| Case Study: Global Crossing
A Provider's Path for Migrating to IPv6 and Offering it to the End User
By Alan Rosenberg
Global VP, Business Development
Global Crossing, Inc.
Overview
In the early 1990s, the IETF recognized that IPv4 was not ideally suited
for the demands of the developing Internet. The explosive growth of Internet
traffic posed several challenges: There was insufficient address space
[aka poor regional distribution] and over 70% of the available addresses
were assigned to North America. The packet header structure was too complex
in that a higher number of fields required significant router processing.
Address space concerns, roaming authentication concerns and the basic
lack of protocol-inherent security capabilities all represented challenges
resulting from the increasing demand for mobile IP.
Additionally, natural market evolution suggested an impending change
in Protocol — more scalable addressing and routing solutions were
needed to meet Internet growth demands. Increased Internet structure complexity,
due to NATs and gateways, were challenging customers and service providers,
alike. The proliferation of fixed and portable IP-enabled devices like
laptop PCs, mobile phones, PDAs, consumer appliances, as well as military
applications and surveillance equipment, were adding to the complexity.
The thinking of the day was that a sooner transition to IPv6, rather
than later, would ensure a new Internet Protocol introduction that was
still manageable and capable of sustaining current Internet requirements.
What is IPv6, anyway?
IPv6 is not an enhanced IPv4, rather an evolution to a new standard. In
1994, IPv6 was approved as the replacement for IPv4 and, in 1998, the
core set of protocols were made into an IETF draft standard. Numerous
IETF working groups worked together to develop and refine the new protocol,
striving to address the main concerns regarding IPv4 including the limited
address space, complex routing, auto-configuration and slow router hand-offs.
IPv6 today address these primary concerns. Address space increased from
32-bit to 128-bit; simplified packet headers with extended header hierarchy,
stateless auto-configuration and fast hand-offs between routers. Additionally,
authentication and encryption extension headers were built in for improved
security and a new Anycast routing option was developed.

With IPv6 packet headers, Traffic Class replaced ‘TOS’ enabling
a priority level for delivering packets, flow labels ensure a sequence
of packets; the payload, which replaces “total length,” is
determined by length of packet following the header and the next header
is specified following the extension header. Hop limits (which replaced
“time to live”), decrement by one at each hop. Fields from
the IPv6 header are removed, including header length, precedence, identification,
flags, fragment offset, header, checksum, options and padding.
IPv6 can also be accredited with introducing extension headers –
representing a new protocol concept. With the additional headers for added
functionality, hop-by-hop options specify hop-by-hop routing. Destination
option headers control the destination router. The fragmentation header
enables packet fragmentation and the authentication header, supports secure
authentication and with the encrypted security payload header, encryption
of the payload is supported.
The advantages of stateless auto-configuration are numerous. Automating
address assignment allows for new network devices to "plug and play,"
local addresses are self-generated, address availability is validated
by neighboring devices and new addresses can be distributed via multicast
broadcast.
The new “Anycast” capability enables easier implementation
and management since, address prefix formats for unicast and multicast
are automatically allocated by address.
Our Strategy
Global Crossing commenced its IPv6 strategy in mid-2001, the objective
being to develop v6 capabilities in advance of market demand. The primary
components of the strategy were two-pronged: Near-term — create
a dedicated overlay test network and long-term — create dual-stack
routers at the edge running over the MPLS core. Trial routers were deployed
in late 2001/early 2002 and IPv6 peering was first established in 2001.

Today, Global Crossing peers with over 20 IPv6 partners. The first trial
customer was provisioned in 2002 and the company currently supports 40
customers on its live network. Beta trials on the production network were
completed in May 2005 and general availability of IPv6 was announced to
Internet access and VPN customers in 2005.
All capabilities currently available on IPv4 are available on IPv6 and
all IP-based services and going forward, all development will include
both protocols.

Migration to the New Protocol
Providers have three main paths to IPv6 adoption:
- Dual-Stack – devices are capable of supporting both IPv4 and IPv6
- Devices are configured with both protocol stacks with two sets of addresses
- GC Juniper IP Platform [Internet] is Dual-Stack v4/v6
- Tunneling – Enabling IPv6 in tunnels over IPv4 infrastructure
- “6-over-4” – Encapsulated IPv6 in IPv4 Packets
- “6-to-4” – Assignment of a Block of IPv6 Addresses
to an IPv4 Node
- Tunnel Brokers – 3rd Party Tunnel Providers
- Translation – Use of IPv6/IPv4 NATs
Global Crossing chose option one, determining that offering this protocol
to our customers via native IPv6 over MPLS using 6PE was more robust.
There may be obstacles to this implementation for other providers, as
many edge routers might require significant upgrades to perform this function,
but it is the most seamless implementation.
Though global implementations of IPv6 like ours are rare at this time,
we expect it to become much more prevalent in the coming years as new
applications will require the functionality of this protocol, such as
large domain users like cable companies and mobile phone providers. The
workarounds put in place for IPv4 are still “working” for
most deployments of IP, but IP address space is like closet space, the
more there is available the more customers will use.
|