[e2e] QoS vs Bandwidth Overprovisioning
pingpan at juniper.net
Wed Apr 25 12:30:32 PDT 2001
There is nothing wrong with overprovisioning. The only problem is that
it can only be done (or guaranteed) within one carrier's network (except
on trans-atlantic links, perhaps). Randy had provided the data in the
past that his network has almost no congestion and delay.
Unfortunately, when you start to check on e2e delay, it can be quite
large. This's because end-user packets go over multiple carriers and
multiple inter-domain connection. There is no guarantee on end-to-end
delay in the Internet today. Period!
By the way, we need to define the meaning of bandwidth utilization here:
most of the providers give you only the average bandwidth usage.
However, the QoS factors are really driven by the peak bandwidth usage,
and the duration of operating at peak. I have seen links that run at
~30% on each 24-hour window, in which it runs above 80% of link capacity
for 5-6 hours.
Alhussein Abouzeid wrote:
> On Wed, 25 Apr 2001, Randy Bush wrote:
> > >> Bandwidth overprovisioning has worked for most of the last decade.
> > > tell me, then why we hear so regularly that the Internet doesn't work?
> > because people make money by saying so?
> > randy
> totally disagree - an advertising major was asking me the other day
> why the heck (excuse her language) the quality of her phone calls to Cairo
> did not improve when she upgraded her "Internet connection" -
> I had to explain to her the harsh reality, that she has overprovisioned
> only part of the path, and, that there's *nothing* she can do
> about it.
More information about the end2end-interest