Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Re: [EFM] OAM developing Geoff's observation.

On Mon, 17 Sep 2001 13:55:18 -0500  Roy Bynum wrote:
> There are a lot of ISP based SPs that are creating a lot of noise.  Is it 
> because they are falling prey to the ".com" syndrome?  Could it be that 
> they do not have the ability to deliver a high margin service and do not 
> want others to realize that?  What about the rest of the industry?  What is 
> the reality of the rest of the industry?  What are service requirements for 
> business versus those for residential?


	I think I understand (and share) your frustration with the
industry.  However, I think it is dangerous to make assumptions on why
businesses fail, or on the industry as a whole, by concentrating on a
few well publicized cases.  It is also dangerous to make assumptions
about how competitive a particular product is based on the health of
the vendor as a whole.  HP can sell great printers but if the rest of
their product line is in the crapper, they are out of business.  From
my perspective, the reason these ISP based SP's have gone out of
business has nothing to do with IP, but rather that they were lousy
businesspeople.  Also, I couldn't help but notice they weren't very
good IP engineers...  Lots of inexperienced folk jumped into the

        So, companies like ours have been successful in our market
niche.  We make money selling high speed IP services.  I don't think
we are shy about explaining what we do and how we do it.  In fact, I
worry sometimes that we say too much...  I feel both GWI and Oregon
Trails have given back to the community by publicizing the reasons for
our success, and giving others a chance to copy and potentially
compete with us.  Traditionally, sharing knowledge this is the
Internet way.

	You are most correct that business requirements may differ
from residential.  I would not want to claim otherwise.

> Think about this simple fact.  The vast majority of small to medium size 
> businesses that have multiple sites on an enterprise network, use several 
> "Private Line" or "Virtual Private Line" links to make up their enterprise 
> networks.  Most of them however only have one Internet type link.  If 
> people simple think thing through and realize how their own enterprise 
> networks are deployed, they would realize what the real priority for EFM 
> supporting business deployment should be.  In this environment slightly 
> higher costs can be justified to obtain better efficiency.

I respectfully disagree on this point.  We don't build our networks
this way.  I am not asking you to agree with me.  I am asking you to
treat my perspective with the same respect I treat yours.  What I
would like to see is an EFM which would allow you to build QoS into
your network, and us not to use your QoS mechanisms.  Then we can let
the market decide the argument, not a vote of a standards commitee.

It may be that there is no way to build QoS without inserting it into
EFM.  However, before adding complicated QoS mechanisms I would like
to see significant proof that it is necessary and at least two working
prototypes of the mechanism; prototypes which can be easily

For an IP connection, QoS on one part of a TCP/IP link does not add
much value.  QoS must be end-to-end to be effective.

> For deployment to residential services, there are other issues.  Only the 
> voice service is symmetrical.

One data point to back up your conclusion, over the last five years,
for our high speed residential IP services we see a fairly steady 10:1
incoming to outgoing.

However, a lesson I have been taught again and again over the last 15
years is that you can't predict the traffic patterns of the customer.
We got into the ISP business before the growth of the web and frankly
I thought the web was an insignificant toy.  Our world changed when
the web became widespread.  None of us know what the next killer app
will be.  I don't feel comfortable predicting future values based on
past performance.

Fortunately, both IP and Ethernet handle changes in traffic patterns

>  The Internet service to residential 
> customers is normally very asymmetrical.  The video/broadcast services are 
> simplex.  This will make for a very different infrastructure and cost model 
> than business services.  Other than for the voice, PPPoE may work very 
> well.

Peer-to-peer traffic such as gaming and teenage-girl-to-teenage-girl
applications are our biggest driver of high speed, low latency
residential services.  None of us can predict the next killer
app... but my gut is that it will not be VoIP nor traditional video.
Spend some time with a teenager and you may develop the same

By the way, this teenage peer-to-peer stuff took us by suprise.  We
used to have some of the local traffic filters other SPs have talked
about on our networks.  We learned the hard way that kids these days
want low latency, high speed links internal to their town.  If you
make their packets take the long way round, they will drop you like a
hot potato in order to switch to the local cable company...  What is
how you say it "low ping bastard?"

> The real question is, can a service provider make any money in this 
> domain?

Yes, we do make money currently and for the last number of years.  I
hope we will continue to make money.  But the future is always
uncertain.  I am trying to "make my own luck" by participating in this

sincere and most respectful regards,