Max, Min & Fair

My tendency is to think that ultimately the Internet will have differentiated packet classes initiated from the edges by requests from customers for services of their choosing. But maybe fatter and more numerous pipes will continue to save the day. I like how Henning Schulzrinne explains how this can be done while not threatening NN. His audio testimony is linked at the bottom of my NN post at http://adam.404.org/2007/09/08/network-neutrality-nonneutrality

In my thinking the following might be even fairer while maintaining the packet content non-discrimination goal:

1) When demand for network bandwidth on shared facilities is low, every user gets as much as he wants.

2) When demand for network bandwidth exceeds supply, every user is allocated bandwidth equitably.

3) Network bandwidth consumed per user is tracked for some window extending back back in time such that it covers a number of low/high bandwidth demand cycles.

4) “Equitable” allocation means something like this: every user requesting less than the average per-user available bandwidth gets what he requests; those who request more get additional bandwidth when it’s available and distributed in inverse proportion to the their consumed bandwidth according to 3 above.

The bandwidth hogs still gain by favoring off-peak times. But those who use little bandwidth but at the high-demand times still get their money's worth.

What do you think?

I almost read that as "deep pocket inspection" :-)

Bob,

My broadband ISP here in Australia implemented a similar concept for many years, which they called 'Flat Rate'.

In essence, the ISP's bandwidth was broken down into blocks; those that wanted committed 'always fast' broadband and those that wanted 'generally fast'.

The 'always fast' option would never be slowed down under any circumstances, you'd purchase x gigabytes per month and you'd get it delivered to you at your full broadband capacity.

The 'generally fast' option would give you the data at full rate if there wasn't any congestion on their network. As soon as the congestion kicked in, a 'generally fast' users data would be prioritised against other 'generally fast' users data on a rolling window x days into the past. So if user A had downloaded more than user B, user B would get their data first and user A would have their packets held/delayed/prioritised by a margin to make sure that users who downloaded less got their data first - up to a threshold so as to minimise/remove dropped packets even when you were being prioritised.

Their plans were hugely successful here in Australia, where we have never been blessed with 'all you can eat' broadband like many other countries. They continued to offer the flat rate plans up until last year, when the leeches where still pulling down 100Gb+ per month and we were seeing soaring international internet transit prices. In the end, they had to stop offering the plan as the high end users were pulling so much data (which was all international) - that it was making the platform unsustainable and unprofitable for them.

All that aside, their implementation of your concept was exceptional and was market leading here in Australia for many years.

Alistair.