Print

[NewtonNet]

Thinking aloud for a moment, I've got a question relating to network design which I'm hoping you might be able to shed some light on....

In IPv4, one has to take into consideration broadcast traffic when designing LAN sizes and limit the number of connected devices to any one segment/subnet/VLAN. Hence, with an absence of broadcast traffic in IPv6 does this mean that these LANs can now be made bigger?

Of course LAN segmentation may be desired for other reasons - security, physical topology, etc - but I'm wondering specifically how this one aspect changes things (if at all).

Mathew

[kcochran]

It doesn't change it too much.  What used to be handled by broadcasts is now handled by multicasts.  To a dumb switch, there's no difference between the two.

[NewtonNet]

Ah okay. I thought this might be one of the 'selling points' of v6 at the LAN level. 

To the devices, however, surely their stacks are less worked given they don't have to unwrap every broadcast packet now that they only need to concentrate on their own unicast/multicast addresses?

Mathew

[kcochran]

Minor improvement, but nothing I'd consider as enough of a gain to permit sticking more hosts on a v6-only LAN than a v4-only LAN.

[NewtonNet]

Gotcha - thanks for that.

[bombcar]

In line with that, if you have a /64, say, from a tunnel, is it simple to slice it up internally and hand out /96s or something?

[jimb]

In line with that, if you have a /64, say, from a tunnel, is it simple to slice it up internally and hand out /96s or something?

Simple, but not recommended as it breaks autoconfiguration for that LAN, and a bunch of other potential issues.  There's debate about using /112 or /126s for p-t-p connections, and /128s for loopbacks, but many still recommend sticking with /64s for those too.

If you need more subnets, request a /48.  That's pretty much what ISPs will be giving end users at home (or possibly /56s depending on the ISP).  Multi campus/site businesses will get multiple /48s, ISPs get /32s, etc.  There's been a lot of discussion about this on NANOG and ipv6-ops lists lately.  

May want to look at a this recent post, also RFC3177 has some guidelines about this.

The meat of the RFC here:

  In particular, we recommend:

      -  Home network subscribers, connecting through on-demand or
         always-on connections should receive a /48.
      -  Small and large enterprises should receive a /48.
      -  Very large subscribers could receive a /47 or slightly shorter
         prefix, or multiple /48's.
      -  Mobile networks, such as vehicles or mobile phones with an
         additional network interface (such as bluetooth or 802.11b)
         should receive a static /64 prefix to allow the connection of
         multiple devices through one subnet.
      -  A single PC, with no additional need to subnet, dialing-up from
         a hotel room may receive its /128 IPv6 address for a PPP style
         connection as part of a /64 prefix.

   Note that there seems to be little benefit in not giving a /48 if
   future growth is anticipated.  In the following, we give the
   arguments for a uniform use of /48 and then demonstrate that it is
   entirely compatible with responsible stewardship of the total IPv6
   address space.


-jim