Re: [p2pi] Thought on congestion

[Jari Arkko <jari.arkko at piuha.net>]

Ted, Henning,

>> I was also a bit surprised by the notion that the top 5/10% of users
>> should simply be cut off in some way. This is not economically
>> efficient; they might, after all, be willing to pay for the benefit
>> they get from all that bandwidth.
>>     
>
> I think a different way to ask that question is:  "Does it make sense
> for users at the top of the usage curve to share infrastructure with
> those in the middle?"  As I noted in the paper I submitted, lots of
> networks are built with specific assumptions about how user
> traffic can be multiplexed over the shared resources.  It might be
> possible to retain those assumptions for a certain period of time
> by pushing those who didn't operate in the expected way on to
> different infrastructure (requiring that they buy leased-line capacity,
> for example); that way the networks' assumptions aren't challenged
> by their activity. 
>
> In case it is not obvious, I think that this does not solve the problem,
> as the current high-end users are likely just  a few years ahead
> of the typical consumer; I think we have to re-think the base
> usage assumptions to get anywhere in the long term.
>   

Indeed. If the world changed, it does us no good to either cut off the
people who were the first to change, or by force make everyone go back
to a traffic mix where statistical multiplexing works. At the end of the
day, networks exist to satisfy a demand, and it needs to do so while
making economic sense for the players involved.

Pricing models and service packaging has perhaps even more to do with
this than technical solutions. For instance, part of our problem is that
Internet service is marketed primarily through nominal link speeds. When
assumptions like the efficiency of statistical multiplexing in today's
traffic mix change that has a big impact on the way the business can be
run. For a given cost, we've actually been selling a combination of
peak, average, guaranteed, etc. service without making that very
explicit. But it could be made explicit very easily, e.g., gold service
= full link speed, silver service = peak can be full, average must be
under 50%, etc. If that doesn't sell, well, I guess we learned something
about actual demand.

This is not to say that there is no need for technical solutions.
Clearly, we can improve service under congested/outside the allowed
limits situations by better informing the network what traffic should be
given higher priority. We can improve the efficiency of the network by
better understanding of its topology. And so on. But the point is that
at the end of the day, you WANT to serve the p2p, video, etc. traffic as
well -- and this may mean increasing your capacity, changing the ratio
of access/backbone capacity, changing your subscriber contract deals and
prices, as well.

Jari

_______________________________________________
p2pi mailing list
p2pi at ietf.org
https://www.ietf.org/mailman/listinfo/p2pi