IPv6 Migration

Question 1

IPv6 migration approaches

Answer 1

• Dual Stack
• Dual Layer

Question 2

Dual stack migration approach

Answer 2

Dual Stack

• Both capabilities in all entities supporting v6
• v4 support can be graudally removed
• Complete duplication of all protocol stack components
  
  • Routing protocols, tables, access lists
• It doesn’t reduce the need for more v4 addresses
• Application need to choose between v4 and v6.

Question 3

Dual layer migration approach

Answer 3

Application are not responsible for choosing protocol to use: TCP/UDP selects most suited protocol (for example using information from DNS records (Type A, AAAA))

Less modifications in the applications

Less used than dual stack.

Question 4

Travering IPv4 only network

Answer 4

Tunneling:

• Encapsulating v6 packets in v4 packets emulating a directs link
• End-point: hosts and routers
• Protocols:
  
  • GRE
  • IPv6 in IPv4
• Setup: manual and automatic
  
  • IPV4-compatible addresses
  • 6over4
  • 6to4
  • Tunnel broker
  • ISATAP
  • Teredo

Question 5

Host centered vs Networks centered vs Carrier-Grade solution list

Answer 5

HC solutions: Dual stack hosts exchange IPv6 packets through an IPv4 network

• ipv4 compatible addresses
• 6over4
• ISATAP

NC solutions: Dual stack and native v6 hosts exchange v6 packets through a v4 network.

• Teredo
• Tunnel Broker
• 6to4

Scalable, CG solutions: Native IPv6 (ipv4) hosts exchange IPv6 (ipv4) packetrs through IPv4 (IPv6) network.

• DS-lite
• A+P
• MAP-T and MAP-E
• NAT64
• 6PE (MPLS-based)

Question 6

ipv6 migration: ipv4 compatible addresses

Answer 6

improperly called automatic tunneling

ipv4-compatible addresses (::/96) are used for v6 communication

process:

1. app send ipv6 packets to ipv6 addresses
2. static route forewards packets to ::/96 through pseudo-interface
3. pseudo-interface encapsulates v6 packets in v4 packets and sends them out

If configured with dual stack router:

1. pseudo-interface in sending host in is configured to terminate tunnels on router
2. encapsulating v4 packets are sent to v4 address of dual stack router

Question 7

6over4

Answer 7

• Host cenetered solution in ipv6 transition
• ipv4 networks emulates a virtual LAN
  
  • broadcast multiple access data link
  • ip multicasting used for the purpose
• Neighbor and router discovery enabled
• v4 address is used for automatic v6 interface ID generation of link local address
• Not very used because of the lacking support for v4 multicast

Question 8

6over4 neighbor discovery

Answer 8

• ipv6 mutlicast addresses are mapped on ipv4 multicast addresses
  
  • 239.192.x.y where x.y are last 2 bytes of ipv6 address

Question 9

ISATAP

Answer 9

• Intra-site automatic tunnel addressing protocol
• ipv4 netowrk as non-broadcast multiple access data link (NBMA)
  
  • No IP multicast suppport
• Interface Id derived from lpv4 address
  
  • Prefixed by 0000:5efe
  • fe80::5efe:0101:0101
• No neighbor discovery: not needed for data-link address discovery as ipv4 address is embedded in ipv6 address
• PRL (Potential Router List) must be provided: router discovery is not possible
  
  • List usually obtained to DNS query to isatap.domain.com (usually domain is obtained through DHCP) or through configuration
  • Each router in the list is infrequently probed to check if they are working and to perform unicast-only configuration.

Question 10

6to4

Answer 10

• Network centered solution for ipv6 migration
• Relay v4 address is embedded in ipv6 prefix by appending it to 2002::/16
  
  • 192.0.2.4 -> 2002:c000:0204::/48
• Basic scenario: not meant for ipv4 host to ipv6 host communication
• Mixed scenario (ipv4 global internet + ipv6 gobal internet): 6to4 relay must be default gatway (192.88.99.1) of 6to4 routers to handle cases of prefix not matching 6to4
  
  • Default gateway (6to4) realy: to be used whener I don’t know what to do with an ipv6 packet (prefix not matching 6to4 prefix)

Question 11

Teredo

Answer 11

• Solution to ipv6 transition with network centered approach
• Architecture similar to 6to4
• IPv6 packets are encapsulated in IP/UDP to be compatible with NAT, since NAT cannot use ports with IPV6 in IPv4.

Question 12

Tunnel Broker

Answer 12

• Solution to ipv6 transition with network centered approach
• Solves the problem of 6to4: I cannot use a general ipv6 prefix
• Tunnel broker server:
  
  • Indentifies tunnel servers (routers) and mediates tunnel setup (gives us the ipv4 addresses to contact)
• IPv6 in IPv4 tunnels
• Tunnel setup Protocol/Tunnel Information Control: protocol used to setup tunnels
• 6to4 doens’t support any prefix, but is completely decentralized, doesn’t need central server, here we need it.
• Who is in charge of the server (who pays for it) when it is shared between ISPs? big problem.
• Pretty similar to ISATAP

Question 13

Goal of Scalable, carrier grade solution to ipv6 transition

Answer 13

Goals:

• need to support
  
  • ipv4 servers possibly communicating with ipv6 hosts
  • ipv4 clients

Question 14

AFTR

Answer 14

Address family transition router:

• allows ipv4 hosts to communcate wit ipv4 hosts over an ipv6 carrier infrastructure
  
  • residential hosts and current providers
• features ipv6 tunnel concentrator and possibly a large-scale NAT
• in use in DS-lite and A+P

Question 15

DS-lite

Answer 15

Dual stack lite

• Dual stack at the edge of the network
• IPv6 only service provider backbone
• Properties:
  
  • Reduces requirement for ipv4 addresses compared to dual-stack approach
    
    • dual-stack requires an ipv4 public address per host
  • Extended NAT enables customer assigned addressing
    
    • IPv6 address of CPE in NAT table
• Limitation:
  
  • NAT is not under control of customer (same as with CarrierGradeNAT)
  • Problematic with servers (static mapping and port forwarding cannot be configured)

Question 16

A+P

Answer 16

• Address + Port
• Carrier grade solution to ipv6 migration
• NAT is under control of customer
• Ranges of TCP/UDP ports are assigned to each customer
  
  • Only ports used by NAT on outside
• Features
  
  • No problem with overlapping private address spaces at customers’
  • Ports acan be assigned automatically to CPE using the Port Control Protocol
    
    • CPE can negotiate more ports at any time
  • AFTR is just IPv4-in-IPv6 (proto-41) tunnel terminator
    
    • NAT44 is not needed

Question 17

MAP ipv6 migration

Answer 17

• Mapping address and port
• Multiple CPE use same public IPv4 address
• set (not range) of ports assigned to cpe
• Client ipv4 address and port set mapped to unique ipv6 address
  
  • prefix routed to cpe
• ipv4 public server address mapped to unique ipv6 address
  
  • prefix routed to border relay
• MAP-E (E stands for Encapsulation):
  
  • IPv4 packets are tunneled
• MAP-T (translation): ipv4 translated to ipv6

Question 18

MAP port set

Answer 18

Port set:

• Each CPE is assigned a unique PSID.
• A > 0 is a bits; PSID is k bits; ; j is m bits.
• a+ k + m = 16bits
• The cpe uses ports with psid value, varying A and j.

Question 19

MAP CPE ipv6 address

Answer 19

• EA(Embedded address) bits contain PSID and partial ipv4 address
  
  • uniquely identifis the cpe

Question 20

MAP border relay

Answer 20

• BR address must be known to CPEs Multiple BRs might have same address
  
  • Anycasting
• MAP-E: BR address terminates tunnel
• MAP-T: prefix associated to BR used for translation of outside IPv4 addreses
  
  • BR prefix is advertised on the backbone
  • Might be advertised by multiple BRs

Question 21

NAT64 + DNS64

Answer 21

Principles:

• an ipv6 prefix is dedicated to a mapped ipv4 address
• dns64 maps A records (ipv4) into AAAA using NAT64 prefix, then serves A and AAAA records to the client
• NAT64 router advertises NAT64 prefix into ipv6 network to attract traffic toward ipv4 hosts

Limitations:

• Possible only when dns is involved
  
  • does not work if user specifies directly ipv4 address
• no dnssec (authoritative dns signs records): since dns64 modifies records

Question 22

NAT64 + DNS64 deployment scenarios

Answer 22

Question 23

nat64 packet forwarding

Answer 23

• Translates ipv6 adress and packet into ipv4
• picks a free ipv4 address/port form its pool
• build nat session entry

Question 24

DNS64 + NAT64 special case of no need for dns

Answer 24

Question 25

6PE

Answer 25

* Only PE routers need to be changed * same mechanism as MPLS-based VPN * very scalable * Minimum configuration required * As more customers require ipv6 connectivity, provider might consider migrating to native support. * Internal routing * 6PE and P routers advertise their ipv4 prefixes with Labels bound to each PE * Topology based label binding. In the image: P-1, P-2: IPV4 only routers PE-1,PE-2: Dual stack IPv4-IPv6 with MultiProtocol-BGP (MP-BGP)

Question 26

MPLS-Based solutions

Answer 26

* Meant for interconnecting ipv6 islads through an ipv4 mpls backbone * most netowrk service providers deploy mpls anyway * IPv6 and MPLS: * Data plane is agnostic to ipv4/ipv6 (labels) * Control plane is not agnostic: destinations are identified through IP address. Solutions: * Native IPv6 over MPLS * IPv6 over Circtuit Transport * 6PE

Question 27

Native IPv6 over MPLS

Answer 27

* Control plane fully upgraded * IPv6 support in: * routing * label distribution protocol * management * Possibly Dual support

Question 28

IPv6 over Circuit transport

Answer 28

* MPLS as layer 2 connectivity * LSP === L2 tunnel * PEs are ipv6 aware: static routes to ipv6 destinations and routing with remote PEs * No changes needed to P routers * It does not scale: * L2 tunnels routes are configured manually * possibly mesh topology of L2 tunnels

Question 29

Propagation and redistribution of routes, announcing ipv6 network with 6PE

Answer 29

* Propagation of routes * 6pe exchange routing info through mp-bgp sessions over ipv4 * 6pe ipv4-mapped address as BGP next hop * Announcing ipv6 networks * CE and 6PE connect through native ipv6 interfaces routing information is exchanged * Redistribution of ipv6 routes * PEs propagate adv to ipv6 networks through IGP * May not be needed if CE-1 uses a default route * Label distribution * For ipv4 dest. among P and 6PE routers * LDP or RSVP * For ipv6 routers: directly between 6PE, using MP-iBGPv4

Question 30

6PE packet forwarding

Answer 30

* CE routers send IPv6 packets to 6PE routers * 6PE has label mapping for destination * was distributed by iBGP * next hop: 6PE identified the BGP next hop * ipv4-mapped ipv6 address * Next hop is reachable through ipv4 (IGP said that) *