The real bandwidth numbers.

Stephen D. Williams
Mon, 28 Jan 2002 23:06:37 -0500

What I and others such as Adam have proposed is, more or less, an open 
source distributed content proxy/cache/router/rebroadcast server that 
all ISPs would run just as they all must run SMTP and DNS servers now. 
 Properly designed and implemented, popular content would be replicated 
in a combination of real-time and staged and end-users would experience 
far more effective bandwidth usage.

ISPs also would experience a greatly leveraged uplink usage since a 
large percentage of bulk transfer data would represent only a single 
stream over their expensive uplink.

There are many aspects to this, as I have discussed before, but let me 
point out one obvious one:

Network TV is perfectly doable over all existing broadband and backbone 
connections with the simple addition of such a rebroadcast server as 
only a single copy of each channel would flow into any single ISP.  The 
killer app here, btw, is that a standard and open method of end-user 
usage tracking, anon or not, would indicate to broadcasters/content 
providers, exactly how many end-listeners there were at any particular 
time.  (Max Headroom-like  Did you guys 
know he wasn't CGI!!??  It was all a latex max and fiberglass suit shell.)

If I thought I could overcome the existing ISP mindset and certain 
monopoly-like interests of AOL and UUNet enough that they'd actually run 
these servers properly, I would help start a Sourceforge project to 
create them.

This eventuality is what would supplant Akamai.  They have to contract 
for distributed bandwidth and then convince their customers that they 
need that capacity.  Most don't most of the time.

Looking at it another way, for every web site that is experiencing a 
temporary peak, there are hundreds that are on their off days.  This 
works out for mass host sites, but it could work at the corporate level 
too with cooperative caching, etc.

How we get there from here is tricky however.


Mark Day wrote:

>>I'm not sure what exactly Akamai does, but this purely enduser software
>>seem to put them out of business on the long run.
>Thanks for the clarification.
>FWIW, I think that particular "long run" is pretty far out. I doubt that P2P
>content distribution is very high on the list of things that really threaten
>them (or other CDN businesses) at the moment.  There are lots of problems in
>CDN-land, but I don't think P2P competition is a big one yet.