North American Network Operators Group

Date Prev | Date Next | Date Index | Thread Index | Author Index | Historical

Re: Scalability issues in the Internet routing system

  • From: Alexei Roudnev
  • Date: Thu Oct 27 13:17:01 2005

If this 500K routes come from upstream, it is just _default_ so can be
installed instantly if configuration is correct.

If this 500K routes are from the peer, you switch (in reality) 10 - 20%, so
it is simpler anyway.

Even if it is multihome customer, there is not any need in _fast_
installation for these 500K routes. You just switch from one
provider to another _some_ of the routes - if it takes 1 minute, nothing
wrong happen.

Then, calculate:
500K routes, say 32 bytes/route (if not compressed by some way), 16MB.
T1 link, 100K/second, 160 seconds, 3 minutes.
100Mbit link, 10MB/second, 2 seconds.

T1 wil not be suitable for full routing of course, so what?

Just agaion - there are many tricks todo things right, out of theoretics of
IPv6 commitees.

----- Original Message ----- 
From: "Blaine Christian" <[email protected]>
To: "Lincoln Dale" <[email protected]>
Cc: "Alexei Roudnev" <[email protected]>; <[email protected]>; "Daniel Senie"
<[email protected]>
Sent: Wednesday, October 26, 2005 6:06 PM
Subject: Re: Scalability issues in the Internet routing system

> >
> > there have been public demonstrations of released routers
> > supporting upwards of 1.5M IPv4+IPv6 prefixes and demonstrations on
> > routing churn convergence time. <
> > document.asp?doc_id=63606> contains one such public test.
> >
> The
> site=testing&doc_id=63606&page_number=6 part may be a bit
> misleading.  For me it would be more interesting to see what happens
> when 500k routes completely disappear from the router then come
> back.   I want to see a 500k route push from a neighboring CRS in
> that amount of time...
> Of course the routes can switch quick when you use a layer of
> indirection (folks have been doing that for a few years now).  My
> question is how fast can you install routes from a standing start (or
> a 1/4 of a standing start if this is 2M prefixes).
> I will leave the question on whether it is actually worth an
> investment in time and resources as an exercise for the reader <grin>.
> Lightreading people,  test it like that!  It will be much more
> entertaining and perhaps even a bit enlightening to see how major
> vendors compare on "brand new" route installation into RIB and FIB.
> They only have to twiddle a couple bits to make indirection work
> quickly.  Having to deal with a brand new prefix is a completely
> different problem.