|
|
 |
Larry Price
High Energy Physics Division
Argonne National Laboratory
High Energy Physics is concerned with the structure of matter
and forces at the most fundamental level. Paradoxically, the quest
to understand ever-smaller and more basic components of matter
has required particles of ever-higher energy. Apart from the theoretical
component, most HEP investigations are carried out at the major
accelerator centers. In the U.S., these centers are the Fermi
National Accelerator Laboratory, the Stanford Linear Accelerator
Center, Brookhaven National Laboratory, and Cornell University's
Wilson Synchrotron. U.S. physicists are active users of accelerators
abroad as well, principally the European Organization for Nuclear
Research (CERN) in Geneva, Switzerland, the German Electron Synchrotron
Laboratory (DESY) in Hamburg, Germany, the National Laboratory
for High Energy Physics (KEK) in Tsukuba, Japan, and the Institute
for High Energy Physics (IHEP) in Protvino, Russia. With the termination
of the SSC Laboratory, U.S. HEP researchers concerned with the
search for the Higgs boson will look toward a major effort at
the Large Hadron Collider (LHC) now under construction at CERN.
At the same time, the B Factory under construction at SLAC
(see sidebar) will be a significant
new U.S. HEP site, and a community of physicists interested in
studying matter-antimatter asymmetry and charge-parity violation
has begun work on a detector for that facility.
Experiments at the major accelerator centers are large-scale enterprises,
typically involving 100 to 500 physicists and at least as many engineers
and technicians during the construction phase. From initial conception
to final data collection, such experiments range in duration from
five to fifteen years. In the next generation of experiments, the
two large collaborations focused on the LHC will include about 1500
members each. These collaborations will involve participation by
multiple institutions, with an international mix.
The primary factor determining participation in a collaborative
HEP experiment is an interest in a certain approach to physics;
there is little regard for geographic proximity in the formation
of such a collaboration. Good communications--particularly via
computer networking and videoconferencing--are therefore critical
in enabling a collaboration to function at all, and communications
must be extremely good if the collaboration is to function smoothly.
Computer networking is especially crucial in HEP experiments because
their complexity requires the use of computers at every stage
of operation. For example, large codes are written to acquire,
store, and analyze large samples of data, and each of these processes
will typically involve collaborators at widely separated institutions.
Fast, reliable, sophisticated networks are indispensable to support
such joint efforts.
The basic services available via computer networking have been integrated
into the operation of current HEP experiments and into the planning
of the next generation of experiments. These services also come
into play in some phases of theoretical work. To serve these purposes,
such services as electronic mail, file transfer services, virtual
terminal service, remote access to files, and remote job submission
and job monitoring must operate with complete reliability and at
high speeds. The entire HEP research community requires access to
these services via TCP/IP, and a significant minority of HEP physicists
continues to require access to them via DECnet. The long lifetime
of HEP experiments means that DECnet support will be required for
this subset of users through the year 2000 and perhaps beyond.
In 1990, HEP groups began an experiment in the use of videoconferencing
for scientific collaborations. With an initial link established
between LBL and SLAC (soon extended to the SSC Laboratory), conference-room
videoconferencing was found to be a highly effective medium for
collaborative meetings. By 1993, HEP usage of videoconferencing
was extensive enough that time slots were difficult to find in
the normal workweek, including the early-evening slot most convenient
for conferences involving Japan. By then, HEP's videoconferencing
system had grown to 17 sites, two of which were in other countries.
In 1994, the HEP system was integrated into ESnet's Video Conferencing
Service, and a transition was made from multiplexed use of ESnet's
leased lines to use of circuit-switched video over the commercial
networks. As videoconferencing evolves technically, its use will
continue to be of great advantage to HEP and will expand as fast
as network bandwidth permits. It seems likely that within two
years, as much bandwidth will be devoted to video as to data.
At the time of this writing, there was a great deal of pent-up
demand for conference-room video service. The extent of this demand
would suggest that the this service will become much more widely
used as its cost continues to fall. However, the lack of universal
operability poses an additional barrier to increased usage of
this service. This problem stems from a lack of interest on the
part of the telephone companies in ensuring that the services
introduced after the breakup of the nationwide Bell system are
as interoperable as those introduced before that time. The lack
of universal interoperability is currently a problem for switched
data services (including conference-room videoconferencing), and
it seems likely that it will be an equal problem for the new services
envisaged as parts of the "information superhighway." ESnet's
VCS is currently supplying interconnection services in cases where
telephone companies are unwilling or unable to ensure interoperability.
Considering the projected visions of the "information superhighway"
and the failure of the telephone companies to address the interoperability
problem, it seems likely that ESnet will need to extend such interconnection
services.
Desktop, or workstation-based, videoconferencing is an alternative
service currently under development by a number of companies.
As its functionality improves and as standards are defined for
its implementation, HEP collaborations can be expected to make
heavy use of workstation videoconferencing. However, the usage
model incorporated into current development plans for this style
of videoconferencing may pose a significant barrier to its widespread
use. This model assumes that workstation-based videoconferences
will involve only a few participants, each using a separate workstation.
The resulting implementations of workstation video will be suitable
for conferences between a few widely-separated individuals but
will not support meetings involving up to 20 people at any one
site. For videoconferences that do involve numerous participants
at a single site, it is strongly recommended that the larger group(s)
meet in a conference room, where they can interact more fully
among themselves, even though they are communicating with others
at remote sites via workstation-based video.
Other new services are needed urgently. One such service is
a complete, readily accessible directory of institutions and individuals.
At present, the lack of such a directory service significantly
reduces the utility of the network. ESnet management and the ESCC
have made a significant start toward solving this problem, but
the emerging directory tools are not widely known within the Energy
Sciences community. The World Wide Web system, which was developed
by the European HEP community, promises to facilitate the use
of directory services and many other kinds of information services
over the ESnet. To cite another emerging requirement, HEP's use
of graphical windowing (via X-11, Motif, etc.) to access remote
computers is becoming routine and can be expected to grow rapidly.
Because of this demand alone, bandwidth requirements for interactive
use will grow by an order of magnitude over the next 2-3 years.
HEP networking needs are largely defined by the requirements of
large accelerator-based experiments. However, the smaller component
of HEP research that is not accelerator-based creates its own distinctive
subset of requirements. These experiments typically seek to detect
cosmic rays or radiation from rare, spontaneous terrestrial events.
Such experiments involve shielding the detector system within a
mountain or deep in a mine to ensure the sensitivity required to
isolate a tiny signal. This aspect of HEP work requires efficient,
reliable network communications to such remote places as Soudan,
Minnesota, Dugway, Utah, the Gran Sasso Tunnel in Italy, or the
Baksan mine in Russia.
HEP groups at laboratories and universities require reliable,
universal connectivity among themselves. At the same time, the
importance of access to major experiments and databases at Fermilab,
SLAC, BNL, CERN, DESY and China's IHEP creates a requirement for
higher-bandwidth connections to those sites. The table at the
end of this section lists the sites of major HEP experiments that
U.S. institutions currently participate in; foreign participation
is also listed for each collaboration. As noted above, HEP requires
fully capable network connections not only between each of the
participating institutions and each of the experimental sites
but also between all of the participating institutions.
Because the HEP field and its collaborations are highly international
in scope, connections to major locations in Europe and Japan need
to be as good as domestic connections. The fact that this is not
true today clearly limits the efficiency and productivity of present
HEP collaborations. The major requirements for international links
during the remainder of the 1990s are as follows:
- A 1 Mbps link to CERN in Geneva, Switzerland, is needed now.
As the LHC program becomes established, this link should grow
steadily in capacity to 10 Mbps by the end of the decade.
- A 0.5-Mbps link is needed to DESY in Hamburg, Germany. This
link should grow in capacity to 1.5 Mbps by the end of the decade.
- A 1-Mbps link is currently needed for general connectivity
to the rest of Europe. This link's capacity should grow to 10
Mbps by the end of the decade.
- A 1-Mbps link is now needed to KEK in Tsukuba, Japan (and
to the rest of Japan); this link's capacity should grow to 10
Mbps by the end of the decade.
- A 0.5-Mbps link is now needed now to IHEP, in Beijing, China,
with connectivity to other HEP institutes in China. This link's
capacity should grow to 1.5 Mbps by the end of the decade.
- A 0.5-Mbps link is now needed to key institutes in Mexico,
Brazil, and Argentina, with connectivity to other parts of Central
and South America. This link's capacity should grow to 5 Mbps
by the end of the decade.
- A 1-Mbps connection is now needed to JINR (Dubna), IHEP (Protvino),
and other points in the Moscow Region in Russia. Connectivity
is also needed to institutes in St. Petersburg and Novosibirsk
and to institutes and universities throughout Russia and other
former Soviet republics.
Estimated bandwidth requirements for new international connections
have been indicated above. A highly accurate method for estimating
HEP's domestic bandwidth requirements was employed by the HEPnet
Review Committee (HRC) in its 1988 report on HEP computer networking.[1]
That analysis led to the conclusion that ESnet's then-current 56
kbps X.25 backbone lines would become saturated by early 1989 and
to the recommendation that plans should be made for an almost immediate
upgrade of the backbone bandwidth to T1 speeds (1.5 Mbps).[2]
Section III of the HRC report further concluded that planning for
the next step in bandwidth above T1 speeds should begin sometime
in 1990.
The subsequent upgrade of the ESnet backbone to T1 lines was
completed in time to keep pace with HEP's bandwidth requirements,
validating the HRC report's estimate. However, the step up to
the next bandwidth level (T3, or 45 Mbps, speeds) has not occurred
fast enough to meet the accelerating HEP demand. As of this writing,
many lines used by HEP institutions are now becoming saturated,
and some emerging uses of the ESnet are on hold until the upgrade
to T3 lines can be completed. Full implementation of the T3 backbone
is therefore a critical priority.
As we have stressed, adequate networking support is a critical requirement
in all phases of HEP research, both experimental and theoretical.
Prompt network access, adequate bandwidth, and essential network
services are fundamental requirements for all HEP researchers. In
addition, ESnet must provide sufficient network management resources
to prevent interruptions of service. ESnet's management must also
be able to forecast requirements well enough to provide the necessary
performance and connectivity before their lack hampers the scientific
program. Finally, it is crucial for management to keep in mind that
the growth in demand for network services is fueled by the emergence
of qualitatively new capabilities as well as by quantitative increases
in usage of existing capabilities. Both trends must be tracked and
taken into account in the planning of further ESnet development.
- L. E. Price, et al., High Energy Physics Computer Networking:
Report of the HEPnet Review Committee. DOE/ER-0372 (1988),
pp. 38-42.
- ibid. p. 43.
Go to the next section, Nuclear Physics
Go to the preceding section, ESnet Today
Go to the Table of Contents
|
|
