Re: IPv6 prefix lengths - how long?

[Michael Richardson <[e-mail]>]

Nick Hilliard <[e-mail]> wrote:
    > Some AMS-IX people wrote a paper about 10y ago about this - "Effects of IPv4
    > and IPv6 address resolution on AMS-IX and the ARP Sponge".  Their
    > analysis

Thank you for the pointers. my google results for the archives:

http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.182.4692 and
https://www.delaat.net/rp/2008-2009/p23/report.pdf
Effects of IPv and IPv address resolution on AMS-IX and the ARP Sponge

Abstract
  Heavy ARP traffic can cause routers to run out of resources. Therefore,
  AMSIX developed a daemon called ARP Sponge. When the number of ARP requests
  for an IP address exceeds a threshold, the ARP Sponge starts sending ARP replies
  which keeps ARP traffic down. The ARP Sponge currently works using IPv only.
  IPv uses Neighbor Discovery (ND) instead of ARP for address resolution. ND
  uses the solicited-node multicast address instead of broadcast, consuming
  less resources from routers. Therefore, IPv support should not necessary,
  though it still can be used to detect network problems and legacy router
  configurations.
  We tested this on an emulation of the AMS-IX platform, both the current
  version and its new version, yet to be introduced, based on MPLS/VPLS and
  with various customer platforms and found that IPv support need (and maybe
  should) not be implemented.

https://www.os3.nl/_media/2012-2013/courses/rp2/p57_report.pdf
Abstract
  The AMS-IX ISP Peering LAN connects more than 600 routers. These routers use
  ARP in order to establish connectivity with each other. With such a large network, it
  can be expected that there are nodes down at any moment. Even if that is not the case,
  there are also ARP Requests for IP addresses no longer in use. This causes severe
  broadcast traffic in the network.
  All the nodes that are listening for ARP messages have to further process the packets,
  meaning that all nodes in the network will spend CPU cycles to process
  these messages. This is inefficient, as most of the time those messages do
  not concern them,
  and some devices even prioritize ARP processing over, for instance, routing protocol
  jobs.
  So far, the solution that AMS-IX has implemented is the ARP Sponge. This
  tool reduces the amount of broadcast traffic on the peering LAN by replying
  to ARP requests for dead addresses.
  The goal of this research project is to investigate the use of OpenFlow to
  further bring down the amount of broadcast traffic on the AMS-IX peering
  LAN.