Question on 95th percentile and Over-usage transit

Mailman List Mirror (Read Only)
1

*//Sorry for the earlier misguiding email subject//*

Dear All,

Thanks for all the replies! I would like to see more, to learn more!

Since I (Research Assistant) am not from network operations and
management domain, I am trying to model the transit pricing function. In
my research work, I am using a pricing model from this work! This
pricing model is as follows:-- Transit price = Constant * (aggregate
traffic)^0.75, which is exactly similar to the one described by Ryan
Malayter in his earlier message. Hence I am wondering, whether the
pricing should be a linear(CDR*[95th peak]) or sub-linear (like the
above)?

With Regards
Pradeep
Research Assistant
Institute IMDEA Networks
Madrid, Spain

2

Yes. :slight_smile:

I think you'll find actual contracts out in the wild that do it either way, and
probably lots of variants as well, because the organizations buying the transit
have differing motivations. Some will want to minimize their monthly expenses
at all costs and hope they don't get a surprise billing spike due to high
traffic, while others may very well be willing to pay 10% more a month for a
"guaranteed no surprises" billing structure for budgeting reasons.

Now, if you have a good model for "how likely is each method to result in
surprises in the real world".... :wink: