The purpose of the Prefix Taxonomy Ongoing Measurement & Inter Network Experiment WG is to consider and measure the problem of routing table growth and possible interim methods for reducing the impact of routing table resource consumption within a network and the global Internet. The first step of the WG is to define the impacts on routing resource consumption and to identify the problems facing routing scalability.
The next step is to develop suggestions for filtering and aggregating prefixes to reduce an individual network's routing table size and route processing load and to suggest possible knobs that result in the least loss of reachability if such methods are determined to be feasible in addressing the problem. This work may possibly define a framework for larger efforts to address the problems facing interdomain routing scalability.
GOALS:
1) To provide a clear definition of the problems facing Internet Routing Scaling today. This includes routing table size and route processing load.
2) To provide a taxonomy to describe prefix information for peer review.
3) To collate measurements of routing table scaling data and publish a reference list.
4) To discuss and document methods of filtering/aggregating prefix information and to discuss and document what support from protocols or vendor knobs that might be helpful in doing this. In addition, to suggest policy guidelines to RIRs, LIRs and/or ISPs for allocations and aggregations,etc. that may be useful.
5) To determine the long and short term effects of filtering/aggregating prefixes to reduce router resource consumption.
6) To develop methods of controlling policy information propagation in order to limit the need for propagation of prefix sub-aggregates.
Some Relevant References:
http://www.antc.uoregon.edu/route-views/ http://www.pch.net/routing/BGP_table_size.html http://moat.nlanr.net/AS http://www.pch.net/documents/data/routing-tables/route-views.oregon-ix. n et/ http://www.employees.org/~tbates/cidr-report.html http://www.telstra.net/ops/bgp/index.html http://www.apnic.net/stats/bgp http://www.merit.edu/ipma
Nov 01 |    | Submit Taxonomy Internet-Draft |
Dec 01 |    | Submit Problem Statement Internet-Draft |
Dec 01 |    | Submit Filtering/Aggregation Internet-Draft |
Jan 02 |    | Submit References Internet-Draft |
Feb 02 |    | Submit Policy Propagation Internet-Draft |
Current Meeting Report
Minutes from ptomaine at IETF52
Randy Bush presented the agenda for bashing
Two drafts to discuss
draft-huston-nopeer-00.txt
http://buffoon.automagic.org/dist/draft-jabley-edge-policy-propagation-control-02.txt
Neither Geoff Huston nor Joe Abley were present to discuss their drafts
Randy asked if anyone else felt compentent to present either draft
No one responded
Randy asked about the future of the working group
If it is to continue, we need a new chair
Sean Doran: Geoff Huston and Joe Abley couldn't be here because of circumstances beyond their control, this is a transient glitch
Mark Knopper and Sean Doran volunteered to fill the role of co-chairs
=======================
Andre Broido presented an overview of variability in BGP announcements http://www.caida.org/outreach/presentations/BGP2001dec/
Questioner: Did you look at the ratio of average and variance of the AS path length distribution? (slide 8)
Andre: I have looked at the variance for these distributions, but I don't have the results here
Randy Bush: Are the statistics for the percentage of routes that are more specifics for just /24s or for all prefixes? (slide 12)
Andre: They are for all prefixes
In slide 15, the values in the table are percentages
kc claffy: Skitter is a CAIDA project that collects forward path information to a large number of destination addresses in the Internet using hop-limited probes (slide 17)
Sean Doran: There's a observational bias in the data collected from BGP tables compared to that collected from actual traffic. There should be significant differences between the connectivity seen by these two techniques
Randy Bush: A bunch of multihomed ASes announce on only one link? (slide 22)
Andre: Yes
Randy: Then this is probably load-balancing, announcing a more specific route to one provider
Ran Atkinson: The data collected from route views is fuzzy, and shouldn't be presented to three decimal places of accuracy (slide 23)
Cengiz Alaettinoglu: You will not see the less specific advertisments in all cases from Route-Views data, since many providers may prefer the other route, and fiter it from their advertisements
Randy: BGP is myopic
Sean Doran: What about oscillating prefixes? Do you think some of the transient entries may be persitently oscillating (slide 25)
Andre: Yes, we only see changes between our samples from Route Views every two hours. We are currently collecting every update, and will present results based on that data in the future.
FP/Kupt = Flips per 1000 uptime units in slide 28
AS752 is Merit, and their position near the top of the list may be due to ongoing research on BGP convergence times
Many of the others are large providers with many prefixes, or in countries outside North America and Western Europe (Romania, Columbia, etc.)
====================================
Sean Doran asked for additional obeservations and predictions to be sent to the mailing list
Slides
None received.