a 10 (TEN) year old document that still seems up to date

[David Farber <farber () central cis upenn edu>]

software development groups.


Some Common Problems


If one looks at the computer architecture and the software production areas, one
sees a number of common problems. The most fundamental problem seems to be the
management of complexity. For example, recently a microcomputer design by a key
U. S. company had to be dramatically downgraded, not because of technological
problems of chip size, line widths, or similar problems, but because the design
had become so large that the designers were not able to control it. These chip
designers found themselves in essentially the same position that many software
groups have found themselves in, namely that things had gotten too complicated,
and  too large for the management and design tools that
were available (and these
designers are reputed to have had some of the most sophisticated tools around).


I feel that there are possible research paths that can provide payoff in the
management of complexity and thus improve the future ability of the nation to
maintain its leadership. Such paths are most likely similar
to those that the SDI
effort must also develop, but again, without a distinct activity targeted at the
civilian sector, our commercial field will most likely not
benefit from the SDI's
advances. It is suggested that the Foundation look seriously at a research
program which, for lack of a better term, I will call ``complex systems
engineering.''  One of the functions of this program would be to help understand
and develop tools for the management of complex systems
development. It will have
other, perhaps equally important roles, which I will touch on shortly.


Scaling


In the academic community, we tend to deal with small problems which are neatly
packaged into three-year Ph.D. topics or two-year NSF grant durations, or, worse
yet, three-year tenure decision times.  Many of the real problems in software or
hardware engineering show up only when one tackles large
problems. The problem of
building a real compiler shows very clearly the issues  in this field, (which is
not much better than it was 10 years ago). Building a state of the art
microcomputer shows the problems of complexity, while building a small prototype
circuit does not. A major challenge facing the academic
research community is how
to undertake research which can help understand the management of complex
technical systems development. Since by the nature of the university, long-term
and large tasks are unattractive, some mechanism must be evolved to expose
academics and students to and involve them in such tasks.


The Convergence of Computers and Communications


Attached as Appendix B is a reprint of a paper, written some eight years ago
(``The Convergence of Computing and Telecommunications
Systems,'' by David Farber
and Paul Baran, Science, 18 March 1977). It explores the convergence of computer
technology and communications technology and points out
that future systems will,
and must blend together these technologies if such systems
are to be competitive.
In the eight years since this paper was published, the argument has become even
stronger. The advent of local computer networks in the university and government
arenas has brought into day-by-day focus this synthesis. At the same time, the
evolution and transition of technology, such as the DOD networking; from a
research tool to a commercially viable and necessary technology has further
strengthened the argument. The commercial importance of this work can be shown
quite dramatically by the movements in the international arena toward standards
in the store-and-forward data communications area.


In the early 70's, the academic research community was a
leader in the definition
and creation of local area network technologies. As is proper, the leadership in
this area has now passed to the commercial area as the
economic importance of the
technologies became obvious.


In the main, the continuing contribution from the academic world has centered on
network modeling and measurement techniques. It is the belief of leaders in this
field that the advent of fiber optic technology as a viable transmission medium
offers the opportunity for a resurgence of fundamental research in local
networking. The ability to have bandwidths that approach those attained on
multiprocessor system buses offers the possibility of a new view of distributed
systems. Research activities being undertaken on a small
scale, such as MEMNET at
the University of Delaware, are attempting to optimize the efficiency of the
network/processor interface by making them more `natural.'  Research towards a
better understanding of the switching of very high speed
communication facilities
is an area that also requires fundamental understanding (for example, how does
one passively switch fiber links?).


The potential for very high speed, yet relatively high latency, transmission
facilities, such as fibers, calls for increased research to better understand
communications protocols and how they can be created or
adapted to perform better
in this environment. There has been some work to date in this area, dealing
primarily with satellite communications. However, I feel that the problems which
arise in local, ground-based systems may require substantially different
solutions due to the extremely high bandwidths involved and to the uses of these
facilities (for example, processor/peripheral and processor/processor
interconnections).


Research involving such high speed media calls for not only encouraging the
development of the appropriate talents within the academic community, but also
for making available state-of-the-art transmission facilities and test equipment
to academic researchers. Also, the very high speeds attained by these systems
makes the fabrication of interfaces inherently dependent on the use of very high
speed logic circuits, with all their attendant design difficulties.


On a broader scale, when one looks in the academic community for people with the
depth of experience and knowledge of the communications world necessary to do
systems engineering synthesis, one finds a startling shortage of people.  Such
shortages are not unique to academia; there is a major and severe lack of
qualified researchers and developers in the telecommunications industry as well.
This lack portends critical problems for an industry which needs to create
systems that blends communications and computing. Actions to rectify this
shortage should be given the highest priority and attention on the part of NSF.


The future holds both technological and application imperatives which will
further the symbiotic relationship of computing and communications. Technologies
such as fiber optics offer data bandwidths which approach those of the internal
bus on modern midscale computers. The adoption of standards such as the ISDN
portend an integrated service offering from the data
communications carriers. The
health of the computer community within the United States may depend, in large
measure, by on ability to react with foresight and imagination to the potentials
of these technologies. While the United States research community is still a
leader in research in distributed processing and in local networking, foreign
suppliers are rapidly becoming leaders in the application of this technology to
the marketplace. The future stimulation of university research in these areas
will depend in very large measure on the availability of substantial hardware
commitments so that they may have an environment in which to perform their
research. Trying to understand the impact of fiber optics on distributed
processing systems while being forced to utilize low bandwidth local networks is
a frustrating and non-productive enterprise.


In addition, there are new areas of research activity that are motivated by the
computer-communications synthesis; for example, the problem of privacy and
protection of information in a distributed computer-based office environment may
be critical to the commercial viability of this important
and  large target area.
Yet, if one looks within the academic community for research activities, both
technical and social, that deal with this area, one finds very, very few. The
reasons for this are complex and relate both to the ``military'' image of
research in trusted systems and to a lack of systems oriented research groups
within the academic community.


What Can Be Done?


In the early part of this document, I suggested some specific actions that could
be undertaken and alluded to paths that could be used to stimulate research in
the information systems engineering areas. The question remains of how to tie
this disparate suggestions together into a program that could motivate important
research to complement that being done in industrial laboratories.


I propose that it would be valuable to stimulate a real, traditional systems
engineering perspective to guide and motivate university research in the
information systems areas. The traditional role of systems engineering is to
provide a synthesis of fundamental research, market needs, and technological
feasibility  to create new products and new understanding of a field. Systems
engineering studies traditionally have had a longer term payoff than the short,
advanced, development type of research activities. These studies also provide a
training ground for the development of personnel who have
an appreciation for all
aspects of the engineering profession. In the case of the information systems
engineering field, this broad, high-level view is critical to the evolution of
the complex systems we have been discussing.


A systems engineering activity can not, however, exist in a vacuum. In the
industrial world, it is motivated by products. In the military world, it is
motivated by broad initiatives, such as SDI. The academic world, also, must be
motivated by a goal. To provide such a goal, I am proposing that the National
Science Foundation formulate a project which can be used as
a vehicle for hosting
both the systems engineering studies as well as the resulting research
activities.


Such a project must integrate communications, computing, software engineering,
human interface design, and information privacy. Further, it must result in
improvements in university information systems design and
fabrication facilities,
as well as yielding insights into the management of complexity.


A project which fits these goals would be the development of a advanced national
network. This network would be based on the most modern transmission technology
including, but not limited to, fiber optics and satellite links. It would
interconnect every engineering researcher in the United States. The network
project would provide these researchers with advanced Engineering Workstations,
high speed multimedia communications both within and without their campuses, and
access to the information and data bases that they need in order to carry out
their day-by-day activities. Further, it would provide an appropriate level of
information privacy protection for all network users.


The engineering workstation would be an excellent vehicle for the exploration of
human interfaces symbolic algebra, and expert systems technology when applied to
a technical environment, as well as providing a test bed for the most advanced
notions in microprocessor architecture. The definition, design, modeling,
construction and performance measurement of such a system would focus the
attention of the information systems engineering research community on:


1)  system level problems


2)  a complex design which will require the development of
tools for dealing with this complexity,


3)  a training ground for future researchers, and


4)  a set of understandings and an environment which would greatly enhance their
productivity as researchers.


Additionally, the fallout of such an activity into the industrial sector in the
form of new product ideas, new ways of using communications, new man-machine
interfaces, etc., would have a stimulating and valuable impact on the national
scene and on our international competitiveness. It would also have the potential
for producing major improvements in the productivity and quality of life for the
academic researcher.


It should be emphasized that this project is designed to develop and mature
fundamental research in the areas covered. It is not intended to just be another
facility; thus, it is important that it be managed as a
research project, drawing
together the best people that the industrial and academic research communities
can offer. We expect that the technology developed within this activity will be
pioneering, and not just another case of rehashed, 10-year old ideas. There are
models of similar projects in Japan and in the United Kingdom. In all these
cases, the activities have had a stimulating widespread effect on the research
capabilities of the countries, as well as having provided a motivation for
effective joint academic/industrial collaboration.


The Alternatives


If we continue at our current level of activity in the information systems
engineering areas, we will become more and more a customer country for advanced
technical products. We already see in Japan and in Europe
strong indications that
this will happen. Further, the academic community's
capability to train people in
information systems technology  will continue to decline as faculty who are
interested in systems-level issues leave for industry. Our faculty will become
more and more comprised of people who are not interested in doing, but just
theorizing. Our future computer engineers will not be
well-trained by exclusively
theoreticians.


Small additions in funding will probably have minor impact on the situation we
have talked about. In order to provide the stimulus for a major push in the
academic community, a significant amount of money must be targeted into a real,
concrete initiative which can fire the imagination and creativity of our
scientists and engineers.


It is my view that the atmosphere in Congress is receptive to such initiatives
and that, ongoing NSF-sponsored activities can provide some, but not all, of the
infrastructure and additional motivation for such an effort.


Acknowledgement


I would like to acknowledge the authors of the two attachments for their
contribution to my thinking processes, as well to Gary
Delp, Peter von Glahn, and Manny Farber for their useful
insights and help.