Re: IP: Pac Bell says Net use may collapse phone system

["Fred R. Goldstein" <fgoldstein@bbn.com>]

  At 04:18 PM 10/28/96 -0500, Dave Reid wrote:
  
  >How much of a percentage of an ISP's customers may have be connected
  >to the ISP at any one time bears no relevance on the argument in any
  >way shape or form. The only thing that is relevant is average usage
  >of any given line, be it residential or business or ISP. The average
  >usage for any given ISP line is approx. 10 times what a residential
  >line is. 
  
  No, the average outgoing usage of an ISP line is zero.
  
  If telcos got their nirvana, a monopoly with all service meausred/timed  and
  leased-line rates in the stratosphere (think "France", for instance), then
  ISPs would pay next to nothing, because they don't initiate calls. An ISDN
  PRI configured for incoming-only costs under US$200 per month in France! Why
  not?  You can't call it without incurring usage charges, so they practically
  give away the lines in order to reduce busy signals!
  
  >An ISP may only be connecting 10% of it's customers at a time but it
  >it is using 100% of it's line's capacity, as opposed to average residential
  >that use maybe 10% of it's capacity. You could have a really bad ISP that 
  >under provisions his lines such that only 5% of his customer's can get online
  >at any one time. That still means that any one of his lines is used at
  >neary 100% capacity. It makes no difference how many customers of ISP may
  try to
  >get to his pool, the real indicator is the average usage of the ISP's lines.
  >This average usage indicates the impact they have on the network. 
  
  Again, no.  An ISP is just like any other hunt group, say, if there were a
  PBX at Telus' headquarters, or a PBX at the University of Alberta, or a PBX
  at the NorTel factory, or a PBX at the hockey arena.  The impact on the
  network is effectively nil, because custom imputes the cost of traffic to
  the caller.  The impact of all those callers *to* Telus, UofA, NorTel or the
  hockey rink is however significant, and spread among those callers, just as
  is the case of calls to ISPs.
  
  >Think about the problem for the standpoint of any single line and it requires
  >in terms of resources from the switch.
  
  Any large hunt group generates predictable traffic, if it's the right size
  for the demand.  I have a Poisson table here to show me the numbers.  ISPs
  are just like any other hunt group user.
  
  Telcos' problem is that flat rate resi tariffs don't generate incremental
  revenue for incremental usage, and they under-predicted average per-line usage.
  
  Also note that many press articles on this are predicated on the innumeracy
  of the average reader.  PacBell, f'rinstance, is quoted as having to spend
  millions of dollars to upgrade its network.  The recent news.com artice
  states, "The price tag for the upgrade needed to support the demand
  generated by the Internet would be a collective $1 billion for the seven
  baby Bells over the next several years" (per Bellcore).  Plus Pacific Bell
  is spending "$14 million to $15 million for switch upgrades this year
  alone". Major money?  Well, PacBell's number is around a dollar a line.  Not
  much when you figure that the capital investment per line is many hundreds.
  And the billion over several years is spread among over a hundred million
  lines, so you're talking a whopping-heavy capital investment on the order of
  ten bucks a line spread over a few years.  In other words, it's probably
  less than they spend on promotional coffee mugs, pens, and
  faux-leather-bound portfolios.  Compare, for instance, to the amount that
  MCI, AT&T and Sprint pay monthly in CCLC charges.
  
  Switch capacity isn't free, but the cost is, in the scheme of things, de
  minimis.  Use the numbers out of context and it sounds like a whole lot more
  than it is.
  ___
  Fred R. Goldstein      fgoldstein@bbn.com  
  BBN Corp.              Cambridge MA  USA    +1 617 873 3850