[LINK] 'true' cost of bandwidth?

Tom Koltai tomk at unwired.com.au
Sun May 3 21:31:08 AEST 2009

> -----Original Message-----
> From: link-bounces at mailman1.anu.edu.au 
> [mailto:link-bounces at mailman1.anu.edu.au] On Behalf Of Leah Manta
> Sent: Sunday, 3 May 2009 6:18 PM
> To: Scott Howard
> Cc: Jan Whitaker; link at anu.edu.au
> Subject: Re: [LINK] 'true' cost of bandwidth?
> At 02:35 03/05/2009, you wrote:
> >What a completely bogus argument.  I'd expect that of 
> Slashdot, but not 
> >of the NY Times.  This is equivalent to claiming that if one day 
> >everyone decided to avoid Victoria Road in Sydney, then the 
> cost to the 
> >government would be exactly the same as on a day when the 
> entire Sydney 
> >population tried to use it.  That statement is completely true, of 
> >course, but it's also completely irrelevant to the true cost of 
> >maintaining the infrastructure involved.
> I think they are specifically saying that "bits and bytes" on the 
> wire cost nothing more or less than the cost of the wire.
> If I have a 100 Mbps pipe and I put nothing over it, I'm still paying 
> for a 100 Mbps pipe.  If I move 100 Mbps data across it, I'm still 
> paying for a 100 Mbps pipe.
> Years ago Data charging didn't exist in the USA.  It was about ports 
> and pipes.  The port cost X and the pipe size cost Y.  Y was defined 
> as the number of TDM slots needed to move your data from A to B.  The 
> more data you needed to move over bigger pipes the more TDM slots you 
> needed on the fibre and hence your share cost more.
> But that doesn't mean DATA moving, or NOT moving, across the 
> pipe costs more.


Actually, I disagree.

The two most expensive running costs of maintaining the net are power
and cooling.
A Cisco 7640 with 0 data across its backplane runs a lot cooler than one
with a full load of 100 Mb per 100 Mb port

So power and cooling alone - are a significant cost in running a network
at empty or under full load.
I have logs on power consumption for networks from 64Kb right through to
OC48 switch rooms. (Sacramento)
About 12 years of data in all.

Additionally, the heat genrated from a router switching a full ACL
privides a high failure rate on cards that were not always designed to
"exceed" the spec (see Adrians post earlier) 

The next highest cost after bandwidth is replacement power supplies
(about one power supplt per fully loaded switch/router every 7 months -
empty routers power supplies last for 4-5 years) replacement cards, QOS
switches (so that when the pipes are fill corp customers still get what
they're paying for.

So Scott, unfortunately your argument is only partly true. Yes, the
bandwidth does not increase.
But the administration management of peering requests, transit/peering
renegotiation costs, QOS tunneling, help desk for QOS issues,
replacement power supplies, replacement cards, disk management (for
cache/squid) etc is a hell of a lot higher under a full 100 Mb than an
empty one.
Bandwidth is a very small part of the cost of running an ISP.


No viruses found in this outgoing message
Scanned by iolo AntiVirus

More information about the Link mailing list