Re: "Network Neutrality" or "Open Internet" and why it costs

[Kevin McAleavey <kevinmca () nsclean com>]

Sean ... THANK YOU! That is perhaps the best explanation of "how radio works" that I've seen lately. I "retired" from 
the New York State Department of Public Service (NYPSC) back in 1996 to move into security software and while I was 
with "the state" I worked in the PSC's Communications Division and was a liason with the FCC on telephony, cable 
television and "internet" as it was cast off from DARPA so they could use the money for "Internet II."

 What you describe is exactly the basis for the assertions by the "reconstituted Ma Bell" (AT&T which is still waiting 
for Verizon and Qwest to come home). POTS is what subsidized internet services as we know it today since the level 1 
networks were capable of providing switched long distance as well as internet. And as long as POTS service was there, 
there was a subsidy stream for construction of additional fibers and the world was a happy place since the majority of 
internet users were on dialup and most level 5 switches had direct access to colocated toll switches which passed off 
the traffic to the backbone under a reasonable "load" with essentially voice circuits working as they had been 
provisioned.

 As broadband took hold, the demands on the PSTN rose astronomically while tolls dropped, and now that VOIP is making 
drastic inroads on the profitable POTS service, the telcos are now required to provide abundant bandwidth and 
construction costs with declining revenue. The cost of trying to find routes for new fiber paths continues to increase 
while capacity remains largely stagnant, even declining overall. This is the reason for the economy of scale of all 
these mergers. And yet the demand for high bandwidth services continues to increase and thus the telcos must now look 
at means of covering the cost of increasing that capacity to meet the demands of so many streaming multicasts. I 
understand the problem and if folks don't want to see even more "throttling" as a result of these bottlenecks, 
something's got to give.

 I know every month when we see OUR bandwidth bills going up that similar is occurring elsewhere and all those neat 
Flash pages and other streaming data are eating the network alive. It'd be real interesting if someone could come up 
with a chart showing how "capacity" which was rarely reached in the recent past is now a commonplace event at most 
switches now. And the problem for the telcos is that they NEED to build out more plant at the same time that revenues 
are on the decline.

 I certainly don't LIKE hearing that high bandwidth customers like google video, Vonage and others are being pressed to 
pay more to offer their services, but at the same time - more pipe needs to be built to accomodate all of these "gee 
whiz" features. The days of message units seemed almost over with, and it looks like a return of "metered service" 
might be the only means of paying for all the necessary construction.

 Just wanted to say VERY thought-provoking and hope your explanation helps to put things a bit more in perspective for 
folks ... the world is changing again and few people actually understand that the "carrier system" can't expand by 
itself.   :)

 With "broadband"

At 01:11 AM 4/17/06, Sean Donelan wrote:
> If you buy a point-to-point, dedicated T1 you will receive exactly a T1
> worth of capacity between those two points.  However, people buying packet
> swtiched or Internet connections are not buying point-to-point, dedicated
> circuits beyond the first network switch/router.
>
> With a phone switch you get 1 voice grade circuit connection or you
> get a busy signal.  The PSTN is not engineered so everyone making a phone
> call at the same time. A phone switch may have 100 POTS lines and for 10
> upstream trunk lines. If more than 10% or so people try to make phone
> calls at the same time, the phone switch will probably give a "fast busy
> signal" or all circuits are busy.  A circuit switched network tends to be
> all or nothing.
>
> With a router you may have 100 downstream T1 (1.544Mbps) circuits and 1
> DS3 (45Mbps) upstream circuit.  Unlike the PSTN, the Internet doesn't
> have an explicit "all circuits are busy signal."  Instead our theoritical
> router with 45Mbps upstream won't give each of the 100 downstream users a
> full T1's worth of capacity, they will only get a portion of the 45Mbps
> of upstream capacity.  This process repeats at every network router, e.g.
> the next router may have 100 DS3 downsream circuits and 2 OC3 upstream
> circuits. In a packet switched network, your end to end bandwidth is
> determined by the most congested network point in the path, not
> necessarily your access link.
>
> ATM and frame-relay networks are sometimes sold with a "Committed
> Information Rate (CIR)" which may be a bandwidth number between 0Mbps
> and the speed of your access link.  Although typically the maximum
> CIR will be less than 80% of the access link speed.  A CIR is part of
> the contract with the service provider to guarantee you will be able to
> get at least that amount of bandwidth between two points on the providers
> network.  Generally service providers only offer CIR guarantees on
> their own network, not across other service provider networks.
>
> Networks with CIR's are more expensive than networks without CIR's, so
> fewer people pay for them.  Most Internet connections, and essentially
> all residential Internet connections, are sold as "best-effort" which
> means they have a "0 Mbps" Committed Information Rate.
>
> Or to put it another way suppose a carrier sold Internet service like
> this
>
>   0Mbps CIR (10Mbps Peak Information Rate): $50/month "Best Effort"
>   1Mbps CIR (10Mbps PIR): $100/month
>   10Mbps CIR (10Mbps PIR): $500/month
>
> Which means on the theoritical router with a 45Mbps upstream connection
> it forward packets in the following order
>
> First, subscribers paying for 10Mbps service, up to 10Mbps.
> Second, subscribers paying for 1Mbps service, up to 10Mbps.
> Third, subscribers paying for 0Mbps service, up to 10Mbps.
>
> The peak information rate is the same in all three cases, 10Mbps. B
> But when there is congestion on the router's upstream circuit the 10mbps
> CIR subscribers always get 10Mbps, the 1Mbps CIR subscribers always get at
> least 1Mbps, and the "Best Effort" 0Mbps CIR subscribers get the remaining
> capacity.  The Best Effort CIR subscribers may get 10Mbps a lot of the
> time, but are the first ones bumped when the router is congested.
>
> Best Effort is much cheaper than CIR, but like buying a "standby" ticket
> on an airplane flight you are also accepting the risk that there might
> not be capacity on a particular flight.  If you want a committed
> information rate, all the carriers are more than happy to sell you a
> different product (dedicated circuits, ATM, frame-relay, or the proposed
> "premium" Internet) with that guarantee.
>
> > Wouldnt adding/removing QoS to the packets violate the ToS? Are they
> > still getting the B/W they paid for?
>
> You can buy a dedicated T1 (1.544Mbps) circuit between two points for
> $2,000/month; you can buy a 8Mbps cable modem connections for $50/month.
> What are you actually buying?
>
> "Best Effort" means on a space available basis.  If you read the "*" in
> the terms of service, it doesn't promise any particular amount of capacity
> on an end-to-end basis.
>
> _______________________________________________
> Fun and Misc security discussion for OT posts.
> https://linuxbox.org/cgi-bin/mailman/listinfo/funsec
> Note: funsec is a public and open mailing list.

----------------------------------------------------
 Kevin McAleavey at your service
 Privacy Software Corporation
 http://www.nsclean.com
 kevinmca () nsclean com

_______________________________________________
Fun and Misc security discussion for OT posts.
https://linuxbox.org/cgi-bin/mailman/listinfo/funsec
Note: funsec is a public and open mailing list.