[e2e] Re: crippled Internet

Christian Huitema huitema at exchange.microsoft.com
Wed Apr 25 19:15:37 PDT 2001

What I see in the graphs is that the max delay is sometime very long,
which raises quite a few questions. How do we explain a delay of more
than one second between two I2 connected universities? Is some gigabit
router somewhere allowing its queues to reach 100 megabytes? Or is some
kind of a bug? And, if this is a bug, is the bug in the measurement
system, or in some routing component? 

Assuming that this is a queuing delay, then we cannot deduce from the
small average and std that an I2 like network is fit for VoIP. A first
problem there is that the std dev may not be mathematically significant,
specially if the delay distribution is heavy tailed; it would be more
interesting to get some distribution indication, e.g. what is the % of
packets that incur delays longer than 200ms, 500ms, etc. A second
problem is the distribution of the outliers; are there concentrated
during rare events, or are they spread throughout the day? It would be
really interesting to run a test program that mimics a VoIP session, and
study how classic jitter compensation algorithm would react to the
observed distribution.

-- Christian Huitema

> -----Original Message-----
> From: Randy Bush [mailto:randy at psg.com]
> Sent: Wednesday, April 25, 2001 12:15 PM
> To: stanislav shalunov
> Cc: end2end-interest at postel.org
> Subject: Re: [e2e] Re: crippled Internet
> > Unidirectional characteristics (with GPS):
> http://www.advanced.org/surveyor/
> i don't think fred would appreciate us using I2 as an example of a
> real-world interprovider isp scenario.  [ btw, the two delay graphs i
> posted
> were each end-site to end-site inter-provider ]
> > Round-trip characteristics:	http://watt.nlanr.net/active/
> although i am not a big rtt fan, and this is a bit I2ish, the data are
> bit
> interesting when you look at the std dev column.  looks to me like
> really good delay except for a few sick sites.
> randy

More information about the end2end-interest mailing list