[Babel-users] jitter and delay in babel

[Dave Taht]

On Mon, Jul 2, 2012 at 1:04 PM, Juliusz Chroboczek <jch at pps.jussieu.fr> wrote:
>> wifi already has enough jitter and delay built into it already, IMHO,
>> as well as means to arbitrate contention to the medium.
>
> No.  The natural jitter of wifi will not prevent synchronisation of
> different routers.
>
>     Floyd, S. and V. Jacobson, "The synchronization of periodic routing
>     messages", IEEE/ACM Transactions on Networking 2, 2, 122-136, April
>     1994.

While I don't necessarily agree yet, I will read. Natural sources of jitter on
wifi include rate changes, interference, retries, competing traffic,
multicast related sleep until power wake up delays, and so on.

>> I was contemplating removing (or making iface type aware) various
>> usleeps in the code,
>
> Huh?

I just grepped, saw usleeps.

> Babeld is event-driven, not threaded, so it doesn't use usleeps.  (The
> small number of usleeps in babeld.c are not in the main loop.)

k.

> What is more, the delays used by babeld have a second purpose -- they
> allow aggregation of multiple TLVs into a single packet.  By removing
> the delay, you'd break that mechanism.

Ah. I do like the shortness of babel routing messages compared to
other protocols that leverage arp...

>> discuss?
>
> Not until you do your homework.

Thank you for doing it for me.
>
> -- Juliusz



-- 
Dave Täht
http://www.bufferbloat.net/projects/cerowrt/wiki - "3.3.8-6 is out
with fq_codel!"