'HOV' Lanes for 'Superhighway'
NIH Joins Next Generation Internet,
Internet2 Development Efforts
By Carla Garnett
On the Front Page...
Almost 30 years old and already the Internet is, well, sluggish
sometimes, especially while performing high-speed, high-capacity
applications. Try juggling a few large-file sites in the heart of the
day, then imagine how long it would take to crunch very large
amounts of data for a research project or conduct a virtual reality
experiment on the Internet. Add in the ever-increasing popularity of
the so-called information superhighway and you easily could find
yourself sitting in the mother of all traffic jams. Or worse, maybe
you have an interesting new concept with 'Net application, but,
anticipating the tie-ups, do not even attempt developing the
potentially valuable task. That's what was happening a little more
than 2 years ago when some of the country's top universities and
science and technology companies teamed up with the government
to begin developing two powerful alternatives to the Internet -- the
Next Generation Internet (NGI) and Internet2.
"Today's Internet suffers from its own success," said Dr. Michael
Ackerman, assistant director for high performance computing and
communications (HPCC) at the National Library of Medicine,
which has been instrumental in getting NIH involved early on with
the NGI and Internet2 initiatives. "Technology designed for a
network of thousands is laboring to serve millions."
Future Shock Absorbers?
To ensure that the Internet will be able to handle the pressure of
future usage, he continued, Vice President Gore announced in
spring 1996 the administration's support of NGI, which marshals
the resources of such government entities as the National Science
Foundation (NSF), the Department of Defense, NASA and NLM.
One of NSF's biggest contributions so far is development with MCI
of the very-high-speed Backbone Network Service (vBNS) in 1995,
which provides a high bandwidth network for research applications.
According to information at its Web site, the nationwide network
operates at a speed of 622 megabits per second using MCI's
network of advanced switching and fiber optic transmission
technologies. At that speed, the project boasts, "322 copies of a
300-page book can be sent every 7 seconds."
High performance interchange of the NGI/Internet2 (ATM
switching equipment) located in NLM's Computer Science Branch, Lister Hill Center
Ackerman notes that as an NIH component, NLM -- purveyor of
the grand Visible Human Project several years ago -- has long
endorsed applying HPCC technology to health research. "The
library has plans to sponsor a variety of NGI healthcare applications
in such areas as advanced telemedicine and distance learning," he
said. The development of NGI is necessary for such applications,
he explained, because they often require the "nearly instantaneous
transfer of massive amounts of data. Perhaps, more importantly,
the transfer must be highly reliable and the integrity of the data must
be rigorously maintained."
What NIH will contribute to both NGI and Internet2 are start-up
investment funds, technological and engineering expertise, and
development of medical research applications that can make the
best of expanded 'Net capability.
No Time for Tie-Ups
"With NIH's biomedical research mission and increasingly more
complex science applications, it's really more critical for NIH to be
able to communicate at high speeds with universities than with most
other agencies of the federal government," said Roger Fajman, head
of the systems development and support section of DCRT's
Network Systems Branch and one of NIH's liaisons to Internet2, a
smaller-scale project developed by a group of universities with aims
similar to NGI. NIH was the first non-university organization to
become a regular member of the more than 115-institution Internet2
consortium whose goal is to build a high-speed, members-only
network devoted solely to science, research and education pursuits.
Currently, the Internet is open basically to anyone with a computer
connection and a service provider. While universal access is good in
most respects, it obviously also significantly increases on-line traffic,
which is not at all good for what are called "data-intensive" or
"data-mining" applications. These applications require higher
bandwidth (connection speed) and larger capacity than is possible
with so many other travelers on the 'Net. Fajman says to think of it
this way: The typical television image is updated electronically about
30 times per second; in contrast, most video images on computers
are updated at varying speeds of about 5 times per second. "Video
images on computers are usually much smaller than TV pictures,"
he explains. "The size of the images affects the data rate." If
Internet2 can bring network computer imaging up to speed, imagine
the boon to telemedicine and teleconferencing alone.
Coming to a GigaPoP Near You
By mid-1998, Internet2 will consist of several GigaPoPs, or
high-speed, high-capacity connection points to the vBNS and other
national networks, around the nation. These will link Internet2
members to each other for communication and, ultimately, scientific
NIH's nearest vBNS connection point will probably be in Perryman,
Md., near Aberdeen, according to Fajman. Internet2 GigaPoPs are
as close to NIH as the University of Maryland, College Park. Since
the cost of connecting at high speed to Perryman is quite high,
connecting to a local GigaPoP can save money and provide other
benefits as well. Other institutions in the Washington and Baltimore
areas are in the process of or are considering connecting to the
Examples of some probable NGI/Internet2 applications include
medical imaging for pathology, radiology and mammography, and
maintenance and retrieval of multimedia data for patient records. A
medical researcher on NIH's campus, for instance, would be able to
consult with a physician at a university teaching hospital across
country in real-time, while they both simultaneously view a patient's
echocardiogram on their computer screens. (See sidebars for a
description of several of NIH's early potential NGI/Internet2 forays.
Also, researchers who want to discuss potential Internet2
applications can contact Fajman by phone, 402-4265, or email,
"Much of the healthcare community is just discovering the
advantages and efficiencies afforded through use of advanced
communications technologies such as the Internet," concluded
Ackerman. "Getting medical practices connected -- in rural areas,
for example -- is also still a formidable problem. NLM advocates
progress in these areas while encouraging, publicizing and
showcasing advanced patient care and medical research applications
that use evolving NGI capabilities."
Aside from developing useful applications to take advantage of the
increased capacity of these "HOV lanes" on the 'Net, what also
remains to be worked out is how the two projects will work
together. NGI, a government endeavor, is currently better funded
and more expansive, but Internet2, the university-led effort, is
further along in its implementation -- Internet2 GigaPoPs have
already popped up in several areas nationwide and many Internet2
members, including NIH, are already in various stages of upgrading
their servers and fiber optic equipment in preparation for
connection. Eventually, according to DCRT and NLM experts, NGI
and Internet2 probably will need to merge. Regardless of when and
how the two link, though, NIH -- as a charter member of both
projects -- is in prime position to reap the benefits for medical
First Steps on
New High-Tech Frontier
The National Library of Medicine has pioneered several key
underlying technologies that are candidates for the Next Generation
Internet and Internet2, according to Mike Gill of NLM's
Communications Engineering Branch (CEB). One of the
technologies being examined currently is asynchronous transfer
mode, or ATM, which could improve transmission speed of live
video, video databases, and large image files beyond the capabilities
the Internet now has. CEB has been test-driving some of its
NGI/Internet2-bound teleconferencing prototypes for a little more
a year now.
Electronics engineer Mike Gill (r) of NLM's Communications
Engineering Branch jokes that the branch's browser-based
medical information retrieval system, or WebMIRS, was "just a
gleam in" Rodney Long's eye about 18 months ago. WebMIRS
developer Long (l) says if all goes well, the Java applet -- which
allows the more than 20,000 x-rays and other key medical data
from the National Health and Nutrition Examination Survey to be
viewed via the Internet -- will be online by late spring.
In December 1996, CEB demonstrated its WebMIRS, a Java
application prototype that allowed digitized x-rays and associated
text datafiles in a multimedia database at NLM to be sent to the
annual meeting of the Radiological Society of North America in
Chicago. Taped and live (digitized) video were also transmitted, and
a live 2-way question and answer video session enabled a scientist
on campus to participate simultaneously in the meeting hundreds of
The second ATM trial -- a collaboration with DCRT -- transmitted
the proceedings of a 4-day gene therapy conference from the
Natcher Bldg. to M.D. Anderson Cancer Center in Houston via
Last June, the ATM trial network carried the TeleHealth Care 1997
conference sponsored by the Texas Health Science Libraries
Consortium at Baylor College of Medicine.
Most recently, NLM -- again teaming up with DCRT -- transmitted
Jan. 21-22 the video proceedings of the "Developing U.S. Public
Health Service Policy in Xenotransplantation" conference,
sponsored by FDA, NIH, CDC and HRSA, to the M.D. Anderson
Center in Houston. [For details on the conference see
"All these trials demonstrated the high capacity service available via
a technology that will be part of the infrastructure supporting Next
Generation Internet and Internet2 applications," said Gill.
The Radiology Consultation WorkStation
By Kenneth M. Kempner
The Radiology Consultation WorkStation (RCWS) is a multimedia,
medical-imaging workstation that has been developed by staff of the
image management and communication section of DCRT's
Computational Bioscience and Engineering Laboratory.
Envisioned as a node in a nationwide telemedicine network
supporting an electronic radiology environment, the RCWS uses a
high-speed asynchronous transfer mode (ATM) network as the
communication infrastructure. Each RCWS system enables the
high-resolution display of medical images and provides a mechanism
for remote consultations between medical specialists in one location
and colleagues at multiple sites.
The Radiology Consultation WorkStation includes four monitors:
two high-resolution black-and-white electronic view boxes (shown
above) for displaying radiographic images, a standard color
computer display, and a high-resolution video monitor to show
video signals from the collaborative site. A pan/tilt/zoom video
camera, mounted atop the video monitor, sends images through
the ATM network to other sites.
Video and sound are transported through a high-speed 155
Mbits/sec ATM link to connecting RCWS systems elsewhere on an
ATM network. Video capability is provided with a higher quality
"S-video" camera and color monitor. Other video input devices will
soon be added to the RCWS, including an S-VHS video cassette
recorder, a high-resolution patient exam camera, and an overhead
projector. Microphones and speakers allow voice communication.
Two high-resolution monochrome image display systems function
as electronic view boxes to show 14x17-inch electronic films.
"Telecursor" contouring can be performed in a manual or
semi-automated consultation mode, allowing each participating
clinician to outline features or regions for discussion and
comparison. These will be transmitted in real-time during the
RCWS consultation session.
Radiotherapy treatment planning is targeted as the initial application
for RCWS. DCRT engineers and programming staff are working
closely with Drs. Laurie Herscher and Rosemary Altemus and other
staff members of the NCI Radiation Oncology Branch. Dr. R. Nick
Bryan and staff members of the Clinical Center's diagnostic
radiology department are also collaborators on the project. The
RCWS is being tailored to help process CT image data before
development of a radiotherapy treatment plan.
RCWS systems will soon be installed at Walter Reed Army Medical
Center and the National Naval Medical Center enabling more
effective collaboration between medical staff at radiation oncology
clinics at these institutions and their NCI ROB colleagues.
An application being planned involves studies related to swallowing
disorders, laryngeal function, and head and neck cancer. Dr.
Barbara Sonies of the CC rehabilitation medicine department plans
to use the RCWS to evaluate patients from St. Louis Barnes-Jewish
Hospital's speech pathology department.
"DCRT has been looking at asynchronous transfer mode (ATM)
ever since it was proposed as the technology of the future, around
'93 or '94," says Jeff Hancock of DCRT's Network Systems
Branch. Specifically, NSB has been looking at the technology as a
way to increase the speed of the NIHnet backbone; offer higher
performing and more flexible connectivity options to the institutes
on and off campus; and transmit voice, video and data over the
Jeff Hancock of DCRT's Network Systems Branch
DCRT has installed ATM switches around campus to support the
division's scientific research and to allow transmission of real-time,
full-motion video between buildings. Conferences from Natcher or
Masur, Hancock explains, can be turned into ATM cells and
transmitted to the rest of the NIH ATM infrastructure (desktop or
remote conference center), to the conventional NIHnet and
Internet, or to any site that connects to DCRT via ATM.
"We have also been actively involved in connecting our ATM
infrastructure to high-speed ATM testbeds in the area," Hancock
continues. "DCRT and NLM share, in all respects -- bandwidth,
costs and technical administration -- a link to the DARPA-funded
and Bell Atlantic-run ATDnet. [ATDnet (http://www.atd.net) is a
high performance networking testbed in the area. Established by the
Defense Advanced Research Projects Agency (DARPA) to enable
collaboration among Defense and other federal agencies, ATDnet
serves primarily as an experimental platform for diverse network
research and demonstration initiatives.]
"This link is six times faster than the current 100Mbit NIHnet
backbone links," Hancock says. "ATDnet has been used to test the
effects of delay upon video transmission and as a way to connect
collaborative sites in the metropolitan (National Naval Medical
Center and Walter Reed Army Medical Center, for example) and
wide (Washington University in St. Louis, if things go as planned)
DCRT is also testing Bell Atlantic's production cell relay service that
uses ATM. This service will allow varying ATM connectivity
options to any off-campus building in the Washington area and will
allow DCRT for the first time to offer the same quality of data
transmission (plus video and voice transmission) to off- campus
users who have typically had to endure lower speed connectivity
due to cost. "Initial tests in Rockledge have been so successful that
the TLCs [technical LAN coordinators] there call us and complain
when they notice we have switched them back to the old
connection," Hancock remarks.
Typically, on the current NIHnet, DCRT connects users to each
other via routers, which are high speed interconnection devices.
These devices stop all traffic and examine it to determine the best
path on which to forward the data. When messages are sent, they
pass through two or more routers, which delay the transmission of
"Savvy institutes have been requesting direct connections between
buildings for years in order to bypass this delay but there is not
enough fiber to go around," Hancock explains. "With current
technology, all LANs are connected to a DCRT router in each
building and communicate over the NIHnet backbone. With ATM
technology and the use of virtual LANs, DCRT could maintain
ATM links that would connect NIH buildings to the NIHnet, but
also provide dedicated bandwidth for direct data connections (called
supernetting) between buildings as well as voice and video traffic.
ATM's traffic management features allow different traffic types to
be identified and run over the same fiber links. So where several
fiber links would have been necessary, only one is needed."
Up to Top