Mikael Abrahamsson <swm...@swm.pp.se> wrote:
>> Can you construct such a network out of laptops, desktops, home
>> routers and NAS? Each have one GbE port and a WIFI interface?
> There are already announced services that are more than 1GE:
>
http://www.pcworld.com/article/2947712/networking-hardware/comcasts-2-gigabit-service-will-cost-300-plus-1000-for-installation-and-activation.html
> So if this is now the case, why would I not want to be able to put
> 2x1GE between my homenet routers to actually make use of this
> bandwidth?
Today, I would manually bond (LACP) those two interfaces at layer-2, which
would result in layer-3 seeing one link, therefore no ECMP needed in the
routing protocol. Not zero-touch for sure!
If we assume that (auto)configuring LACP is hard, and want to do in the
routing protocol, I absolutely buy your situation; and your gigabit uplink
while gigabit to NAS makes total sense to me.
It could be that we detect multiple single hops between routers in either the
routing protocol or the HNCP, and we would put both links into the same
broadcast domain. (The architecture says we are supposed to do that)
Given that there might be an arbitrary amount of homenet-unaware L2-switching
gear between A and B, that seems to imply that one should enable LACP
negotiation in this situation, and it might succeed, or it might not, and if
it fails, then ECMP at layer-3 would make sense.
My impression is that from the point of view of router F (connected behind
your node B), that it would not see any additional adjancy in the routing
TLVs that it sees from B. Maybe the metrics would be better than if there was
only a single link.
What I don't see is a situation where there are actually two equal cost
paths (multiple hops, some not identical) between two end-points within the
home, such that there would be a benefit to ECMP.
In your ABCD+NAS situation, add a wifi link between the NAS and NAS client C.
Assume that in the future OS on C is willing to upgrade to get better
functionality. My claim is that C will get better benefit from implementing
MPTCP for the connection to the NAS, so that it can go faster when plugged
in, and yet still function if all it has is wifi... LISP is also an
alternative for the C<-->NAS traffic. But, if the GbE is plugged in,
(and especially if as you suggest A==B is >1GbE), then I don't see C<-->NAS
traffic ever desiring to use wifi at the same time as wired.
{I have 50Mb/s DSL2 at home now, and in anticipation, I had previously
replaced my 24-100Mb/s "smart" switch with a major-brand 24-port GbE SOHO
managed switch in 2012... Damn thing actually can't keep up with back to back
packets between NFS server and desktop... even at 100Mb/s... Turing QoS *OFF*
reduced my drop-rate from 1 in 2, to 1 in 10... Ridiculous.}
--
Michael Richardson <mcr+i...@sandelman.ca>, Sandelman Software Works
-= IPv6 IoT consulting =-
signature.asc
Description: PGP signature
_______________________________________________ homenet mailing list homenet@ietf.org https://www.ietf.org/mailman/listinfo/homenet
