FW: ISPs slowing P2P traffic...

Mike Lewinski mike at rockynet.com
Tue Jan 15 19:13:36 UTC 2008


Geo. wrote:

> Guys, according to wikipedia over 70 million people fileshare
> http://en.wikipedia.org/wiki/Ethics_of_file_sharing
> 
> That's not the fat man, that's a significant portion of the market.
> 
> Demand is changing, meet the new needs or die at the hands of your
> customers. It's not like you have a choice.

A few years ago I worked in a startup WISP environment - a coop that had 
a total of 5 T1s across three PoPs servicing some 100+ households over a 
50 sq mile area.

We had a customer share his 25GB music collection through gnutella, and 
then leave for a weekend vacation skiing. Meanwhile, other neighboring 
members who were actually still at home and attempting to use the 
network interactively, suffered until rate limiting was put into place. 
Contracts aside, given this is a coop, who has more right to use the 
bandwidth, at an ethical level?

The fact is that for shared content, as in customer-seeded torrents or 
shared gnutella files, demand scales at what *appears* to be an 
exponential rate in proportion to the amount of shared content. Please 
explain how a rural WISP can buy a pipe that will scale with that type 
of demand, and how they can recover the associated costs. Because in my 
experience, adding new upstream capacity often just exacerbates the 
overall problem, especially for nodes with strong signals to the PoP.

Note, this was in a rural area where the other available Internet 
options are:

1) satellite - expensive and wretched latency for the telecommuting 
crowd trying to get to their company resources over VPN

2) dialup - also poor latency that, when combined with VPN overhead is 
nearly useless for the majority of telecommuters

3) ISDN - $150-200 / month for better latency and moderately better 
throughput. I used a combination of satellite and ISDN prior to 
inception of the Coop to meet both low-latency needs for shell access 
and satellite for downloading large files. Then I looked at the combined 
costs (I didn't pay the ISDN) and realized I could almost be on a T1.

4) Fractional T1 or FR - $400-600+ / month

The only economically feasible way of providing broadband residential 
services are shared T1 distributed by wifi. There's no cable, no DSL, no 
FIOS. And there's zero likelihood, given distribution density and 
engineering challenges, that any of those offerings will ever make it up 
there.

At the time we started, there were three other WISPs in the same county 
competing with each other, all created in the same year, that had to 
contend with the P2P issues, as well as radio interference (I knew most 
of the parties and promoted the idea of "mutally assured destruction" if 
we didn't work out a channel schema, fortunately coverages didn't 
overlap in too many places).

Bandwidth may be commodified if you live in a major metropolitan area, 
but I don't see the rural WISPs having any other economically feasible 
options apart from QoS to control the P2P issues.

When I was on the BoD of the Coop, I promoted education as the primary 
method to change user P2P behavior, backed up by filters and QoS as 
needed. Others wanted Packet-shaper ($$$) automatic enforcement, but I 
felt that over time the P2P would migrate to encryption as a result of 
shaping (this back in ~ 2003 and it seemed like it was just around the 
corner, so perhaps I'm wrong on this).

As an alternative to education and QoS, I felt that metering and billing 
off of bandwidth used per-house was a good approach that would have it's 
own educational value. When bandwidth isn't actually a commodity, 
suddenly torrent is no longer the best means to acquire new "content" 
(and from my experience I can tell you, yes the customers were mostly 
downloading music and videos they didn't own).

P2P is IMHO a perfect example of the "tragedy of the commons". Too many 
users feel like "If I don't grab as much for myself as I can, some other 
fool is going to do the same thing and then I'll be left out". That 
mentality itself is also a form of mutually assured destruction in its 
own right.

So, to answer the "meet the new needs or die" poster I say - show me one 
other viable option for the rural WISP that doesn't involve QoS or 
filters and still makes a profit. Because I know a lot of WISPs who 
would love a better solution.

Finally: most of the people posting on this list aren't the ones writing 
the marketing or developing the actual product offerings. So all the 
people responding that "you shouldn't be selling unlimited if you can't 
afford to provide it" need to back off for a minute and recognize this 
simple reality. If engineers were running the sales and marketing I'm 
sure the world would be just fine ;)



More information about the NANOG mailing list