:: Re: [DNG] Request for information -…
Góra strony
Delete this message
Reply to this message
Autor: Jim Jackson
Data:  
Dla: Dan Purgert
CC: Dng
Temat: Re: [DNG] Request for information - - re: networking


On Tue, 6 Jun 2023, Dan Purgert wrote:

> On Jun 06, 2023, Jim Jackson wrote:
> > On Tue, 6 Jun 2023, Didier Kryn wrote:
> > > Le 06/06/2023 à 04:02, Dan Purgert via Dng a écrit :
> > > > [... a bunch of stuff describing general breakouts for 1k hosts
> > > > as an aside discussion to the OP's question ...]
> > >     Didn't you forget that all these sensors don't speak to each other,
> > > but they instead only speak with one single host. Given that, I'm not
> > > sure breaking down the traffic into many local loops would bring much
> > > improvement.
> >
> > From the OP's description of his proposed setup, I agree.
> >
> > Interestingly IPv6 over ethernet was designed to make it easier for
> > one lan to have most hosts - it uses multicast instead of broadcast so
> > it does depend on switches being able handle multicast reasonably
> > inelligently. In this case I suppose it could be possible to run the
> > setup using IPv6 link local addresses :-)
>
>
> You still end up getting inundated with ARP and other types of cruft. I
> haven't read anything that really indicates that v6 is any better at
> handling >1k hosts in a single broadcast domain than v4 is; but then
> again I also haven't kept as closely up-to-date with it as I did up til
> about 2018 or so.
>
> (references / new reading material would be appreciated ^_^ )


As far as I remember, IPv6 was designed to be more efficient on LANs. It
was designed when many LANs were still esentially collision domains!!!
For most of us "collision domains" are a thing of the past, though I still
have a couple of bits of gear that use thin ethernet and a thin ethernet to
Twisted pair repeater. But they haven't been powered up recently.

IPv6 doesn't use broadcast, it uses Multicast. ARP in IPv4 is replaced by
Neighbour Discovery in IPv6.

https://www.cisco.com/c/en/us/td/docs/ios-xml/ios/ipv6_basic/configuration/xe-3se/3850/ip6-neighb-disc-xe.html

The closest you get to a broadcast is sending data to the All Nodes
multicast group.

So the question is - how good are cheap switches at implementing multicast
at link level efficiently. So you need switches that support MLD snooping -
MLD being the IPv6 version of IPv4's IGMP. I understand that there is an
equivalence between IGMP and MLD.

It probably eventually means instead of number of MAC addresses a switch
can store for switching purposes, there may be some limit on space to store
where all the members of the multicast addresses are, on a large network.
Each new host creates at least one new multicast group.

I came across this ...

https://gist.github.com/njh/5c74614a92eb6c088ae9334db70df76a

but didn't see a date. It appears a lot of managed switches support
snooping but not sure about unmanaged cheap switches.

This discusses some of the issues ...

https://blogs.infoblox.com/ipv6-coe/how-many-ipv6-nodes-can-you-have-on-a-lan/