Emerging Technologies : University Information Systems

IPv6

Background

Although not nearly a crisis, Harvard University has begun to feel the effects of the limited address space IPv4 offers. The 128.103.0.0/16 network has approximately 66% of its address space allocated, and contiguous blocks of greater than 512 addresses have been exhausted. Aggressive use of VLSM (variable length subnet masking) has made for more efficient use of existing address space, but the multitude of new devices on the horizon – VoIP phones and Internet-enabled PDAs, for example – ensure that address usage will increase at more than a linear rate in the future. Harvard, of course, is not alone. The development of IPv6 was launched in the early 1990s as the limits of the IPv4 address space became apparent. The use of CIDR (classless inter-domain routing) blocking and tighter restrictions on IP address allocation has shifted the timetable forward, but address depletion still remains a serious issue

Harvard has established an IPv6(Internet Protocol version six) test bed. As the Network Operations Center for the Northern Crossroads we received an IPv6 address allocation from the Abilene NOC and we’ve established an IPv6 tunnel into the Abilene network and the 6Bone (The IPv6 Backbone Test Network). Each of the members of the Northern Crossroads have been allocated address space from the Abilene assignment to the NOC. Harvard is the first to establish a test bed via the Northern Crossroads and a description of the test environment follows.

IPv6 still remains a cutting-edge technology. Many vendors (including Microsoft and Cisco) have yet to release production-quality, IPv6-capable versions of their products. In order to get familiar with IPv6 deployment and operation, the NOC has established the testing environment.

Address Allocation

The Northern Crossroads has been assigned a block of addresses from Abilene, the Internet2 NOC. These addresses are not “valid” and will need to be changed at some point in the future. At present, there are only three TLA allocations:

2001::/16	Early production allocations
2002::/16	6to4 prefix
3FFE::/16	Test Address Space for the 6bone

The NoX has been assigned 2001:0468:0600::/40, the details of which follow:

2001:0468:0600::/48	Nox Core
2001:0468:0601::/48	Nox Reserved
2001:0468:0602::/48	Harvard
2001:0468:0603::/48	UNH
2001:0468:0604::/48	University of Maine
2001:0468:0605::/48	Dartmouth College
2001:0468:0606::/48	University of Vermont
2001:0468:0607::/48	Tufts University
2001:0468:0608::/48	Northeastern University
2001:0468:0609::/48	Boston University
2001:0468:0610::/48	MIT
2001:0468:0611::/48	Yale University
2001:0468:0612::/48	Uconn
2001:0468:0613::/48	Brown
2001:0468:0614::/48	University of Rhode Island
2001:0468:063f::/48	Nox Pt to Pt Links/Tunnels

As shown above, Harvard has been assigned the 2001:0468:0602::/48 network. The allocation details of the block assigned to Harvard are as follows:

2001:0468:0602:000::/56	Harvard Testbed
2001:0468:0602:100::/56	Harvard NOC
2001:0468:0602:200::/56	Tunnel Test (David LaPorte)
2001:0468:0602:300::/56	Unallocated
…
2001:0468:0602:FC00::/56	Unallocated
2001:0468:0602:FD00::/56	Unallocated
2001:0468:0602:FE00::/56	Unallocated
2001:0468:0602:FF00::/56	Point-to-Point Links

The above provides for 254 networks to be assigned to internal Harvard organizational units, with the remaining space allocated for point-to-point links. Each allocation may be partitioned as the local IT staff see fit.

Hardware

The IPv6 testbed has been built primarily from decommissioned or surplus NOC hardware. It consists of two Intel-based systems (running FreeBSD 4.2 and RedHat Linux 7.1) and one Alpha-based system (running Tru64 5.1). Network connectivity is provided through a Cisco 7507, Cisco 7505, and Cisco 2900XL.

The hardware is mounted in two cabinets of a rack in the Network Operations Center. The following is a diagram of the physical location of each piece of equipment within the cabinets:

Access

Each system is multi-homed to the IPv6 testbed and IPv4 management (128.103.242.0/24) networks. Each system utilizes a dual stack model, meaning it can interoperate with both IPv4 and IPv6 nodes. The dual stack model also provides resolver libraries capable of dealing with the IPv4 A records as well as the IPv6 equivalents, allowing for IPv6-only delegation chains once IPv6-enabled root servers are deployed. Although not required, special care has been taken to run only one protocol, IPv4 or IPv6, on each physical interface. This allows us to manage the IPv6 network out of band, easing troubleshooting and minimizing the risk of connectivity loss during disruptive maintenance.

Implementation

All systems are properly secured and are running SSH2 on each IPv4 and IPv6 interface for remote administration. Two of the systems, ipv6-linux.ipv6.harvard.edu and ipv6-freebsd.ipv6.harvard.edu, function as name servers and are authoritative for the ipv6.harvard.edu domain (the contents of which are listed in the appendix). One server, ipv6-linux.ipv6.harvard.edu, is running an IPv6-only webserver (available at http://www.ipv6.harvard.edu). Although it lacks any real content, it provides an easy way to verify IPv6 client connectivity.

Many services, BIND is SSH2 being notable examples, are not IPv6-ready out of the box. Source code patches must be applied to enable this functionality. Patches for many popular daemons are available at ftp://ftp.kame.net/pub/kame/misc. As these systems are not considered production, additional services may be installed or removed at any time, although we will always provide DNS and HTTP services.