[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[RRG] Re: Not moving the problem to the global mapping system



Brian E Carpenter wrote:
On 2008-04-10 17:54, Michael Meisel wrote:
...
I think we should ask this question in terms of the problems we are
trying to solve. As we all know, one of these problems is the frequency
of BGP updates today, which are largely coming from edge network. To put
it a different way, the problem is that the reachability information for
even the smallest of edge networks is announced globally. Each update
requires a significant amount of processing by each node that it passes
through. So the mapping system better not suffer from this same problem,
or we haven't done any good!

So, specifically, we should be asking:

(1) Is the mapping function successful in preventing edge network
reachability from being propagated into the global routing system?

(2) If yes, does it do so without simply moving the problem to the
global mapping system?

Note that proposals that both (a) put reachability information into the
mapping system and (b) involve any sort of push model start to look a
lot like BGP, and therefore are going to have a hard time answering
"yes" to (2) convincingly.

I can't say how much I agree with this. If the mapping system
degenerates into a reachability-driven routing system, we might
just as well switch to two layers of BGP immediately.

I think I wasn't clear, because it sounds like we very much agree. I think we should expect a "yes" answer to both questions, and a mapping system like you describe is exactly the sort that would have trouble doing so.

I would suggest turning this into a concrete goal, such as:

<strawman>
The update rate in the mapping system should be at least
two orders of magnitude less than the update rate in
the BGP4 system, at any point in time.
</strawman>

Yeah, I think something like that is reasonable, though I think "update rate" is probably both too vague and too narrow a metric. We need to take into account the order of magnitude in number of devices doing the processing, processing required per update, and probably other things I haven't thought of.

But I still wonder -- is it truly a useful step towards solving the scalability problem to develop a list of such constraints, in and of itself? Or is it only useful if we use those constraints to evaluate concrete designs? I'm tending towards the latter.

-Michael


--
to unsubscribe send a message to rrg-request@psg.com with the
word 'unsubscribe' in a single line as the message text body.
archive: <http://psg.com/lists/rrg/> & ftp://psg.com/pub/lists/rrg