[Babel-users] Rate limiting [was: New Version Notification...]

[Dave Taht]

On Wed, May 27, 2015 at 3:30 PM, Juliusz Chroboczek
<jch at pps.univ-paris-diderot.fr> wrote:
>> (by the time you wait for 10,000 nodes to show up in a deployment, it
>> would already be too late to fix)
>
> Just like you wouldn't deploy 10 000 OSPF or IS-IS nodes in a single area,
> you wouldn't deploy 10 000 Babel nodes without some form of aggregation.

Yes, that would be the right way, and a separate interesting problem.
Certainly distributing less p2p routes in favor of /64s and other
forms of aggregation (much like how bgp users now look askance at /24s
and the like) would become a problem that would be nice to automate in
favor of gradually (and algorithmically) filtering out more distant
p2p routes in favor of aggregate ones. there is a great deal of
activity of late in the bgp world experimenting with partial routes
and offloads from core routers (with limited table sizes) to other sdn
boxes (with full tables).

in the current microcosmic deployment of babel....

hnetd's default is not to aggregate the /60 or /56 it gets from
comcast but to announce all /64s on all interfaces and the /60 it
gets. My own network is big enough now for that (+ the /24s) to not
fit in a single packet, and I sometimes have gone through great (e.g.
anal) lengths to keep it to one, trying to get stuff to announce a
bare minimum covering route by default at even this scale.

(no I have not rolled out ss ipv6 outside the testbed yet)

> Or would you?

Heh. Would I today? No. but if you want to speculate (I am just having
fun), imagine a world where 802.11ad takes off, which has a VERY short
range in the spectrum it is allowed. Or UWB, which shares the same
problem, in a large set of organic, unmanaged, meshy deployments that
gradually build, over, say a decade. In both cases bandwidths are much
higher in potentia (as it is with koruza) in which the present hard
rate limiter (as opposed to one actually measuring the rate at which
updates could enter the network successfully) could slow down the
speed of propigation.

(this is increasingly not specific to the source specific bits of the
protocol, but full routing table updates in general)

>
> -- Juliusz



-- 
Dave Täht
What will it take to vastly improve wifi for everyone?
https://plus.google.com/u/0/explore/makewififast