Tech Notes

Addressing

VLANs

  • Most peering requirements can be accommodated on the layer 2 fabric without resorting to VLANs.
  • However, tagging and private VLANs can be configured where required, but only one MAC address will be accepted per physical port.
  • HFXIX will assign VLANs as required for private peering.

IPv6

  • IPv6 has special requirements:
    • IPv6 interfaces must disable auto config (Cisco “ipv6 nd prefix default no-autoconfig no-rtr-address”)
    • IPv6 interfaces must disable router advertisements (Cisco “ipv6 nd suppress-ra”) and should disable receiving them (Linux “echo 0 > /proc/sys/net/ipv6/conf/eth<N>/accept_ra”)

BGP

  • Participants are strongly encouraged to establish BGP peering with the route servers in order to gain maximum benefit. Route distribution can be controlled with communities (see below).
  • BGP sessions should be established with the HFXIX ASN 13770.
  • There are two route servers for redundancy: 206.130.15.2 and 206.130.15.3 for IPv4, and 2001:504:37:10::20 and 2001:504:37:10::30 for IPv6. BGP sessions should be established with both for consistency and reliability.
  • BGP exports are unchanged from the import. Communities, AS paths, etc. are exported as they were imported. This means that AS13770 will not show up in exports, and modern routing engines will object to the BGP peer AS being absent. To avoid this check in Cisco IOS, add the statement “no bgp enforce-first-as” to the BGP configuration. Other servers (Quagga, JunOS, etc) have similar statements.
  • Communities can be used to control how your routes are advertised. Route Server communities are applied in the table order.
    Communities 
    0:peer-asPrevent announcement of prefix to AS peer-as
    13770:peer-asAnnounce prefix to AS peer-as
    0:13770Prevent announcement of prefix to all participants
    13770:13770Announce prefix to all participants (default action)