IPv6 Addressing, Exams of Engineering

Download IPv6 Addressing and more Exams Engineering in PDF only on Docsity!

IPv6 Addressing

IPv6 Transition Strategies and

Technologies Workshop

th

th

August 2014

Vientiane

Last updated 16 June 2014^1

Agenda

p Recap: how it worked with IPv

p Getting IPv6 address space

p Constructing a scalable IPv6 address plan

p IPv6 addressing on LANs

p IPv6 address plan example

How did it work for IPv4?

p Up until 1994:

n Operators applied to InterNIC for address space p 1993 onwards: included RIPE NCC and APNIC n Class A: Big organisations n Class B: Medium organisations p From 1992 onwards, multiple class Cs often handed out instead of single class Bs n Class C: Small organisations

p From 1994 onwards (classless Internet)

n Address space distributed by InterNIC (replaced by ARIN in 1998) and the other RIRs n Distribution according to demonstrated need (not organisation size or want)

IPv4 address plans (pre 1994)?

p Prior to 1994, doing an address plan in IPv4 was

very simple

p Class C was used for one LAN

n If entity had more than one LAN, they’d normally get a class B

p An organisation with a class B had 256 possible

LANs

n And that was more than most networks had in those days

p Organisations with more than 256 LANs tended to

be Universities, big IT companies, etc

n They either had multiple class Bs, or even a class A

IPv4 address plans (post 1994)?

p In the classful Internet days, IP address planning

didn’t really exist

n The address space was big enough for most needs, as the number of devices and LANs were small

p With the arrival of classless Internet, and IPv

runout in the early 90s

n IP address planning was needed n Organisations got address space according to demonstrated need p A previous class B might now only get a / p LANs no longer were automatically /24s p etc

IPv4 address plans (post 1994)?

p Advent of NAT assisted with delaying IPv4 runout

n End-user got single public address, and NATed on to that address p (End-users could get lazy again)

p Operators became more careful:

n RIR policy required “demonstrated need” p Further allocations made only when existing allocations were proven to be mostly used up n Started assigning address space across backbone according to the needs of the infrastucture p No gaps, but still no real plan p /30s for point-to-point links etc p Although the “plans” often separated infrastructure address space from what went to customers

IPv4 address plans (today)

p More serious issues – because of the lack of

structure, lack of planning:

n Infrastructure security filters become very hard to manage p Adding yet another small block of IPv4 addresses to perimeter and control plane filters n Traffic engineering is more challenging p Lots of small blocks of address space to manage and manipulate p With impacts on size of the global routing table too! n Infrastructure addressing is difficult to manage p Loopbacks and backbone point-to-point links no longer out of one contiguous block n Access address pool resizing p Broadband access pools renumbering, reassigning, etc 10

IPv

p IPv6 changes all this

p Address space delegations are generous

n Reminders of the “old days” of classful IPv

p No NAT

p Address planning is very possible

p Address planning is very necessary

p Documentation is very necessary

p Operators accustomed to handling IPv4 in the

1980s and early 1990s might be able to use

those old skills for IPv6!

Where to get IPv6 addresses

p Your upstream ISP

p Africa

n AfriNIC – http://www.afrinic.net

p Asia and the Pacific

n APNIC – http://www.apnic.net

p North America

n ARIN – http://www.arin.net

p Latin America and the Caribbean

n LACNIC – http://www.lacnic.net

p Europe and Middle East

n RIPE NCC – http://www.ripe.net/info/ncc

Internet Registry Regions

Getting IPv6 address space (2)

p From your upstream ISP

n Receive a /48 from upstream ISP’s IPv

address block

n Receive more than one /48 if you have more

than 65k subnets

p If you need to multihome:

n Apply for a /48 assignment from your RIR

n Multihoming with provider’s /48 will be

operationally challenging

p Provider policies, filters, etc

Using 6to4 for IPv6 address space

p Some entities still use 6to

n Not recommended due to operational problems n Read http://datatracker.ietf.org/doc/draft-ietf- v6ops-6to4-to-historic for some of the reasoning why

p FYI: 6to4 operation:

n Take a single public IPv4 /32 address n 2002:<ipv4 /32 address>::/48 becomes your IPv address block, giving 65k subnets n Requires a 6to4 gateway n 6to4 is a means of connecting IPv6 islands across the IPv4 Internet

Nibble Boundaries – example

p Consider the address block 2001:db8:0:10::/

n The range of addresses in this block are: n Note that this subnet only runs from 0010 to 0017. n The adjacent block is 2001:db8:0:18::/ n The address blocks don’t use the entire nibble range 2001:0db8:0000: 0010 :0000:0000:0000: to 2001:0db8:0000: 0017 :ffff:ffff:ffff:ffff 2001:0db8:0000: 0018 :0000:0000:0000: to 2001:0db8:0000:001f:ffff:ffff:ffff:ffff

Nibble Boundaries – example

p Now consider the address block

2001:db8:0:10::/

n The range of addresses in this block are: n Note that this subnet uses the entire nibble range, 0 to f n Which makes the numbering plan for IPv6 simpler p This range can have a particular meaning within the ISP block (for example, infrastructure addressing for a particular PoP) 2001:0db8:0000: 0010 :0000:0000:0000: to 2001:0db8:0000:001f:ffff:ffff:ffff:ffff