Howard,
We are purposely hoping/intending these requirements
to lead to new, different, novel, ... architectures
and protocols. Things like draft-ietf-bmwg-conterm-01.txt,
for example, are written from the perspective of the
current architecture and protocols (in particular BGP).
If we use them as a basis for the requirements document
we may tend to bias people to thinking "they just want
to hack BGP a bit more" (ok, I'm being a bit overdramatic
here to make the point) when in fact what we want is the
exact opposite.
Frank Kastenholz
At 12:40 PM 3/5/02 -0500, Howard C. Berkowitz wrote:
>Quoted material from document.
>>2. System-wide
>>Defining a single target for maximum convergence time for the real
Internet is absurd. As we mentioned earlier, the Internet is large enough
and diverse enough so that it is quite likely that new changes are
introduced somewhere before the system fully digests old ones.
>
>Do we have an agreement on what convergence time means? In the current BMWG
work, we have tried to make that definition more rigorous, and also
recognize there are several kinds of convergence time with different scopes.
These different times need distinct names to avoid confusion. While
draft-ietf-bmwg-conterm-01.txt is officially focused on BGP, it really does
try to come up with some fairly general terms.
>
>>
>>So, the first requirement here is that there must be mechanisms to limit
the scope of any one change's visibility and effects. The number of routers
that have to perform calculations in response to a change is kept small, as
is the settling time.
>
>Suggested clarification:
> The Internet consists of a multitude of scopes (Sx) of information. A
scope
> [perhaps a better name is needed] is an area that is aware of detailed
> topology, other constraints on that topology, and changes within that
area.
> A scope contains Rs routers and a variable number, En, of basic network
> elements E (such as prefixes or links/path)
>
> It is a requirement that the convergence time of a scope be constant,
> within a range TBD.
>
> This implies that convergence time either:
> 1. Is not a function of the number En, but of Rs.
> 2. The overall mechanism will automatically divide a scope
> that grows too large to meet the convergence time objective,
> without impact on the administration of that scope or its
> connectivity to other scopes.
>
> There may be a notion of partial convergence within a scope, in which
> it can be said that for Z% of the changes (with average interarrival
> time Tc), within Y% of the Tc seconds, X% of the routers have converged
>
>**** let's either formally define "settling" or call it a specific kind of
> convergence *****
>
>**** and if En is TBD, does that imply that all routers have the same
> control plane processing power? ******
>
>>The second requirement is based on the following assumptions
>>- the scope of a change's visibility and impact can be limited. That is,
routers within that scope know of the change and recalculate their tables
based on the change. Routers outside of the scope don't see it at all.
>>- Within any scope, S, network changes are constantly occurring and the
average inter-change interval is Tc seconds.
>>- There are Rs routers within scope S
>>- A subset of the destinations known to the routers in S, Ds, are impacted
by a given change.
>>- We can state that for Z% of the changes, within Y% of Tc seconds after a
change, C, X% of the Rs routers have their routes to Ds settled to a useful
answer (useful meaning that packets can get to Ds, thought perhaps not by
the optimal path -- this allows some 'hunting' for the optimal solution)
>>X, Y, Z, ARE TBD
>>This requirement implies that the scopes can be kept relatively small in
order to minimize Rs and maximize Tc.
>>
>>The growth rate of the convergence time MUST NOT be related to the growth
rate of the Internet as a whole. This implies that the convergence time
either
>>1. Not be a function of basic network elements (such as prefixes and
links/paths), and/or
>>2. That the Internet be continuously divisible into chunks that limit the
scope and effect of a change, thereby limiting the number of routers,
prefixes, links, and so on involved in the new calculations.
>>The growth rate of the convergence time MUST NOT be related to the growth
rate of the Internet as a whole. This implies that the convergence time
either
>>1. Not be a function of basic network elements (such as prefixes and
links/paths), and/or
>>2. That the Internet be continuously divisible into chunks that limit the
scope and effect of a change, thereby limiting the number of routers,
prefixes, links, and so on involved in the new calculations.
>
>
>I think we need a more specific definition of "growth rate of the
internet".
==================================================
My preferrred signature is:
This information is for the sole use of
whoever receives it and may contain confusing,
enlightening, enraging, entertaining,
irritating, or just plain stupid information,
including without limitation, double-secret-
probation information belonging to [CENSORED
BY THE NSA/FBI/MOUSE]. Any unauthorized review,
use, disclosure, or distribution outside of an
establishment serving alchohol is prohibited on
days that do not end in Y.
But our ******'d lawyers would rather have:
=======================================
This email message is for the sole use of the intended recipient (s) and may
contain confidential and privileged information, including without
limitation, Confidential and/or Proprietary Information belonging to
Unisphere Networks, Inc. Any unauthorized review, use, disclosure or
distribution is prohibited. If you are not the intended recipient, please
contact the sender by reply email and destroy all copies of the original
message.
This archive was generated by hypermail 2b29 : Mon Aug 04 2003 - 04:10:04 EDT