IPv6 Transition: Update for Operators & Providers

Bob Gilligan

Table of Contents

• Stan Barber's Notes
• February 1996 NANOG Home Page

Stan Barber's Notes

Factors Driving Transition

• 1.4 billion out of 3.8 billion IPv6 addresses are assigned.
• When will they run out?
  • No consensus prediction
  • Estimates cluster at 2005 +/- 5 years
• Little user preception of addressing problem
• Future address comsumption
  • New low-end Internet Devices Low cost = high volume = lots of addresses
• Internet extension to the home
  • Multiple Internet hosts per home
• Transition must be mostly completed before address run-out
  • All notes requiring global connectivity should be converted before run-out to avoid discontinuity
• Delaying start of transition compresses the transition period
  • Same amount of work, less time to do it.
• New Features
  • Address auto-configuration
    • Plug-and-play (user benefit)
    • Eases re-numbering (operators benefit)
  • Multicast support standardized
    • All IPv6 routers support multicast
    • Unicast and Multicast topologies are the same (operators benefit)
    • Multicast applications work everywhere (users benefit)
  • IP Layer security is standard (users benefit)
  • Flow Support (users and providers benefit)
    • Auto/video conferencing becomes possible.

IPv6 Implementation & Testing Status

• Host products
  • Digital Unix
  • Digital VMS
  • Solaris 2
  • HPUX
  • Streams
  • FTP Software
  • BULL
  • BS2000
• Router products
  • Digital
  • Bay Networks
  • Cisco
  • Ipsilon
  • Penril
  • Telebit
• Research implementations
  • 4.4 BSD (Inria)
  • 4.4 BSD (NRL)
  • BSDI (WIDE)
  • Linux

• IPv6 Testing
  • First IPv6 test event held last week at University of New Hampshire.
  • Ten implementations participated.
  • Basic IPv6 functionality tested.
  • Next test scheduled for June.

Transition Design

• Avoid upgrade dependencies (e.g. don't make it necessary to upgrade node x before node y)
• All incremental upgrade
  • Must be able to upgrade one machine at a time
  • IPv6 will become available in different products at different times
• Let site transition at their own pace
• Provide options
  • One size does not fit all
• Make it easy for the end-user

Transition Mechanisms

• New nodes support both IPv4 and IPv6. (All hosts and routers should do this.)
• Same transport layer and applications above both IPs.
• Provides complete interoperability with IPv4 installed base.
• Upgrading nodes from IPv4 to V4/V6 dual doesn't break anything.
• IPv6 over IPv4 tunnels will be used to build the initial IPv6-based Internet
• Configured Tunnels (much like the current MBONE)
• Automatic Tunnels (used a address resolution protocol with an embedded IPv4 address)
• The IPv4 compatible address structure in IPv6 has the IPv4 address prepended with 96bits of zeros. Is not able to take advantage of advanced IPv6 addressing capabilities.

Transition Planning

• Maintain complete V4 routing system until run-out
• Upgrade IPv4 router to dual V4/V6
• Incrementally build up V6 backbone routing system
  • Use V6-over-V4 runnels to construct V6 backbone
  • Grow like the MBONE
• Shutdown V4 in areas where there is no need for V4.

Initial options for operators

1. Do nothing
   • Wait and see
   • Let V6 users tunnel across you
   • Can't provide V6 service or V6 addressing
2. Build a tunnel V6 infrastructure over your V4 system
   • No change to internal V4 routing
   • V4/V6 routers at edges
   • Can provide V6 service and addressing
3. Do V6 fully

Initial options for users sites

Summary

• IPv6 implementations are under way
• Transition is flexable -- Understand your options
• Start transition planning early

Questions & Answers

Is Sun going to put CIDR in the kernel? Solaris 2.6 will have that in it.

What about routing protocols for IPv6? Long term, the IPv6 address space can contribute to having more hierarchy.