Re: Mail regarding draft-litkowski-rtgwg-spf-uloop-pb-statement

[Martin Horneffer <[e-mail]>]

Hello Stephane, Mike, and group,

just one comment (see below) since I get to see more and more different 
networks from the operator's point of view.

Am 11.05.15 um 16:22 schrieb [e-mail]:
> 4. "   Routers have more and more powerful controlplane and dataplane that
>
>    reduce the Control plane to Forwarding plane overhead during the
>    convergence process.  Even if FIB update is still reasonably the
>    highest contributor in the convergence time for large network, its
>    duration is reducing more and more and may become comparable to
>    protocol timers.  This is particular true in small and medium
>    networks."
>
> I don't understand what is meant by "may become comparable to protocol 
> timers"? Are you suggesting that the FIB update latency WAS greater 
> than the protocol timers, but has now been reduced to a comparable value?
> [SLI] Right, even if it may be not true for all the networks, this 
> tends to be the case
>
>
> The reference to small and medium networks is also confusing, since in 
> my experience it is actually the small and medium networks which are 
> subject to the LARGEST FIB update times as a result of the deployment 
> of under powered hardware.
> [SLI] Yes and no …
>
> I may say that small/medium networks have less powerful hardware, but 
> also less routes (except badly designed networks J). Large network 
> have more powerful hardware but more routes to handle.
>
This really depends on many corner conditions such as exact choice and 
age of hardware, network design in term of topology and routing 
architecture and, of course, size of the network.

While it might be possible for a small or medium sized network with a 
very clean design, a small number of internal routes, and modern 
hardware to have a very fast FIB update, there are numerous reasons why 
this could fail: old hardware or software, network design which includes 
many parts of the aggregation, access and service generation area in the 
IGP, and/or routing architectures which for one reason or the other 
include some service routes in the IGP.
For very large networks on the other hand it will definitely not be 
possible to keep FIB updates very fast.

IMHO it would be a good idea to remove  the reference to the size of the 
network. And don't even try to specify which kind of network has shorter 
or longer FIB update times. Just indicate that depending on size and 
exact design of a network it MAY have short FIB updates times, but make 
clear that by no means this is always the case.

Best regards, Martin