On Wed, Dec 31, 2008 at 4:40 PM, William Herrin <[email protected]> wrote:
> On Wed, Dec 31, 2008 at 1:30 AM, Tony Li <[email protected]> wrote:
>> |<[email protected]> wrote:
> Do the folks on this group clearly understand the harm that the
> Regional Internet Registries do by denying number resources to
> multihomed registrants based on our assurance that it's a necessary
> evil in order to keep BGP stable?

There's a fairly constant swing between the RIR folks saying: "Please
send out some guidance (/48, /56, /35, /32... pick pls) and the IETF
saying: "Well, we don't want to make something in stone/rfc else it'll
be a PITA to change later, and we KNOW it'll have to change
(ala-cidr-switch)"

I think folks, in general, get that hampering resource allocation ==
hampering network-growth == slowing the goodness that IP/Internet
is/could-be.  That said, there's pressure to make sure that we don't
blow up while trying to get to the end-state.

> If we still can't reliably quantify a BGP scaling limit then our
> request that the RIRs suppress BGP growth by suppressing resource
> assignment is simply unconscionable.

In 2005-ish at the IAB Workshop in AMS there was a good several
presenations on scaling/growth (one by Tony Li about hardware and why
moores-law doesn't apply to the gear in backbone networks).

>> We know of no hard upper bound to BGP or (more importantly)
>> the routing architecture as it currently exists.  It is apparent that unless
>> there is some significant progress somehow, the cost and/or complexity of
>> running the current architecture is going to start to climb.
>
> Climb to what?

'more expensive'.. I really don't think anyone can put a hard number
on this, but the running theory is that network devices will start
costing significantly more per device than they do today for the same
performance envelop PPS-wise, to handle a much larger routing
table/FIB.

(Keep in mind that not only is RIB size important but RIB
stability/rate-of-change is as well, and these two things hit two
different parts of the device today...)

>
> Assume a $2k COTS PC purchased 12/31/08 with component choice
> optimized for routing.
>

you can't make this assumption as 2k cots devices can't forward at
+100gbps, a core backbone device today can, and do in several
networks.

> Assume the ratio between BGP routes and updates per second will follow
> whatever growth pattern has been demonstrated in the ratio of routes
> to updates over the last 10 years. I expect this ratio is constant or
> near constant.
>

it's actually not constant, there are a number of factors (I think
Geoff Huston has some reasons/numbers here as well) involved. One
factor being 'laziness/incompetence/bad-actors' another being a
business model which is essentially arbitration of inbound traffic
costs via route updates...  Some of these reasons perhaps are constant
pains, some are new pains.

> Assume the PC builds a trie-based FIB from the BGP RIB. Trie based
> FIBs are known to exhibit a growth in resource consumption which is
> linear relative to the traffic switched, regardless of the size of the
> trie.
>
> How many routes can we pack in before we either fill memory, can no
> longer sustain both the 500mbps routing rate and keep up with BGP
> updates? Surely we can answer this question with engineering accuracy!
>

One point Tony made was that soon, perhaps, you won't be able to make
a lookup across the memory device holding the FIB fast enough to
service a packet on the fastest known interfaces today, presuming the
FIB grows in memory at some set rate (look at the graphs Vince Fuller
or Geoff Huston have for approximations of the rates).

-Chris
(IAB Workshop results RFC:
<http://www.ietf.org/rfc/rfc4984.txt>
workshop contents:
<http://www.iab.org/about/workshops/routingandaddressing/> )
_______________________________________________
rrg mailing list
[email protected]
https://www.irtf.org/mailman/listinfo/rrg

Reply via email to