Re: drop rates, graphs, triangles
Neal Cardwell (cardwell@cs.washington.edu)
Fri, 26 Jun 1998 10:08:33 -0700 (PDT)
Wow. On the face of it, this seems to say that for 40% of paths, we can
reroute the packets to cut the drop rate by at least a factor of four
(>=16% down to 4%), which should more than double the bandwidth that TCP
sees, all other things being equal. Am i getting that right? Of course
maybe once we put load on those alternate paths, they'll suck too... And
there's John's point that people with traceroute servers are not a random
sample of the internet -- they may be the ones who are so puzzled by their
sucky net performance that they set up a traceroute server to figure out
what's going on...
neal
On Fri, 26 Jun 1998, John Snell wrote:
>
> There's now an updated graph for packet loss. It's conservative, of
> course. See the top graph bullet.
>
> http://www.cs.washington.edu/research/networking/detour/local/triangle.html
>
>
> _____________________________________________________________________________
> "The human mind is a 400,000-year-old legacy application...and you expected
> to find structured programming?" -- Randall Davis, 1996 AAAI Pres. Address
>
>