IRTF Routing Research Group
31 July 2000, 1:00pm.
==========================
notes by jhawk and ahuja
Intro - abha
------------
Who we are. What we are.
Reviewed list of topics.
Added load-based routing to the list of topics.
Topic list:
routing convergence, stability and scalability
fault tolerance
Quality of Service routing
multicast routing
Extremely dynamic contraint-based routing
Traffic engineering
NAT and IPv6 routing
optical networks and routing
operational concerns of routing
load-based routing
Goal of this meeting: introduce who we are, what we are about and
encourage discussion about routing issues.
Sean's Talk
-----------
IRTF generic.
Steal from the end2end research group model
Four obvious streams of routing research
a) current behavior
b) projections of lifetimes
c) what next
d) broad sugggestions
Noel: group could recommend to IAB and IESG as to how we should procede.
need to make fundamental change.
IAB rubberstamping/charter-adjustment stuff.
DWDM - limit of 40 giga-bits/channel, up to 512 channels per fiber.
parallel but not fate-sharing links
2 approaches:
a) expose all channels to routing system
b) abstract all
links into one huge link/line.
ugly scaling issues: too much state? inefficiency and error recovery.
Curtis: solved problem issue? explicit routing/mpls bandwidth
advertisement.
Layer integration. Optical integrating with ip/sdh etc.
Control plane: many more routers to configure, etc. hundreds of
thousands of devices.
"Swedish Problem": 3 million households, 5mbps full-time bandwidth to
each household. Lack of ability to change ISPs, etc.
Other jurisdictions [cable, gas, etc.]. Can this stuff scale?
"An exchange point in every neighborhood?!" "opportunity cost" of
multihoming is getting very cheap!
Geoff: concern about carrier motiviation to have resiliency. carriers
have no motivation to provide resiliency.
ALE-type expectancy of the routing system?
Noel: TANSTAAAFL for multihoming. Lack of cost-feedback for multihoming
via the routing system.
Curtis: aggregation lossage. Aggregation is not as good as it could be.
Aggregation across providers.
Topic in current routing research:
=================================
Convergence - abha
------------------
Measurement of BGP Convergence. Craig Labovitz, Abha, Farnam Jahanian,
abhijit Bose.
Path vector algorithms slow to convergence.
psuedobeliefs:
"Restoral not an issue...
bad news travels fast
aspath limitting fixes converegence
more band
convergence study how? failure/repair data?
factorial-space convergence problems happen in the 6bone!
Curtis: "link state sucks too". dkatz presentations. flooding scaling
problems.
BGP Table Analysis - Geoff
-----------------
Geoff Huston:
http://www.telstra.net/ops/bgp.html
exponential routing table growth since 99.
unique ASes.
/32s per AS from 16k to 13k in 9 months.
33k-routes covered by other aggregates.
only 10% aggregatable by telstra in AU.
PIARA brief discussion.
Observation of 100mbps deployment in 1998.
See the web page for tables, etc.
Nimrod - Noel
-------------
Noel: "String theory described as 21st century physics that fell into the
the 20th century; nimrod as 21st century routing fell into the 20th
century." { Nimrod talk }
Explicit routing architectures: nice testing, one node can update
algorithms.
Topology and abstraction are TWO DIFFERENT THINGS.
No. 1 goal: extend lifetime of the system.
Van: Architecture has to match with business models.
//abha ;)
++++++++++++++++++
abha ahuja
<ahuja@umich.edu>
*teehee*
This archive was generated by hypermail 2b29 : Mon Aug 04 2003 - 04:10:04 EDT