHz/subscriber
  Maintenance cost per Mbit/sec/MHz/subscriber/year

I am not sure how to define "a given area."  Should subscriber density
be taken into account?  Should we consider terrain characteristics?
How do we define expertise; do we have to?

For point-to point Systems
* Resources required to deploy a network connecting two points
  Time
  Money
  Expertise (scarce knowledge resources)
  Intangibles (spectrum, etc)

Ultimately expressed in:
  Infrastructure cost per Mbit/sec/MHz/mile
  Maintenance cost per Mbit/sec/MHz/mile


One of the risks of doing this type of analysis is that we may find
that under many circumstances, wireless is not currently the most
efficient choice for network connectivity.  This could also be a
reward-in-disguise because it would allow WISPs to use that analysis
to determine whether a new market has a reasonable chance of being a
successful deployment.  If we do the analysis correctly we can use it
to determine when wireless would become the best choice (ie: if
equipment prices dropped or if spectral efficiency increased.)

Another risk is that we won't get comparable data from other forms of
connectivity -- I foresee certain organizations intentionally
providing incorrect data.  I'm not sure how to reconcile that.  Are
these numbers something that we can back into using publicly
available, illegal-to-misreport information?


This type of information will be invaluable to the future of wireless
connectivity and time is of the essence.  When the regulators decide
to auction off spectrum there is a ~0% chance that it will later
become available as a freely shared resource, but if we can
demonstrate the benefit of unlicensed/shared spectrum before the next
auction is approved, there is a good chance that future spectrum will
be considered for similar use.

Best,
Tony

[1] MIMU = Moniker that I made up.  One of her NANOG presentations was
basically a demonstration that a significant portion of the data
claiming to represent aggregate Internet traffic was largely invalid.
 She has pointed out on numerous occasions during other NANOG
presentations that either a commenter's information or the presenter's
information was based on a less-than-desirable data set (and usually
provided a pointer to the correct data.)  Combining this with "Keep It
Simple Stupid" yields the words above.

[2] Imagine if we had to pay what the cell companies have said that
the spectrum was "worth" at auction... could any of us make an
economic case for $30/month accounts if we had to pay $4 billion for
20MHz of spectrum covering New York state [a]?  Assuming a population
of 18.98M and an average household size of 2.61 [b], that's $550 for
every household regardless of whether they use your service.  I don't
think that the cell companies can even justify it by saying it's only
$211/person...  I think that they are betting that their investors
won't realize how huge a number that is given the fact that it may not
secure their ability to be a monopoly in the market.  Additionally,
the auction process really amounts to a lump-sum-tax on consumers.  As
license-free wireless has demonstrated in small areas, current
unlicensed "junk" spectrum can function in a very similar capacity to
auctioned spectrum.  If the allocated spectrum is large enough or
coordinated well enough, the exact same functionality can be provided
without the multi-billion dollar cost which is passed on to the users.
 [a]
http://www.findarticles.com/p/articles/mi_m0UKG/is_2001_Feb_5/ai_70709354
[b]
http://factfinder.census.gov/servlet/GCTTable?_bm=y&-geo_id=&-ds_name=DEC_2000_SF1_U&-_lang=en&-mt_name=DEC_2000_SF1_U_GCTH6_US9&-format=US-9|US-9S&-CONTEXT=gct

[3] By "properly implemented" I mean a system that has been
professionally engineered and has a reasonable expectation of being
adopted by others once they understand the merits of the design.  For
instance, a point-to-point backhaul network of narrow-beamwidth
antennas would be a "properly implemented" network under most
conditions.  A point-to-point link using omnis would hardly every be.

[4] These may be totally off the wall, but I needed a starting point.
 These examples are too generic because they don't take into account
different types of deployments (fixed wireless vs. mobile access.)
They are also too specific because they limit the data to figures that
can't be easily derived from aggregate information like financial
statements.

--
Tony Weasler
Attron Networks
cel: 414-202-1027
888-8-ATTRON x101
http://attron.net/


On 9/27/2005 11:47 PM, k claffy created:
> [not reading this list regularly, but tom hit a nerve]:
> 
> tom et al
> 
> caida (www.caida.org) is an internet data analysis/research organization
> whose mission includes informing public policy, aimed toward improving
> policy 'toward congruence' with our best empirical (scientifically grounded)
> understanding of the relevant technological issues/constraints/parameters.
> 
> i am no expert on spectrum policy, but afaict the difference between
> having huge effect and having no effect is sufficiently formalized
> reporting/analysis of Real World Operational Experiences (this means
> you), written in way that will convey to scientists (this means me), as
> well as to the public, what happens when technology gets deployed in
> reality.  one underutilized option is collaborating with university
> researchers to quantitatively document (1) potential deliverables under
> various regulatory scenaraios (2) successes and failures under existing
> regulatory scenarios.
> 
> caida Really wants to help support forward motion here, but we are
> desperately lacking hard data.  emergency situations are obviously not
> the time to talk about research, but i want to make it clear that if you
> still don't have what you want by the time this emergency is over, please
> don't underestimate the value of hard data and careful articulation
> of the experiences you have had, so that scientists can come in and help
> compile them into comprehensive and unassailable demonstrations to their
> funding agencies of why change is essential.  
> 
> i believe the right kind of analyses/reporting could reduce the
> length of this fight from 10 years to 2.  (ok, maybe 20 to 4...)
>  
> but the research community and the deployment communities are going 
> to have to [find time and resources] to work together. we've never
> needed eachother more.
> 
> k
-- 
WISPA Wireless List: wireless@wispa.org

Subscribe/Unsubscribe:
http://lists.wispa.org/mailman/listinfo/wireless

Archives: http://lists.wispa.org/pipermail/wireless/

Reply via email to