From: k claffy (kc@ipn.caida.org)
Date: Mon Feb 19 2001 - 12:24:16 PST
On Wed, Feb 14, 2001 at 05:18:45PM -0800, Neil Spring wrote:
I think there would be utility in being able to relate
flow endpoints by more than just their uniquely encrypted
IP addresses. There are bigger pictures out there.
I had a conversation with a colleague who is writing
tools to locate shared bottlenecks. This involves
finding network segments that are shared between flows
and correlating the behavior (latency) of those flows.
I suggested monitoring existing traffic to provide similar
measurements without introducing additional network load.
I'm explaining his work abysmally; the point is that
understanding network topology would be important for
deriving results and validating conclusions. Encrypting
IP addresses in a simple, highly anonymous way, removes
this information.
neil
you/your colleage might want to check out
http://www.caida.org/analysis/performance/bandwidth/
which is still in larval stages, but you
can get the idea
Internet routing practice is still strange and mystical to
me, so I don't know what information can be found easily,
how well such information reflects network topology, and
finally how to store such information in an anonymous
way.
However, if topology-related information can be preserved,
I'm all for it. I like having more information than I
know what to do with. In an ideal world, researchers like
us wouldn't have to have friends in the right places to
collect the information-rich trace data we need, it would
all be on MOAT's servers.
This archive was generated by hypermail 2b30 : Thu Sep 27 2001 - 16:24:41 PDT