Please send me your comments

[Philip Smith]

Hi Cathy,

At 13:18 02/11/00 -0800, CJW wrote:
>Hi everyone out there in rtma-land,
>
>I think it's time I try again to get you all involved.  I'd like to
>come up with a list of agenda topics and perhaps meet as a working group
>at the next ARIN meeting.  So far some topics I can think of are:
>
>ways to summarize the routing table info:
>    a) discuss Scott Marcus's info on AS number depletion
>       (perhaps compare it to Geoff Huston's data presented
>        last week at the APNIC meeting)
>    b) how do assigned ASNs compare to the ones showing up
>       in the routing table?
>    c) how many prefixes are being assigned per ASN?

A study of ASNs usage/assignment/profile/depletion would be good. And like 
Randy suggested, a hint in the direction of idr to move from 16 bit ASNs 
to >16 bit ASNs within the next 4 years or so would be useful. Failing 
that, efforts to reclaim the ASNs assigned but unannounced?

>    d) perhaps compare registered prefixes and length to
>       advertisements.  For example how many advertisements
>       are generated for each assigned prefix?

It would be interesting to see this,... It would prove that the degree of 
multihoming is increasing and dare I say the level of aggregation clue is 
decreasing?

>    e) perhaps query experts on when the existing network
>       equipment will no longer be able to handle the size of
>       the routing table (not just # of prefixes but number of
>       paths, not just amount of memory required, but what happens
>       if there are flaps)

I really think the only way we are going to find this out is by trying it. 
We all can do theoretical studies but real life tends to throw up its own 
little tests... If anyone has done analysis work of BGP convergence rates 
versus CPU for the major platforms, it would be interesting to hear...

>Any other thoughts?  What info should we be gathering? (if any)
>
>Thanks for your help!
>---Cathy

Other ideas...

- It might be interesting to see information or summaries about the number 
of BGP announcements per minute in various parts of the "default free 
zone". And a summary of what caused them (infrastructure/power switch/human 
error) - or at least an attempt. Even an attempt to correlate that BGP 
churning in different parts of the Internet. RIPE NCC's RIS springs to mind 
as a possible site which is gathering this sort of info (but I don't think 
they are analysing this particular aspect).

- Might be worth discussing the importance of BGP flap damping, if any? 
Like how many providers are doing so, what parameters are they using. 
Comparing route processing performance on non-flap damped networks with 
those which have flap damping at the edges... RIPE-210 documents 
recommendations from RIPE Routing WG, but maybe the net needs to be widened 
to figure out if these are satisfactory, could be improved, &c... Can 
vendors to anything to improve in this arena - can BGP be improved any to 
improve stability... etc.

- Maybe expert discussion on why appearing in Tony's CIDR Report no longer 
has the stigma it used to have? Does it even matter any more? (Most new 
ISPs have never heard of it, and even then many don't understand what it is 
trying to achieve...) Do we just assume that the whole routing thing will 
keep on working, just keep beating on the vendors to produce ever bigger 
faster pricier boxes to keep up? :) We started talking about this here in 
Brisbane last week, might be worth carrying it on...?

some ideas...,

philip
--