Ironically_ the need for the Internet2 project has - School of

Report as inappropriate Your report has been received

Shared by: pengxuebo

Tags

Stats

views:: 0
posted:: 12/22/2012
language:: Unknown
pages:: 29

Public Domain

Document Sample

The New Internet

Jeffrey R. Ellis, Adam P. Uccello, Richard C. Gronback
University of Connecticut
Computer Science and Engineering
CSE 245 – Computer Networks
December 22, 2012

Abstract

The Gartner Group has predicted that a large minority of the more than 4,500 Internet Service
Providers (ISPs) in the United States “will be forced out of business in the next five years”
(Gar98, 1). Additionally, “[b]etween one-third and one-half of U.S. households will not be able
to afford … the $50 to $60 monthly cost of cable modem access by century’s end” (Gar98, 2).
This is due to the inevitable spill over of new internet technologies into the commercial market as
the government and academic community invest in future networking technologies. Currently,
there are two major initiatives in this quest for alleviating the bandwidth-constrained research
and academic communities who now share with commercial markets what was once their
exclusive network: Internet2 (I2) and Next Generation Internet (NGI). Even though these are
two separate programs, they have many commonalities not just in concept, but in physical
hardware. This paper will explore each individually and also take a look at the features they
share, thereby giving a comprehensive overview of what lies ahead in the new Internet and why
the Gartner Group may be correct in their assessment of its impact on commercial Internet
activity.
CSE245 – Computer Networks J. Ellis, A. Uccello, R. Gronback

1. Introduction
As anyone living outside a cave in the last 4 to 5 years can attest, the Internet has revolutionized
many facets of the way we live, work and do business. Many of those same people, prior to
1995, had no idea that a cross-country computer network existed and was being used daily by
researchers in both government and academic environments. The reason being that prior to the
National Science Foundation’s (NSF) NSFnet going public in 1995, no one would have known,
or likely cared. Now, however, as the personal computer and its Internet browser(s) are found in
almost as many homes as the television, we have come to expect this network not only to exist,
but also to perform up to our increasingly higher standards. Of course, as our “commercial” use
increases, the institutions that formerly had exclusive access to this network have seen their
available bandwidth squeezed to unacceptable levels. So, the cycle continues as these
institutions are planning to develop a new network – much of which is actually implemented on
top of existing infrastructure – to get back their “private” access.
The original NSFnet allowed for a relatively fast T3 (45 Mbps) connection between many
university and government research facilities spanning most of the U.S. The NSF was the key
player in developing this backbone and regional IP networks. Later, to encourage further
development of the Internet and networking technologies, the Foundation partnered with MCI to
develop the very-high-performance Backbone Network Service (vBNS), which opened for
“business” in 1995. More on the particulars associated with the vBNS and its role in the new
Internet will follow below. For now, it is important to realize that vBNS is important to both the
I2 and NGI initiatives.
Both I2 and NGI address the problem of increased congestion on the present Internet.
The difference between the two lies mainly with their approach to a solution. Internet2 is a
bottom-up initiative and a project of the University Corporation for Advanced Internet
Development (UCAID). It is comprised of some 120 universities and 25 corporate sponsors, all
of which pay “an annual fee of between $10,000 and $25,000 and must demonstrate that they are
making a definitive, substantial, and continuing commitment to the development, evolution and
use of networking facilities and applications in the conduct of research and education” (Fin98, 2).
On the other hand, Next Generation Internet is a top-down initiative, originating in the White
House, “involving federal agencies, that reaches down to academia and the user community.

The New Internet Page 2 of 29 12/22/12
CSE245 – Computer Networks J. Ellis, A. Uccello, R. Gronback

Essentially, its mandate is to remove roadblocks to continued American dominance in
technological innovation” (Fin98, 4). While I2 is an off-the-shelf approach using existing
technologies, NGI is interested in the research and development of new networking technologies.
Both programs will initially act as a test-bed, as did the original NSFnet, and probably be
released for commercial use, as was the current Internet.
Given all of the hype over the present Internet, it is not surprising to find ourselves
subjected to even more hype over the next generation Internet. Often, there is confusion about
how this new Internet is coming into reality. This paper will hopefully clear up some common
misconceptions. First of all, there is not currently any new cabling being laid across the country.
Both I2 and NGI will utilize existing backbones. Additionally, to connect these backbones in an
efficient manner, both initiatives will need to use GigaPOPs. A GigaPOP is a gigabit-capacity
Point of Presence that will allow for discriminate points of access to the newer high capacity
backbone. An example of an existing GigaPOP in North Carolina will be detailed below. Also,
as the expandability of IPv4 has about come to an end, the new Internet will incorporate the
newer IPv6. A brief overview of this protocol is also given below. Before an overview of
vBNS, GigaPOPs and IPv6 is given in section 4, a detailed look at the specifics of I2 and NGI
will be addressed in sections 2 and 3, respectively.

2. Internet2 (I2)
Ironically, the need for the Internet2 project has grown out of the commercial success of the
original Internet. Whereas the Internet once provided ample opportunity for universities and
research organizations to study, experiment, and download and share information, the rush of
private companies and corporations to use the network for commercial activity has clogged the
information expressway and forced researchers to evaluate alternative methods for
experimentation and research. And while capitalism and business usually drive advances in
technology, problems associated with the developed Internet’s architecture are too massive for
any company or consortium of companies to successfully combat.
Therefore, in late 1996, thirty-four universities met to examine the current Internet
situation and plan how to develop new networking technologies and applications. The result of
this meeting was the plan for Internet2, whose mission is to:

The New Internet Page 3 of 29 12/22/12
CSE245 – Computer Networks J. Ellis, A. Uccello, R. Gronback

Facilitate and coordinate the development, deployment, operation and
technology transfer of advanced, network-based applications and network
services to further U.S. leadership in research and higher education and
accelerate the availability of new services and applications on the Internet.
(In2M98)

The universities pledged to set up this “second” Internet as a way to further research and
development of networking issues and technologies. Running on high-performance network
backbones, separate from the mainstream Internet, and utilizing new Point-of-Presence access
junctions of exceedingly high data-rate capacity, this new internet would serve as a test-bed for
experimental technologies and application programming. And, as it would be a private network
accessible only by cooperating universities, it would suffer from none of the bandwidth overloads
of the original Internet.
Today, Internet 2 boasts over 130 member universities and institutions of learning.
Overseen by the University Corporation for Advanced Internet Development (UCAID), Internet2
has also drawn support from corporate sponsors, who have pledged $30 million to the project.
Member universities are providing over $60 million each year in equipment, personnel, and
funding. Since the announcement of Internet2 in October 1996, these institutions and this
funding has transformed the idea of a next-generation Internet into a reality, through two
enormous network backbones and a dozen GigaPOP access nodes.
UCAID has identified three major goals for the Internet2 project. First and foremost is
the development of a cutting-edge research network. As universities have often pioneered
advanced networks, and contributed to the success of the original Internet, UCAID believes it to
be a responsibility of academia to create and maintain an advanced network. Identifiable aspects
of a cutting-edge network include large bandwidth backbones, high-capacity PoPs, experimental
technologies for data transfer, adaptability, reliability, security, and groundbreaking hardware
devices. It is expected that advances in many of these areas will occur as a direct result of I2
development, that university researchers will use current technologies until newer technologies
are stabilized, and then the flexible network will be “upgraded” to take advantage of these new
offered services. In that way, I2 is designed using a state-of-flux mind-set.

The New Internet Page 4 of 29 12/22/12
CSE245 – Computer Networks J. Ellis, A. Uccello, R. Gronback

A second goal set forth for Internet2 is the development of revolutionary networking
applications. Although the current Internet does support client/server application programming,
mostly utilized via applets and CGI-scripts on the World Wide Web, full-fledged network
applications are few and far-between on today’s Internet. I2 hopes to remedy this by providing a
development environment for computer scientists and researchers to create applications that
utilize all the resources and services developed in this advanced physical network. This focus of
the project signals the shift away from client/server applications to a fully distributed
programming paradigm. This shift is hoped to lead to more efficient programming practices and
in general, a better use of available resources.
The final goal set forth by UCAID involves the community at large: the transfer of all
network advances to the commercial Internet. UCAID recognizes that although research for the
sake of increasing knowledge is important, practical application of the knowledge to mainstream
society is beneficial. All research results of Internet2, therefore, are public domain results and
companies in private industry are encouraged to adopt and use the gains of I2 for commercial
advancement. Advances in network design, as obtained by the first goal outlined for the project,
are expected to be implemented by mainstream Internet service providers and companies
responsible for the maintenance and creation of the actual physical network connections.
Likewise, it is expected that standards organizations will adopt any new standards and
technologies that prove successful on I2. At the same time, all web-based industries and
organizations are encouraged to attempt to incorporate the new application media into their
existing online applications. Only if the technologies birthed by the I2 project are adopted into
and fostered by the community-at-large will practical value result from the researches.
Although these are the main aspects of the Internet2 project, UCAID has outlined several
other objectives that it would like to meet. As I2 is a nonprofit, research-based network,
members of the project are expected to cooperate, not compete, on various portions of research.
It is hoped that I2 will facilitate the coordination of standards and practices among the members.
In addition to cooperating on standards issues, it is desired that researches will create and use
new applications that facilitate the sharing of research and experimental data, so as to cut down
on data duplication and redundant experimental overhead, and to foster the spirit of coordination

The New Internet Page 5 of 29 12/22/12
CSE245 – Computer Networks J. Ellis, A. Uccello, R. Gronback

between members. In addition, universities will partner with government and private-sector
organizations to develop software beneficial for all.
I2 is primarily an educational tool, and great advances in education are encouraged.
Virtual proximity, the concept of high-quality, real-time video communication among individuals
separated by physical distance, has become one of the mainstays of I2 development. A recent
demonstration of I2 technology occurred when a handful of doctors assisted in gallbladder
surgery from across the U.S. I2 will allow for the further enhancement of virtual proximity in
education services. Students and other researchers will be encouraged to experiment with
communication technologies. The impact of new networking paradigms and researches will also
be studied at educational institutions, with the results being analyzed to prepare for regular
Internet adoption.
Quality of Service is a hot topic in today’s Internet industry, and I2 has a special taskforce
assigned to develop network services that allow for more assurances that data will be delivered
properly to the correct client. QoS infrastructures will be developed and deployed on I2 to try to
address this common problem. Advanced application programming will be encouraged by the
development of middleware and other tools that facilitate design and coding. And of course, all
technological advances of I2 will be actively encouraged in migration to the regular Internet.
These additional objectives, nine in all, when combined with the goals already explored, set forth
a complete picture of the development of I2, and enumerating them gives a quantitative measure
for the project’s continued success. In short, I2 will develop the Internet of the future, one piece
at a time.
The hardware that supports Internet2 must, by the project’s definition, provide the most
advanced physical network available. First under consideration is the network’s backbone. The
backbone, of course, is the major pipe through which the non-local data is transferred. Existing
Internet backbones do not provide the bandwidth necessary for the goals of I2, so UCAID had to
find other backbone resources. The two alternatives were to find an existing one and upgrade it
as necessary, or to create one from scratch. The first option would be faster and cheaper, but it
would not allow for UCAID’s total control of data transmission nor would upgrades occur solely
at their discretion. In the end, both strategies were implemented, as the Internet2 project began to
use the existing vBNS backbone and pioneer the Abilene backbone.

The New Internet Page 6 of 29 12/22/12
CSE245 – Computer Networks J. Ellis, A. Uccello, R. Gronback

The vBNS, or very high performance Backbone Network Service, was launched by MCI
and the National Science Foundation in April 1995 to provide a backbone for advanced research
and application development. The vBNS was designed to communicate via standard IP over the
cell-switching technology called ATM, through the Synchronous Optical Network (SONET)
standard. These technologies allow the vBNS to operate at a capacity of 622 Mbps (OC12),
approximately 403 times today’s standard T1 link. Upgrades to the vBNS are planned to
increase the bandwidth to OC48, a fourfold increase to 2.488 Gbps, by the turn of the century.
In addition to the high bandwidth opportunities of the vBNS, there were several other
aspects of the backbone that proved enticing to Internet2 partners. The backbone had already
demonstrated low latency, high throughput, stability, wide coverage, and regular decongestion.
MCI provided both a production network and a test-bed network for experimentation.
Mechanisms were already in place for measuring network traffic and performance, and regular
monthly reports were distributed to all interested parties. A public archive detailed all previous
performance activities, network tests, and engineering experiments. Important work in
multicasting, quality of service, and the next generation IP standard (IPv6) was already
underway. The vBNS proved a natural fit for the ideals of I2, and over 50 universities have
connected to the vBNS as part of the Internet2 initiative.
However, the vBNS was only a partial solution to Internet2’s need for backbone services.
As MCI and the NSF were the leaders of the vBNS project, UCAID could not force the
organizations to upgrade their network bandwidth to OC48 (2.4 Gbps) to meet the demands of I2.
Further, the agreement between MCI and the NSF was only for five years, expiring in April
2000, and although a continuation of the project is likely, UCAID did not want to depend on a
further agreement between the two organizations for I2’s continued existence. Also, the fact that
the backbone was sponsored by the NSF meant that research institutions not directly related to
Internet2 were also using the network and taking advantages of the bandwidth benefits. In April
1998, therefore, Vice President Gore announced that UCAID would be developing a new high
performance backbone called Abilene. Less than a year later, on February 24, 1999, Abilene was
activated.
Partnering with Qwest Communications, Nortel Networks, Cisco Systems, and Indiana
University, UCAID has created an OC48 (2.4 Gbps) network in support of Internet2 research and

The New Internet Page 7 of 29 12/22/12
CSE245 – Computer Networks J. Ellis, A. Uccello, R. Gronback

experimentation. UCAID outlined three goals for the new Abilene network. It would be a
network that would support the demands of up-and-coming advanced research applications,
would enable the testing of new network capabilities, and would provide opportunity for the
conducting of network research. Abilene is not set to compete with the vBNS for subscribers,
but rather addresses the concerns that UCAID members had with the vBNS. The contract
between Cisco, Nortel, Qwest, and Indiana University outlasts the NSF-MCI parntership. The
network is being developed at the OC48 rate that will eventually result in the vBNS, and
developers of Abilene are planning on migration to OC192 links (9.6 Gbps). Also, UCAID
would have strict control over who would be able to access Abilene. And as Abilene would
utilize different topologies from the vBNS, member universities would have more of a choice in
technology development by choosing either network.
The major difference in technology between Abilene and the vBNS is that in UCAID’s
project, IP is transferred directly over SONET, eliminating the overhead of the ATM switching.
Advances in routing technology since the creation of the vBNS enabled this alteration, and the
routers provided by Cisco are state-of-the-art. Abilene will provide links to the vBNS, however,
to ensure maximum productivity by member organizations. Abilene’s launch was entirely
successful, and marked by the demonstration of the virtual proximity gallbladder surgery.
Both backbones provide high performance network traffic, but backbones alone do not
solve the bandwidth woes of the common Internet. Equally as important is the Point-of-Presence
that connects a local network to the backbone. Points-of-Presence can be thought of as the on-
ramp to the so-called information highway. And of course, to take advantage of the high
performance of the backbone, the PoP should have sufficiently high data capacity as well. Thus,
the concept of a GigaPOP is born.
The GigaPOP is a term simply stating that any connection to the backbone must occur at
a rate high enough to handle gigabit transactions. The connection rate does not have to be 1
Gbps or more, but it must be higher than a standard Internet PoP. One of Internet2’s defining
characteristics is that each member university must connect to a backbone through a GigaPOP. It
is in development and deployment of the GigaPOP that prevents universities from getting
actively involved in Internet2; Strawn and Luker predict that regular cost of GigaPOP and
Internet service may cost some universities up to $500,000. Therefore, it is in GigaPOP

The New Internet Page 8 of 29 12/22/12
CSE245 – Computer Networks J. Ellis, A. Uccello, R. Gronback

technology that is currently the major bottleneck for Internet2 participation. The backbones exist
and are ready for users, but universities have to fund GigaPOP access junctions before they can
take advantage of, and contribute to, the services of the vBNS and Abilene.
Of course, Internet2 is about more than just the underlying hardware. It is very profitable
for industry to develop groundbreaking hardware, so the research focus of I2 is more involved in
software, communication methods, protocols, and the like. UCAID and member universities
have identified nine different Internet2 working groups, where ongoing research in particular
areas is collected and concentrated by various researchers. Although universities are encouraged
to develop their own research and development projects, experiments related to one of the
working groups would have better facilities and more data if they joined the current working
group. Each of these working groups publishes regular updates for public consumption
(In2W99):
IPv6: The University of Nebraska is working toward the full production of the next IP
standard (128-bit instead of 32-bit). Working in tandem with the IPv6 Research and
Education Network, the IPv6 group is attempting to secure the deployment of a
production-quality IPv6 system on one of the backbones.
QoS: The Quality of Service group is coordinating and developing the QBone Initiative
(explained further in this paper). Quality of Service refers to a packet-delivery paradigm
that is more reliable than the current Internet packet-delivery system. Ensuring data
delivery is one of the key aspects of Internet2, as advanced applications will not be able to
lose any data upon network migration. The QoS group is involved in QBone
development, testing, and standards-setting.
Measurements: The University of California is leading the Measurement working group,
studying network performance. The group is working on developing measurement
standards for GigaPOPs. Currently, the Measurements group is working with the QoS
group to study QBone test results and develop performance measurements.
Network Storage: The University of Tennessee, partnering with IBM, is studying network
storage. The Distributed Storage Infrastructure (DSI) initiative (discussed further in this
paper) is a result of this working group.

The New Internet Page 9 of 29 12/22/12
CSE245 – Computer Networks J. Ellis, A. Uccello, R. Gronback

Multicast: Multicasting is the ability of one host to transmit data to several clients at the
same time. Currently, the Internet supports this only by clients forwarding the packets of
information onto other clients, essentially becoming hosts themselves. This group strives
to develop true ways of multicasting, as opposed to the Internet’s read-and-forward
approach.
Topology: This group is focused on communicating with the Next Generation Internet
initiative. As many of the goals of the NGI effort are synonomous with those of
Internet2, the Topology group discusses where and when various networks can be joined
to provide better services to members.
Routing: Based in the University of Washington, the Routing group examines the latest in
routing issues and technologies, including the topics of routing registries for different host
locations and explicit forwarding of data packets.
Network Management: The Network Management group focuses on the Abilene backbone.
This group is taking long-term responsibility for Abilene’s continued operation, as well as
architecting future management tools. This group is currently working with the Routing
group to develop common routing registries.
Security: Out of the Pittsburgh Supercomputing Center, the Security group is working on a
test authentication scheme with the Corporation for Research and Education Networking.
Security, like Multicasting and QoS, is often brandied as a major focus of I2 research.
These working groups represent a good cross-section of current technologies being
studied by I2 researchers. They provide delineated action items for lofty, general goals. And
they are a great starting point for future research. Such research has already led to three I2
initiatives, each an outcropping of ideas presented from the working groups. The initiatives,
QBone, the Digital Video Network, and the Distributed Storage Infrastructure, are finding
success and practical use from I2-developed technologies.
QBone is the name for a test-bed for the Quality of Service testing on Internet2. Develped
by the QoS working group, the QBone attempts to use Differentiated Services to separate high-
impact computing from general network traffic. Basically, DiffServ works as follows: high-
import data and application code can be marked as a special high priority, whereas common
network services (html pages, e-mail, non-real-time data) are not marked. When routers forward

The New Internet Page 10 of 29 12/22/12
CSE245 – Computer Networks J. Ellis, A. Uccello, R. Gronback

packets of data, they will queue up the common service data in favor of forwarding the high-
priority data. This invites constant streams of data to not become lost or slowed down because of
volumes of less important data. Such QoS implementation is necessary to Internet2’s success;
early implementations at the backbone and GigaPOP level ensured QoS by allocating exorbitant
amounts of bandwidth to deserving high-priority data, a policy that will not scale onto the
Internet, whereas this DiffServ approach is very applicable on the commercial Internet. The
QBone provides not only the DiffServ, but methods for tracking and measuring the effects of the
implementation.
The Digital Video Network (DVN) initiative is one service that expects to take advantage
of progress made from QBone. Although real-time video exists on the Internet, performance
problems such as Quality of Service occur. The DVN initiative exists to provide more advanced
video networking. In addition to application development, the DVN will become a production
system that will support member universities’ need for teleconferencing, courses, and other
shared information. The aforementioned gallbladder surgery was made possible through the
advances of the DVN service.
A third initiative, the Digital Storage Infrastructure (DSI) initiative is exploring how to
incorporate storage resources directly into the network. Understanding that data transfer will
continue to burden network traffic, the DSI researchers are utilizing intelligent replication of data
to alleviate this problem. DSI is also studying the concept of I2 channels to provide consistent
data links from server to user. The new channels would extend the power of Internet channels by
providing the passing of non-web services through the data stream. The DSI’s advances are
intended for general use of course, but the main goals are directed at producing the next century’s
education network.
The goals, topology, and ongoing projects of Internet2 have all been discussed on a
national level. But individual universities have as much importance to the success of I2 as any of
the working groups or initiatives already mentioned. The University of Connecticut is a working
member of the Internet2 community, and although it is not at the forefront of I2 development, it
will be taken as an example of what a single University can accomplish through I2 membership.
Led by Dr. Peter Luh and the UConn Computer Center (UCC), the UConn I2 program consists of

The New Internet Page 11 of 29 12/22/12
CSE245 – Computer Networks J. Ellis, A. Uccello, R. Gronback

six different projects (Luh98). Each requires the advances inherent in the I2 paradigm, and each
will provide useful research and data in its associated field.
First, the School of Education proposes to study the affects of distance learning on grades
K-12 students. Implementing the Virtual Proximity ideas, project leader Michael Young believes
that only two-way interactive distance learning actually promotes learning, whereas one-way
video-teaching promotes only boredom. Young argues that distance learning could affect the
choices of careers that young people make, and that science and technology occupations could be
impressed onto children via Virtual Proximity. His proposal includes the study of Virtual
Proximity effects on teachers and students alike. The tenets of the Internet2 program are met in
his proposal; a justifiable use of network bandwidth, an educational content, and relevance to the
community at large.
Krishna Pattipati and Peter Willett from the Electrical Engineering department plan to use
the features of I2 to work on network-based monitoring and fault diagnosis. The idea is that in
large, production systems, when a fault occurs, it is nearly impossible to diagnose the fault with a
minimum of downtime. The engineers suggest that the advances of network science may be able
to revolutionize system monitoring and remote diagnosis by collecting system state data from
widespread data collectors. They will be testing new monitoring algorithms and reporting tools
as part of a joint project with NASA.
Dr. Luh and L. Thakur from the Booth Research Center intend on studying network-based
scheduling and supply chain coordination. Understanding that competition in the marketplace of
today is more dependent on time considerations than cost or quality, Luh and Thakur have
developed methods for comparing different aspects of project development, and identifying
which aspects are responsible for timing issues. They propose to develop an “integrated
planning, scheduling, and supply chain management method” tool on using I2 services for
communications. They propose to develop a prototype for testing at various I2 member
universities. After successful there, a further version of the tool may be redeployed to corporate
sponsors.
Distributed Services Telemedicine is the topic of study for Dr. Ian Greenshields and Dr.
G. Ramsby. Greenshields and Ramsby suggest that simple video communication between
doctors at remote sites is no longer good enough, but images should be able to be analyzed

The New Internet Page 12 of 29 12/22/12
CSE245 – Computer Networks J. Ellis, A. Uccello, R. Gronback

quantitatively, digitally enhanced, and diagnosed. Further, the scientists believe a fully
distributed environment to be more advantageous than a client/server methodology. A
Java/CORBA tool is being developed and the researchers expect to test the product on I2.
Dong-Guk Shin and Peter Gogarten are cooperating on research for the Human Genome
Project. Explaining that the Genome project has produced a multitude of data, Shin and
Gogarten state that only a network with many available resources would have enough power to
analyze genome data. They propose the creation of a virtual genome center, through which
participating researchers can share the important data analyses.
UCC and Robert Vietzke plans on the creation of a multimedia routing network. Vietzke
presents the network to combine analog and digital signals into one multimedia network device.
The advances of I2 of which he plans to take advantage include the improved bandwidth
capabilities of the backbone, and the Quality of Service option which does not exist on the
standard Internet. This project, like all the UConn-sponsored I2 projects, fit the ideals of I2
perfectly.
As can be seen from the preceding discussions, Internet2 is a forward-thinking project
that is already starting to make huge strides. The goals of the project are to advance the science
of networking, and to apply these advances to the common Internet. As can be seen from the
hardware contributions of companies like Cisco and MCI, the studies of the national working
groups and initiatives, and the individual University of Connecticut I2 projects, the goals of
UCAID are truly being realized through I2. Internet2 is certainly helping develop the next
generation’s Internet.

3. Next Generation Internet (NGI)
History has proven that in order for technology to advance effectively, it must be fueled with a
large amount of capital. Thus, a precedent has been established in the government taking on, and
consequently funding, large next-generation research and development projects. Just as the
original ‘Internet’ was born as a DARPA project, the Next Generation Internet (NGI) project is
beginning its life as a large government funded R&D project. As for taxpayer justification, the
NGI program is said to be “essential to sustain U.S. technological leadership in computing and
communications and [to] enhance U.S. economic competitiveness.” (LSN98, 1)

The New Internet Page 13 of 29 12/22/12
CSE245 – Computer Networks J. Ellis, A. Uccello, R. Gronback

Developing new networking technologies is no small task. Recognition of this fact has lead to the
coalition of many government, commercial and academic institutions. Among these institutions
from the government side are the Defense Advanced Research Projects Agency (DARPA), the
National Science Foundation (NSF), the National Aeronautics and Space Administration
(NASA), the National Institute of Standards and Technology (NIST), the National Library of
Medicine (NLM), and the Department of Energy (DoE). As is visible, the NGI has some very
large institutions working towards its success. It is also backed by a very large sum of money.

The NGI’s budget falls under the Large Scale Networking (LSN) Working Group of the
Subcommittee on Computing, Information, and Communications (CIC) under R&D. For the
fiscal year (FY) 1998, the NGI budget was $100 million. For FYs 1999 and 2000, the current
proposal calls for $110 million per year. With the farthest deliverable due date currently
projected to be 2002, it is evident that the NGI project is one that will evolve for a number of
years most likely receiving an increase in funding as network superiority becomes an increasingly
important topic in the goal of U.S. technical dominance.

In a sentence,
The goal of the NGI initiative is to conduct R&D in advanced networking technologies, to demonstrate
those technologies in testbeds that are 100 to 1,000 times faster than today’s Internet, and to develop and
demonstra[te] on those testbeds revolutionary applications that meet important national needs and that
cannot be achieved with today’s Internet. (LSN98, 1)

More formally, there are three goals that have been defined and under which research is currently
being conducted:

1. To advance research, development, and experimentation in the next generation of networking
technologies to add functionality and improve performance.
2. To develop a Next Generation Internet testbed, emphasizing end-to-end performance, to
support networking research and demonstrate new networking technologies. This testbed will
connect at least 100 NGI sites – universities, Federal research institutions, and other research

The New Internet Page 14 of 29 12/22/12
CSE245 – Computer Networks J. Ellis, A. Uccello, R. Gronback

partners – at speeds 100 times faster than today’s Internet, and will connect on the order of 10
sites at speeds 1,000 times faster than the current Internet.
3. To develop and demonstrate revolutionary applications that meet important national goals
and missions and that rely on the advances made in goals 1 and 2. These applications are not
possible on today’s Internet.
(LSN98, 2)

As is very apparent, these goals are fairly broad. This allows the NGI initiative to encompass
many different aspects of the future of the Internet and of communications technologies as a
whole.

3.1 Goal I: Experimental Research for Advanced Network Technologies
The thrust of Goal 1 is to design, develop and deploy advanced networking technologies. It is
touted to be the ‘pathway’ to terabit-per-second speeds over wide area advanced networks. This
will be accomplished through partnerships with industry that will allow for the construction of an
infrastructure that can be used profitably by new advanced applications. Advancements made in
Goal 1 will be continually inserted into the Goal 2 testbed. In this way, the new technologies can
be experimented with ‘on-the-fly’ thereby decreasing the life-cycle time between successive
iterations of an emerging technology. The goal is broken down into three major sub-goals:

1. Network Growth Engineering
2. End-to-end Quality of Service (QoS)
3. Security

3.1.1 Network Growth Engineering
This sub-goal is concerned with the evolution of the new network topologies. It seeks to
implement a scalable architecture that has built into it the mechanisms necessary for it to grow.
This task is again broken down into three major areas:

1. Create and deploy tools and algorithms for planning and operations that guarantee predictable
end-to-end performance at scales and complexities of 100 times those of the current Internet.

The New Internet Page 15 of 29 12/22/12
CSE245 – Computer Networks J. Ellis, A. Uccello, R. Gronback

2. Facilitate management of large-scale internetworks operating at gigabit to terabit speeds
supporting a range of traffic classes on a shared infrastructure.
3. Create an infrastructure partnership through which lead users (government and research)
share facilities with the general public thereby accelerating the development and penetration
of novel network applications.
(LSN98, 9)

In support of this task are the following areas of concentration:

Planning and Simulation: This concentration seeks to formalize and possibly automate the
planning necessary for network growth and maintenance. It proposes to define a ‘network
planning description language’ that can be used for this purpose.

Monitoring, Control, Analysis, and Display: This concentration seeks to provide next-
generation analysis tools for viewing network traffic in real-time at super-high data rates. The
ultimate goal is to be able to provide real-time summaries of traffic and communication patterns
that can be used to configure better network topologies to support a particular network’s
behavior. These tools will also be invaluable in the ‘debugging’ and testing of the new
technologies.

Integration: This concentration is concerned with the seamless integration of these new
technologies. Compiling requirements from users, it will be responsible for the clean synergy of
these requirements into the NGI testbed.

Data Delivery: This concentration is concerned with the methods and protocols by which
information is delivered. It plans to provide tools that will allow a network engineer to adjust the
‘strategy tradeoffs’ that best meet their needs. Tradeoffs include everything from routing and
switching methods to priority traffic support, to virtual circuit support, to flat rate versus variable
costing options.

Managing Lead User Infrastructure: This concentration seeks to incorporate into the
infrastructure the ability to support certain ‘lead users’ of a given communications medium. In
this way, big industries such as telecommunications companies could use a large portion of the

The New Internet Page 16 of 29 12/22/12
CSE245 – Computer Networks J. Ellis, A. Uccello, R. Gronback

bandwidth of a given line while seamlessly allowing other traffic to use the remainder of the
available bandwidth.

3.1.2 End-to-end Quality of Service
This sub-goal seeks to provide effective end-to-end QoS support to the new networking
technologies. It plans to provide the framework of models, languages and protocols necessary to
permit the delivery of QoS throughout all layers of the network. Built into these protocols will be
the necessary ability to negotiate for different confidence levels and bandwidth/latency tradeoffs.
At a higher level, API’s to this framework will be defined so that applications can make
immediate use of the developed architecture.

3.1.3 Security
The last of the sub-goals of Goal 1 is to provide comprehensive security in the new technologies.
A major part of this goal is centered around cryptography. The plan is to incorporate an extensive
Public Key Infrastructure (PKI) that will interface with the industry-wide interface to effectively
provide the much needed services of authentication, data integrity, data confidentiality and
nonrepudiation. Along with the PKI, there are plans to develop new protocols which are much
more secure than the protocols used today.

It has been recognize that in order for NGI to truly succeed, there must be a high level of
confidence in the services listed above. To this end, the third focus of the security sub-goal is that
of security criteria and thorough testing. Going beyond simple functionality tests, tests will be
conducted to ensure a one-to-one mapping between what a given functionality was defined to
provide and what it actually does. In this way, things such as unintentional or intentional
‘backdoors’ in a given functionality can be discovered and disposed of.

3.2 Goal II: NGI Testbed
The proposed outcome of Goal 2 is twofold. The first and major component is that of High
Performance Connectivity. This component plans to connect at least 100 sites at speeds 100
times faster (end-to-end) than today’s Internet. The goal of this component is to provide a “full

The New Internet Page 17 of 29 12/22/12
CSE245 – Computer Networks J. Ellis, A. Uccello, R. Gronback

system, proof-of-concept testbed for hardware, software, protocols, security, and network
management that will be required in the future commercial Internet.” (LSN98, 3) The second
component is more concerned with advanced R&D. This component, Ultra High Performance
Connectivity, plans to develop ultrahigh speed switching and transmission technologies that are
capable of providing connectivity at 1+ Gbps end-to-end. This component hopes to connect
approximately 10 sites at these speeds and will lay the groundwork for future tera-bit-per-second
communications.

The NGI will essentially be a large, distributed laboratory where the technologies of Goal 1 and
the application requirements of Goal 3 are put to the test. It is expected to be an extremely
transient environment that will be constantly changing as the technologies and applications
evolve. Goal 2 is concerned with the integration of the different technologies and making sure
that everything is able to work together. The goal has been broken down into seven sub-goals,
which are detailed below.

Infrastructure: This sub-goal seeks to deal with the transient environment mentioned above. In
order to have an effective working environment, this sub-goal seeks to create a “leading edge but
stable” environment. This type of environment is difficult to achieve and will require things such
as basic backup services that can be used if a higher level service goes down. While occasional
inconsistencies (and possible failures) in the network will be inevitable and expected, the
infrastructure sub-goal will do its best to provide an environment where these ‘outages’ are
limited.

Common Bearer Services: Like the above, this sub-goal is concerned with the stable migration
to new technologies from current ones. In this way, initially, IPv4 will be the underpinning of
NGI. Once IPv6 demonstrates “stable” performance, it will be integrated. Similar actions will be
taken with all new technologies coming out of Goal 1.

Interconnection: This sub-goal is concerned with providing the seamless ‘nationwide fabric’
necessary to provide consistent QoS over all NGI sites.

The New Internet Page 18 of 29 12/22/12
CSE245 – Computer Networks J. Ellis, A. Uccello, R. Gronback

Site Selection: This sub-goal is primarily deals with NGI administrative issues. It requires that
NGI sites are properly adhering to NGI plans and goals and that the sites have plans to eventually
move to commercial funding. The sites must also ensure that a connection to a GigaPOP is
available and that all associated costs, etc… are in ‘concert’ with the overall NGI plan.

Network Management: This sub-goal is responsible for taking on the daunting task of network
management. Some key features of this effort include Distributed Help Desk,
Security/Authentication Methods, A Distributed GigaPOP Network Operation Center (NOC),
and Network Monitoring and Management Tools.

Information Distribution and Training: The final sub-goal is concerned with keeping
interested parties up-to-date on the current state of affairs in the NGI world. This will include
everything from web sites, to training classes, to conferences, etc…

3.3 Goal III: Revolutionary Applications
As with many types of technologies, the success of the NGI initiative will be measured not by the
maximum bps rate that can be guaranteed end-to-end, but instead by the ‘neat’ and ‘eye-popping’
applications that can be run on top of it. The NGI community realizes this and is therefore
spending a considerable amount of time and effort in selecting the test applications that will
show the world ‘what this thing can do’. There is a very extensive selection process whereby an
application proposal must prove that it ‘requires’ NGI technologies and that it is of general use to
the U.S. as a whole. The selection process will be selected through the collaboration of four
groups. These are, the NGI Funded Agency Missions, the NGI Affinity Groups, the Federal
Information Services Applications Council, and Broader Communities.

NGI Funded Agency Missions will solicit the mission specific application proposals that would
benefit each of the involved agencies directly. Since the NGI project spans so many agencies,
they all would like to see applications relevant to their needs be the chosen applications. To help
deal with this and identify projects that span multiple fields, the NGI has created the NGI

The New Internet Page 19 of 29 12/22/12
CSE245 – Computer Networks J. Ellis, A. Uccello, R. Gronback

Affinity Groups. There are Disciplinary Affinity groups (Health care, Environment, Education,
Manufacturing, Crisis management, Basic science, Federal information services) and Application
Technology Affinity groups (Collaborative technologies, Distributed computing, Digital libraries,
Remote operations, Security and Privacy). The Federal Information Services Applications
Council will seek to include application proposals from government groups outside of the NGI
initiative. Finally, the Broader Communities group will look outside of the government
altogether. These groups will work together to select applications that they feel are best.

Some examples of current application proposals include:
(From the ngi.gov web site as of 04.19.99)

Distributed Positron Emission Tomography (PET) Imaging: Enhance the ability of
biomedical scientists to conduct animal research through the development of high resolution PET
scans and ATM transmission of the resulting reconstructed 3-D images.

Real-time Telemedicine: Provide a means of remote medical consultations through the use of
real-time analysis of medical diagnostic procedures involving motion.

Medical Image Reference Libraries: Create medical reference libraries where images, movies
and sounds are digitized and accessible remotely.

Telerobotic Operation of Scanned Probe Microscopes (SPM): This project, among its
technical goals, aims to demonstrate and implement capabilities for the remote operation of
scanned probe microscopy (SPM) systems at various levels of control using standard data
representations and controller interfaces for collaborative measurement, research, and diagnostics
purposes associated with nanometer-scale dimensional artifacts.

Advanced Weather Forecasting: To add the new advanced Doppler weather radars to the suite
of observing systems used to initialize and update numerical weather models. This will provide
key additional data which is expected to make rapid storm-scale modeling possible, thereby
providing additional warning of weather related hazards and for crisis management related to
these events.

Chesapeake Bay Virtual Environment: To enable scientists at dispersed sites to study the
Chesapeake Bay and other marine environments using real time control of the simulation and
multimodal presentation.

As is visible, the NGI project is one that will push networking technologies to the n-th degree
and will be a major player in the future of communications as a whole.

The New Internet Page 20 of 29 12/22/12
CSE245 – Computer Networks J. Ellis, A. Uccello, R. Gronback

4. I2 and NGI Synergy
Although I2 and NGI are distinctly different programs that happen to be geared toward a
common goal, they share certain technologies that will be looked at more closely in this section.
First, as was described above, the vBNS is important to the success of both programs, so it will
be described in some detail. Second, as the concept of a GigaPOP is imperative to the
interconnecting of future high capacity backbones, we will explore the specifics of one such
GigaPOP currently in use in North Carolina. Lastly, this section will give an overview of IPv6,
as it will be the Internet Protocol version of choice in the future.

4.1 The vBNS
With the popularity of the Internet, the NSF wanted to ensure that the country’s research
communities were able to access a private, high-speed network to further develop
internetworking technologies that would eventually benefit not only academia, but commercial
interests as well. The very-high-performance Backbone Network Service was launched in April
1995 as the result of a 5-year cooperative agreement between MCI and the NSF. The vBNS
provides the following core services: “a high-speed best-effort Ipv4 datagram delivery
service[,]… an IPv4 multicast service, an ATM switched virtual circuit logical IP subnet service,
and ATM permanent virtual circuits across the vBNS backbone as needed. Among services
under development, are a reserved-bandwidth service and a high-speed IPv6 datagram delivery
service” (Jam98, 6).
Like the NSFnet that preceded it, the vBNS is a closed network that connects NSF-
sponsored supercomputer centers (SCC) and NSF-specified network access points (Fig. 1).
Originally, only 5 SCCs and 4 network access points were available. This high-speed
interconnectivity enabled researchers to link two or more SCCs, forming supercomputing meta-
centers. In fact, as pointed out in (Jam98), the NSF created the High Performance Connections
(HPC) program to sponsor access to the vBNS for R&E institutions in order to reach the “critical
mass” of the vBNS. Only then would the number of applications and experiments running on the
vBNS be able to exploit its potential and bring more advanced services to the commercial
Internet. The vBNS will ultimately host over 100 institutions, including links to other research
networks in the U.S. and abroad.

The New Internet Page 21 of 29 12/22/12
CSE245 – Computer Networks J. Ellis, A. Uccello, R. Gronback

Looking at the architectural layout of the vBNS, it is implemented as IP-over-ATM
running on over 25,000 km of a Synchronous Optical Network (Sonet) OC-12-622.08 Mbps
backbone. A collection of ATM switches and IP routers interconnect 12 vBNS POPs, located at
MCI terminal facilities, and four POPs located at the following SCCs: the National Center for
Atmospheric Research (NCAR), the National Center for Supercomputing Applications (NCSA),
the Pittsburgh Supercomputing Center (PSC), and the San Diego Supercomputer Center (SDSC).
Each of these POPs typically provides access via User-Network Interface (UNI) ports on a Fore
ASX-1000 ATM switch. In addition to this ATM connectivity, “[f]rame-based connections to
the Cisco 7507 router are also available[,] as are ports which support Packet-over-Sonet” (Jam98,
2). To allow supercomputers access via High-Performance Parallel Interface (HIPPI), their POPs
also have Ascend GRF 400 routers. In order to provide a measure of the network’s performance,
each POP has a Sun Microsystems Ultra-2 workstation with an OC-12 ATM NIC to run nightly
tests on each backbone link of the vBNS. Interested readers can find the plotted output of these
tests on the vBNS web site at http://www.vbns.net/stats.
The ATM layer actually rides on top of MCI’s Hyperstream network, which is also used
for commercial applications. “The vBNS was the first ‘customer’ on MCI’s Hyperstream
network, is the only customer with an OC-12 access rate, and will continue to be the first
recipient of advanced capabilities” (Jam98, 3). A set of Permanent Virtual Paths (PVPs) form a
full mesh topology, on top of the vBNS, connecting each node with each other. Each PVP
carries a number of Permanent Virtual Circuits (PVCs), which connect some 23 IP routers
running the Border Gateway Protocol (BGP), the internal BGP (iBGP), and the Open Shortest
Path First (OSPF) protocol.
In addition to the directly attached institutions, the vBNS supplies high-speed connections
to many other federally funded research networks and to R&E networks in Canada, Germany and
Singapore. “This high-speed (mostly 155.52Mb/s) Ipv4 connectivity between the vBNS and
other large Federal Networks provides a valuable broadening of the vBNS community, and is an
integral part of the vBNS’ participation in Internet2 and the Next Generation Internet” (Jam98,
5).
As the three projects have overlapping goals, the vBNS, NGI and I2 are commonly
confused; they all aim to provide the R&E community with a top-of-the-line network so that they

The New Internet Page 22 of 29 12/22/12
CSE245 – Computer Networks J. Ellis, A. Uccello, R. Gronback

may further networking technologies. Not only do the goals of these projects overlap, but so do
the participants, thereby adding more confusion. “The NSF is part of the multi-agency Next
Generation Internet initiative as well as the sponsor of the vBNS and the High Performance
Connections program. MCI is party to the vBNS cooperative agreement as well as an Internet2
corporate partner. Most universities with NSF grants to connect to the vBNS are also members
of Internet2. Some of these universities are also involved in NGI research” (Jam98, 2). Goals
and partnerships aside, it is clear that as the physical vBNS is a year and a half older than either
I2 or NGI, it will in many ways serve as a prototype for the both of them. The vBNS is already
running some of the types of applications that NGI and I2 aspire to support. As put in (Jam98):

One of the goals of the vBNS project is to accelerate the pace of the deployment of
advanced services into the commercial Internet in order to advance the capabilities
of all Internet users. The vBNS is an environment in which new Internet
technologies and services can be introduced and evaluated prior to deployment on
the large-scale, heavily-loaded commercial backbones. Examples…include native
IP multicasting, a reserved bandwidth service, and the latest version of IP, IPv6.

With respect to NGI’s goal of delivering 100 times the performance of the current
Internet with over 100 participating entities, the “vBNS represents the NSF’s efforts to meet that
goal” (Jam98, 2). And with respect to I2’s goal of providing academic institutions the ability to
develop advanced network technologies, they will utilitze “existing networks, such as vBNS, to
connect members to each other and to other research institiutions” (Jam98, 2).

4.2 The GigaPOP
Another collaborative effort between I2 and NGI is the GigaPOP. As the expression “a chain is
only as strong as its weakest link” promises, a network can analogously only be as fast as the
smallest bottleneck it is forced to transmit through. From (Col99) and Internet2, in order for a
GigaPOP to provide the desired interconnectivity, it must:
 have at least 622 Mbps capacity;
 provide high reliability and availability;
 use the Internet Protocol (IP) as a bearer service;
 also be able to support emerging protocols and applications;

The New Internet Page 23 of 29 12/22/12
CSE245 – Computer Networks J. Ellis, A. Uccello, R. Gronback

 be capable of serving simultaneously as a workaday environment and as a test bed;
 allow for traffic measurement and data gathering;
 permit migration to differentiated services and application-aware networking.

As some general features of GigaPOPs and their requirements with respect to I2 (and NGI) were
given in the discussions above, here we aim to expand upon these with a specific implementation
of a GigaPOP: the NC GigaPOP.
Four institutions, Duke University, North Carolina (NC) State, the University of North
Carolina (UNC) at Chapel Hill, and MCNC, teamed up with Cisco Systems, IBM, Nortel (later
Nortel Networks) and Time-Warner Communications came together as the North Carolina
Networking Initiative (NCNI). Their goal was to create a new regional “independent network in
the Research Triangle, because such a network would let the universities and MCNC conduct
computing and networking research in an environment unconstrained by congestion,
undifferentiated services, and other limitations of the commercial Internet” (Col99, 2). The
NCNI was formed in May 1996 and forwarded its first packets on the NC GigaPOP in February
1997, becoming one of the first implementations of a GigaPOP.
The NC GigaPOP has four primary nodes at NC State, Duke, UNC Chapel Hill and
MCNC (Fig. 2). These primary nodes will serve as connection points to the vBNS and upcoming
Abilene Network, while secondary nodes connect other NCNI partners. In deciding on the exact
architecture of the GigaPOP, the NCNI had to carefully consider two issues: topology and fiber-
optic infrastructure.
With only four primary nodes, the GigaPOP could easily have been configured in a full
mesh topology, which would have provided for robustness in the event of a link failure. The
problem with mesh topologies is in that they are not very scalable. A ring topology was
implemented due to its scalability and its resemblance to what phone and cable companies call a
metropolitan network (MAN), which allows for the use of hardware and software equipment that
has been optimized for this configuration.
The issue of fiber-optic infrastructure came down to cost. In order to lease the required
four OC-12 links to form the ring, the monthly cost would be $276,000 a month ($3.3 million per
annum). And, the local exchange carriers did not currently have the fiber nor the switching
capabilities required for the GigaPOP. Thus, NCNI made a deal with Time Warner
Communications to provide a private four-fiber ring infrastructure (two in, two out).

The New Internet Page 24 of 29 12/22/12
CSE245 – Computer Networks J. Ellis, A. Uccello, R. Gronback

Having a topology and cabling infrastructure, the next decision was in the networking
technologies. The NCNI decided on the same setup as the vBNS; IP atop ATM over Sonet.
Nortel provided Sonet add/drop multiplexers (ADMs) for each of the nodes, providing a total of
2.488 Gbps (OC-48) in both directions around the ring. In order to connect IP routers to the
Sonet ADMs, there were two possibilities: a 155 Mbps Cisco router that mapped IP packets
directly into Sonet (POS); or, a 155 Mbps Cisco router that mapped IP packets into ATM before
mapping them to Sonet. As alluded to earlier, the IP-to-ATM over Sonet was selected to
preclude the need for a second router “at each node to logically connect each of the four routers
and take advantage of the distributed architecture. Since ATM is virtual-circuit technology,
however, three virtual circuits could be spread across two 155-Mb/s adapters to create a fully
connected IP network, and only a single router would be needed at each node on the ring. The
virtual-circuit nature of ATM also mapped well to the circuit orientation of the Sonet
multiplexers” (Col99, 4). Additional details of the connections can be found in (Col99) and at
http://www.internet2.edu/.
The NCNI submitted a proposal to the NSF in order to have the NC GigaPOP connected
to the vBNS and was awarded $1.4 million to do so. In order to provide connectivity, the first
link in the Southern Crossroads (SoX) GigaPOP-to-GigaPOP network was made, connecting the
NC GigaPOP to the developing GigaPOP at the Georgia Institute of Technology. This link was a
45 Mbps (DS-3) dedicated line that now enabled the NCNI to share an OC-3 ATM connection to
the vBNS. The remaining problem with the GigaPOP was how to restrict commercial network
traffic, as the NCNI, Abilene and vBNS networks are designed to be research networks that
avoid commercial Internet congestion.
Since current IP routers can only implement one routing policy, NCNI provided separate
routers for each policy. Although not a scalable solution, it will have to do until explicit routing
technology based on source-destination pairs is available. For now, all network traffic
originating on the NCNI systems is directed to a single router. This router contains a list of
routes to vBNS destinations. If a destination doesn’t have a route by way of the NCNI and
vBNS, it is forwarded to the commodity Internet. This problem of policy enforcement has led I2
(In297, 2) to expect two general types of GigaPOPs:

The New Internet Page 25 of 29 12/22/12
CSE245 – Computer Networks J. Ellis, A. Uccello, R. Gronback

 Type I gigapops, which are relatively simple, serve only I2 members, route their
traffic through a one or two connections to another gigapops, and therefore have little
need for complex internal routing and firewalling; and
 Type II gigapops, which are relatively complex, serve both I2 members and other
networks to which I2 members need access, have a rich set of connections to other
gigapops, and therefore must provide mechanisms to route traffic correctly and
prevent unauthorized or improper use of I2 connectivity.

Although there is no exact blueprint for the construction of a GigaPOP, the NCNI
implementation seems to fulfill the requirements of I2. For every GigaPOP that is needed to
construct the new Internet, there will probably be a unique solution. The key is in the planning
and utilization of the highest speed equipment available, so as to prevent the congestion
problems which plague the commercial Internet. In collaborating on GigaPOPs, it is clear how
the I2 and NGI initiatives complement each other.

4.3 IPv6 Overview
Although subnetting and Classless Interdomain Routing (CIDR) have helped limit Internet
address space depletion and routing table size, it is clear that an improvement to IPv4 is needed
to overcome the scaling problems associated with the Internet’s rapid growth. IPv6 hopes to
provide a solution to the current problems while anticipating the need for future adaptability.
IPv6 provides a 128-bit address space, which will allow it to address 3.4 x 1038 distinct
nodes. “Based on the most pessimistic estimates of efficiency…, the IPv6 address space is
predicted to provide over 1500 addresses per square foot of the earth’s surface, which certainly
seems like it should serve us well even when toasters on Venus have IP addresses” (Pet96, 254).
Figure 3 shows the IPv6 header. In addition to the expanded address space, other features
planned for IPv6 as outlined in (Hin95) are:
 Expanded Routing and Addressing Capabilities: along with the increase from 32 to
128-bit addresses, IPv6 will provide more levels of addressing hierarchy and allow for
simpler auto-configuration of addresses. An additional “scope” field will add to the
scalability of multicast routing.

The New Internet Page 26 of 29 12/22/12
CSE245 – Computer Networks J. Ellis, A. Uccello, R. Gronback

 Anycast Addresses: this new type of address will identify sets of nodes where a packet
sent to an anycast address is delivered to one of those nodes. This will allow IPv6 source
route to allow nodes to control the path which their traffic flows.
 Header Format Simplification: some of the IPv4 header fields have been dropped or
made optional. By header simplification, even though the size of the IPv6 address is four
times that of IPv4, its header is only two times longer.
 Improved Option Support: the IPv6 header options are encoded to allow for more
efficient forwarding with less stringent limits on the length of options and greater flexibility
for the additions of new options in the future.
 QoS Capabilities: packets can be labeled as belonging to a particular traffic “flow” for
which the sender requests special handling, such as non-default QoS or “real-time” service.
 Authentication and Privacy Capabilities: IPv6 includes the definition of extensions
which provide for authentication, data integrity, and confidentiality.

4 8 12 16 20 24 28 32
Version Priority FlowLabel
PayloadLen NextHeader HopLimit
SourceAddress

Destination Address

Next header/data

*
*
*

Figure 3 (IPv6 Header)

With all of its improved capabilities, some of which challenge the features that make ATM
an attractive alternative to IP, IPv6 is clearly needed in any future Internet endeavor. For this

The New Internet Page 27 of 29 12/22/12
CSE245 – Computer Networks J. Ellis, A. Uccello, R. Gronback

reason, both I2 and NGI have embraced IPv6. The only challenge facing IPv6 is a smooth
transition from IPv4. Although several transition plans have been proposed, with tunneling as
popular technique, the problem is of the scale and nature that has been likened by many to the
Internet version of Y2K.

5. Conclusion
Although the two main initiatives, I2 and NGI, are distinct, they both aim toward increased
networking bandwidth for educational, research and government interests. It should be
interesting to see how the future of the Internet unfolds as these “new” networks are constructed
and later made available to an increasingly computer-centric society. If the success of the last
unveiling is any indication of future success, the Gartner Group may be right in predicting the
decline of many Internet service providers (although we’re not sure the shift from our ISPs to our
local cable company for service is a necessarily a good thing). As newer networking
technologies are proven and introduced into the commercial market, the promised “information
superhighway” may very well resemble a highway, and not Lombard Street in San Francisco.
This should make many of us happy, as we are ultimately financing their creation.

The New Internet Page 28 of 29 12/22/12
CSE245 – Computer Networks J. Ellis, A. Uccello, R. Gronback

Works Cited

(Col99) Collins, John C., et al, “Data Express: Gigabit Junction with the Next-Generation
Internet,” IEEE Spectrum, February 1999:
http://www.spectrum.ieee.org/spectrum/feb99/ngi.html.

(Fin98) Finley, Amy, "Untangling the Next Internet," SunWorld, April 1998:
http://www.sunworld.com/sunworldonline/swol-04-1998/swol-04-internet2.html

(Gar98) "GigaPOP - Lynchpin of Future Networks - Will Add Scalability; Wide Range
of Price/Performance Choices," Gartner Group, 19 Aug. 1998:
http://www.techmall.com/techdocs/TS970819-8.html

(Hin95) Hinden, Robert M., “IP Next Generation Overview,” IETF, 14 May 1995:
http://playground.sun.com/pub/ipng/html/INET-IPng-Paper.html.

(In297) "Preliminary Engineering Report," Internet2, 22 Jan. 1997:
http://www.internet2.edu/html/engineering.html

(In2M98) “Internet2 Mission,” Internet2, 1998: http://www.internet2.edu/html/mission.html.

(In2W99) “Internet2 Working Group Reports: February 1999”, February 1999:
http://www.internet2.edu/html/wgreport-9902.html.

(Jam98) Jamison, John, et al, "vBNS: Not Your Father's Internet," IEEE Spectrum, July 1998:
http://www.vbns.net/presentations/papers/NotYourFathers/notyourf.htm

(LSN98) Large Scale Networking , Next Generation Internet Implementation Team, NGI
Implementation Plan, February 1998
http://www.ngi.gov/implementation

(Luh98) Luh, Peter B. and Vietzke, Robert ,“UConn & Internet2 – Project Summary,”
University of Connecticut, 1998: http://abraham.ucc.uconn.edu/internet2.

(NGI99) The Official NGI Web Site
http://www.ngi.gov

(Pet96) Peterson, Larry L., Davie, Bruce S., Computer Networks: A Systems Approach. San
Francisco: Morgan Kaufmann, 1996. http://www.mkp.com/books_catalog/1-55860-368-9.asp.

(Sch98) Von Schweber, Erick, "Projects Promise IS Plenty," PC WEEK, 09 Feb. 1998:
http://www.zdnet.com/pcweek/reviews/0209/09ngi.html

The New Internet Page 29 of 29 12/22/12