Portfolio: Technical articles

Moving R&D into the Networking Fastlane

Laurel M. Sheppard
Lash Publications International

Internet 2 promises to revolutionize the way researchers across the nation, whether from industry or academia, solve the problems facing society today.

Gigapops…tele-immersion…. Collective Entity…vBNS… Sounds like technobabble straight out of a Star Trek episode, right? Wrong! These are just a few of the terms used to describe several components and technologies of a national program currently in progress dubbed Internet 2 or I2. The program is a partnership of colleges, universities, government and industry that is developing advanced Internet technology and applications vital to the research and education missions of higher education. As with the first Internet, this technology will eventually be transferred to the commercial sector.

A Fast Start

I2 had its beginnings in the summer of 1995, when the Monterey Futures Group was formed to address the advanced networking requirements of higher education. The objective was to develop a system that could support real-time audio and video interactions--of high resolution and with no time lags-- between any campus location from the desktop. A little over a year later, after a series of meetings and financial commitment by participating universities, a steering committee was formed to draft a plan and I2 was born. In fact, the I2 program was announced one week before the federal government announced its Next Generation Internet (NGI) project.

The first official meeting was held in January 1997 and today I2 consists of over 130 leading U.S. universities and several hundred people working directly on this project. I2 is the primary project of the umbrella organization, University Corporation for Advanced Internet Development (UCAID), which was formed in the fall of 1997 to administer the program. I2 is intended to be a standards-based but pre-competitive production network.

Industry partners total over a dozen companies, including 3Com Corporation, Cisco Systems, AT&T, MCI WorldCom, IBM, and Qwest Communications (see page for a complete list). "Internet 2 represents a giant step in the Internet’s evolution, offering the promise of integrated voice video and data for next-generation network users," according to Eric Benhamou, chairman and CEO of 3Com.

UCAID’s university members have committed over $60 million per year in new funding for Internet 2, with corporate members promising to provide over $30 million over the life of the project. Cisco Systems, the first corporate partner, has provided $10 million in equipment grants alone.

By the end of 1998, nearly all participating universities will have stable Internet 2 connections. Some Internet 2 project members are already able to communicate using multimedia and other advanced services over the existing NSF very high performance Backbone Network Service (vBNS) infrastructure. vBNS connectivity is about 24,000 times faster than a standard 56 kilobyte modem.

In September, another high performance network called Abilene made its public debut. Developed jointly by UCAID and Qwest Communications, Nortel and Cisco Systems, the $500 million network will formally launch in January 1999 as an Internet 2 backbone network. Within a year, nearly 70 universities will be connected.

Shared Virtual Environments

One of the key objectives of I2 is to provide collaboration between researchers that are in different locations; in other words, a virtual laboratory. The components of a virtual laboratory include: Computer servers capable of handling very large scale simulations and data reductions; data bases that contain application specific information-- these data bases are both dynamic and distributed and often are very large containing terabytes of information or more; scientific instruments that are connected to the network; collaboration tools simulation technology; and software assets. Each virtual laboratory is based around specialized software for simulation, data analysis, discovery and reduction, and visualization. Most of this software was originally designed for stand-alone use on a single machine.

Multi-disciplinary design and manufacturing is just one example of a virtual laboratory environment. In this case, a company involved with the production of a large and complex product, such as an airplane, must be able to direct simulation processes to interact with design data bases which contain technical and manufacturing specifications. The design and simulation may require the simultaneous access to hundreds of subcomputations, which are provided by subcontractors at different locations. Such a collaboration can result in the most cost effective and safest product manufactured to customer specifications.

Another example is facilities engineering or city planning, which usually requires multiple civil engineers of various expertise working together to design the development or facility. A vast amount of data must also be obtained from various city and county offices. I2 would greatly simplify the process by providing one location for accessing all data that is needed, and by enabling virtual collaboration between engineers. Two or more civil engineers could view a plan simultaneously from different locations and make changes as needed, using a variety of simulation tools.

Earlier this year, one of several virtual collaborations was demonstrated. Ohio Supercomputer Center (OSC) researchers in Columbus showed how two surgeons at different locations hundreds of miles apart can collaborate using a computer simulation of an actual medical case. Researchers used a 3D computer simulation of a tumor created from magnetic resonance imaging and computer tomography data of an actual patient. As one individual removed tissue in the computer simulation, the other individual—more than 400 miles away in Washington DC—immediately "saw" the removal of the tissue. Such a computer-generated environment simulates a real-life situation and can include the senses of sight, touch and sound to give both parties the same experience.

Similarly, this concept could be applied to rapid prototyping for manufacturing. One engineer would interactively perform simulated milling on an object represented in volumetric form, while the other engineer at another location would interactively change their view of the process as well as explore the object, using a special force reflecting joystick. Thus, the two engineers could collaborate from different locations and perhaps come up with an even better design in far less time than conventional communication would allow.

I2 Architecture

A major technical challenge is the overall architecture for the Internet 2 infrastructure. There is a need to minimize overall costs to the participating campuses by providing for access to both the commodity Internet and advanced services through the same high-capacity local connection circuit. In addition, other campus programs and projects can be accommodated by means of a flexible regional interconnection architecture.

The overall architecture of Internet 2 is shown in Figure 1. The key new element in this architecture is the gigapop (or gigabit capacity point of presence) -- a high-capacity, state-of-the-art interconnection point where I2 participants may exchange advanced services traffic with other I2 participants. Campuses in a geographic region will join together to acquire a variety of Internet services at a regional gigapop. Each campus (such as Alpha and Baker in Figure 1) will install a high speed circuit to its chosen gigapop through which it will gain access to commodity Internet services as well as advanced Internet 2 services.

Among the gigapops, the wide area interconnect service must support differentiated Quality of Service as well as highly reliable, high-capacity transport. Since these capabilities are not yet available in the commodity backbone Internet, a special purpose inter-gigapop interconnect network will be established. This interconnect was initially provided by the NSF vBNS and is now being augmented by Abilene to give I2 a redundant and comprehensive set of connections.

The gigapop concept can greatly increase market competition among Internet service providers and help ensure cost-effective I2 services over the long run. It might become a common way for end-user networks to acquire a wide variety of communications services, from basic Internet transport through caching and content provision

Other Engineering Challenges

According to Larry Dunn, Technical Development Manager of Cisco’s Advanced Internet Initiatives Division, three of the major engineering challenges are routing policy, quality of service, and multicasting. Current routers tend to receive information from multiple sources about possible paths or routes to a destination. The router then chooses to use the "best" of these for packets headed for that destination. "The problem comes when a router upon which several feeder-sites converge should implement policy that might require the use of more than one of the viable paths to a common destination," explains Dunn.

Assume two senders converge on a router that has two paths to one destination. If the higher-performance path is used, and both senders are allowed to use it, all is fine. But if one sender is supposed to use the lower performance path instead, there is a problem. This case comes up in Internet 2 when an aggregating point, such as a gigapop, has traffic coming from several institutions. Many institutions might be allowed to use a particular Internet 2 backbone (vBNS or Abilene), but some might not, depending on the "acceptable use policy" of that particular backbone.

To overcome this problem, several options are available. Policy routing, where the router looks not only at the destination address, but also the source address—provides the router with enough information to forward the packet properly. However, this approach is often much slower than a destination-address-only-based one. Another approach is to require a separate router for each policy-class of traffic converging at the gigapop. For instance, if there are 10 institutions coming in, and two classes (high performance vs low performance), then two routers suffice. This method has acceptable performance, but requires more hardware.

A promising approach is MultiProtocol Label Switching or MPLS. Currently under development by the Internet Engineering Task Force (IETF), it will be deployed in the next 3-9 months and will be able to handle complex routing policy with good performance. MPLS allows short "labels" or identifiers to be applied to packets so that the correct path is selected by the router. A related approach is to include hardware and algorithms that allows a single box to have more than one forwarding table. (The router consults the forwarding table to figure out where to send the packet on the next hop towards its destination—each destination path will have its own forwarding table.) This promises to facilitate both complex routing policy and high performance, while not requiring multiple boxes.

Quality of Service

Though the term "Quality of Service" (QoS) may mean many things to many people, one of the key concerns is bandwidth. Does an application appear to have access to a certain required bandwidth end-to-end? And what happens to that application's data flow when some portion of the network experiences congestion? (remember the busy signals encountered at AOL)

Several approaches are possible, and tend to involve tradeoffs between the precision and strength of guarantees to individual flows, versus the amount of information that must be kept in the network to provide that guarantee. The rule of thumb is that more precision and stronger guarantees require more information to keep track of. Since more information means more memory or processing in the core network elements, the scalability of the network may be limited.

Two approaches being developed by the IETF are "int-serv" (Integrated Services", and "Diff-serv" (Differentiated Services). Int-serv tends to lean towards more precise guarantees, and a potential cost of scalability. Diff-serv leans towards less precise guarantees for individual data flows, but pretty strong guarantees to aggregate data flows, and has good scaling properties. I2 is establishing a program that will investigate the ability to apply diff-serv techniques end-to-end across the backbones used by Internet 2.

Some of the problem areas to be explored and resolved include: How to make QoS work across a campus, with a variety of different network devices present; how to sustain QoS across administrative domains; how to provide both local and inter-domain resource allocation control; how to assure that the above requirements work with equipment from multiple vendors; and how to make resource allocation & QoS techniques available to application developers in a useful way. QoS is an enabling technology that applications have not had easy access to in the past. However, it will likely only be useful to applications developers if they can use QOS features without making major changes to their programs.

Since many of the interesting applications contemplated by Internet 2 members include video or audio, a highly functional multicast infrastructure is desired. Multicast networking is making progress rapidly, in areas of efficient construction of multicast distribution trees or structures, efficient discovery of senders and receivers, high-performance implementations, etc. However, the current technology available still has limitations with regard to scalability and performance. Both the vBNS and Abilene engineers are working hard in this area, as are the Internet 2 institutions. Dunn expects much progress during 1999.

Of course, not all of the significant issues in Internet 2 are necessarily engineering. Examples according to Dunn include: lead times on new high-speed links, entry into the market of alternate providers of bandwidth, and politics of forming coalitions within or among states to optimize dollars, performance, service to constituents, etc. Another important issue is the social implications of advanced networking, which will be addressed for the first time at the Sociotechnical Summit in March of next year, with support from the John D. Evans Foundation, Cisco Systems, and Advanced Network & Services. The Foundation is the first private foundation sponsor of an Internet 2 initiative.

This summit, which will convene at the University of Texas at Austin, will focus on such issues as smart interfaces, effective interfaces, spatial media, tele-relating, information processing, and social and organizational uses of interactive technology. "Understanding the communication dimensions of high performance digital experiences will have profound implications for the design of interactive technology and for the social and organizational uses of it," explains Dr. Larry Faulkner, university president and a trustee of the Internet 2 project.

Future Outlook

Technologies from I2 are expected to become a commercial reality in about half the time—about 4 years—it took the first Internet technologies to become widely available to the private sector, according to Doug Gale, Director of OARnet (OSC's Networking Division Ohio’s GigaPOP manager) and one of the founders of I2. As with Internet 1, which created a multi-billion dollar networking industry, a similar outcome is expected. Gale believes that I2 could result in 10X as many jobs as I1 did, and obviously many of these will be for engineers.

And behind the scenes Internet 3 and Internet 4 are already in the works. "You can use the analogy of the development of flying to describe the technical progress of the Internet," explains Gale. "Internet 1 is propeller technology, Internet 2 jet technology, Internet 3 space flight, and Internet 4 intergalactic flight." How this translates into applications can only be imagined, but one thing is for sure—scenes right out of Star Trek will soon become a reality.

Acknowledgments: The author wishes to thank the Ohio Supercomputing Center, Internet 2 staff, and Cisco Systems for their assistance with this article.

A Whole New Vocabulary or More Technical Jargon/Technobabble As with most new technologies, especially anything related to computers, the Internet has practically created its own language. Here are just a few of the acronyms and definitions that have resulted.

CAVE (Cave Automatic Virtual Environment)--A room-sized advanced visualization tool that combines high-resolution, stereoscopic projection and 3-D computer graphics to create the illusion of complete "immersion" in a virtual environment for one or more users.

Collective Entity—An organization of gigapops who join together to acquire and manage connectivity among themselves.

Gigabits per second (G/bps)--Refers to a billion bits (or 10 to the 9th power --1,000,000,000 -- bits of information) when used to describe data transfer rates.

Gigapop--Regional network aggregation points being formed by Internet 2 universities to connect to a variety of high performance, and other types of networks. Gigapops provide scaleable high-speed connection points. About twelve Gigapops are now up and running.

Jitter--the variation in the amount of latency among packets being received

Latency--the amount of time it takes a packet to travel from source to destination. Together, latency and bandwidth define the speed and capacity of a network.

Multicast--to transmit a message to a select group of recipients. A simple example of multicasting is sending an e-mail message to a mailing list. Teleconferencing and video conferencing also use multicasting, but require more advanced tools and networks

NGI-Next Generation Internet--a federal government initiative focused on developing the revolutionary applications and networking capabilities needed by federal mission agencies such as the National Science Foundation, NASA, the National Institutes of Health, and the Department of Defense.

Packet--a piece of a message transmitted over a network. Large chunks of information are broken up into packets before they are sent across the Internet.

Quality of Service (QoS)--The ability of an application to receive a predetermined level of end-to-end performance from a network. This may include a particular amount of bandwidth or guarantees of maximum latency or jitter.

Tele-immersion—technology that allows individuals at different locations to share a single virtual environment. For instance, mechanical engineering students or industrial engineers working together to design a new bridge or robot arm would be able to interact with other group members while sharing the virtual object being modeled. The individuals could share and manipulate data and simulations and jointly participate in the simulation, design review or evaluation process. very high performance Backbone Network Service (vBNS)-a network that will connect around 100 research institutions--and already links five NSF supercomputer centers--at 2.4 gigabits per second by the year 2000. Begun in 1995, the vBNS is an investment of up to $50 million in a 5-year National Science Foundation project with MCI.

Virtual Reality-An artificial environment created with computer hardware and software and presented to the user in such a way that it appears and feels like a real environment. In addition to feeding sensory input to the user, the devices also monitor the user's actions. Goggles, for example, track how the eyes move and respond accordingly by sending new video input.

Source: www.internet2.edu

A Who’s Who of Corporate Partners
AT&T
Advanced Network & Services
Cabletron Systems
Cisco Systems
FORE Systems
IBM Corporation
Lucent Technologies
MCI WorldCom
Newbridge Networks
Nortel/Bay Networks
Qwest Communications
StarBurst Communications
StorageTek
Torrent Networking