[e2e] e2e delay characteristics

Konstantina Papagiannaki Dina at sprintlabs.com
Mon Aug 18 15:02:09 PDT 2003

This is not exactly what you are looking for, but it may give you an idea of 
what the delays look like for packets that transit the core of a backbone 
network. At http://ipmon.sprint.com/delaystat/ you can see the distribution of 
delays experienced through the Sprint IP network. Equal Cost MultiPath 
generates sometimes distributions with multiple modes, depending on how many 
equal paths exist through the network and what their propagation delays are 
(an example is http://ipmon.sprint.com/delaystat/delay.php?date=20021121&linkpa
ir=results/20021121/sj-01.0-to-nyc-27.0). There is very little evidence of 
congestion in the core, so if you are looking for congestion and loss, this 
type of statistics are definitely not what you are looking for:-)

Hope this helps...


> Hi all-
> I'm interested in having a good model of e2e delay, delay variation, and
> loss. I've been using Bolot's 1993 paper (End-to-end packet delay and loss
> behavior in the internet) as my primary source, but I'm concerned that 
> things may be different a decade later.
> I've looked at Vern Paxson's thesis and from that (ch. 16), it also
> appears that Bolot's suggestion of an exponential variation in the delay
> does not match the characteristics that Paxson found.
> Does anyone have some good references for this? I'm primarily interested 
> in a steady flow of packets with about 50 to 500 ms between each packet, 
> rather than single probes or file transfers. I will use what I can get, 
> however.
> thanks for your time-
> Matt

More information about the end2end-interest mailing list