Optical Bus Communication Modeling and
Brian J. d’Auriol
Department of Computer Science
The University of Texas at El Paso
El Paso, TX, USA 79968
El Paso, TX, USA
Abstract—Multiple independent communications are studied
on optical bus parallel computing models. These models make
use of an optical bus interconnect to provide for communication
in a multi-processor system. The communication traffic studied
in this paper results in potential frequent bus collisions. The
well known collision detect mechanism is ineffective however due
to the communication pipelining aspect of the optical bus. An
analysis of the conditions necessary for bus collision and the
subsequent communications model that represents the necessary
information to determine potential bus collisions are presented.
A simulation based on this model as well as results from the
simulation are discussed. The results indicate collision patterns
and trends in sequences of general communications. The model
and simulation considered in this paper establishes a mechanism
that enables future analysis of general communication usage of
Parallel computational models provide a tool to design and
analyze algorithms for multiprocessor computers. Many paral-
lel models have been proposed in the literature, for example,
PRAM. The optical parallel bus model is an emerging type of
parallel computational model that takes advantage of optical
communications technology in its interconnection structures.
Optical bus models provide for improvements in terms of
speed and bandwidth over other types of interconnection
networks used in multiple-processor systems. The growing
popularity of these optical bus models is reflected in the more
than seventy-five publications in the area in the past decade.
An extensive bibliography appears in .
Optical bus parallel models consist of an array of processors
that communicate through an optical bus interconnection net-
work. Most such models have a one dimensional linear topol-
ogy of processors, although multi-dimensional configurations
also are possible. The interconnection network uses one or
more waveguides, i.e. optical fibers. Specific waveguides are
allocated for messages or for addressing.
Optical bus models represent an improvement over con-
ventional electrical bus parallel models. Advantages include:
pipelined concurrent access as opposed to exclusive write
access, low error probability and gigabit transmission capac-
ity . Two important properties of light that lead to these
advantages are the unidirectional propagation and predictable
propagation delay per unit length. We note that some of these
advantagesas described in the literature may now be somewhat
offset due to technological developments in newer electrical
buses at very high frequencies.
Many algorithms have been devised for various optical bus
models. Such algorithms share a common characteristic, that
bus communication occurs in a defined communication phase
within particular bus cycle time frames. One major reason
for this is that bus collisions due to multiple independent
communications are eliminated. However, many of the optical
bus models allow for more general communications, for exam-
ple, as would be encountered in MIMD operations. Collision
detect, a well known solution to bus contention, does not
work with optical buses due to the inherent pipelining of the
communication. This is because a communication could be
already on the bus, but ‘behind’ the processor initiating a new
This paper considers a communication model that specifies
the necessary conditions for bus collisions. In particular, a
simulation of the optical bus communications is developed
and some preliminary results from the simulation regarding
bus collisions are presented in this paper. We believe that this
represents the first simulated study published regarding bus
collisions arising from multiple independent bus accesses.
This paper is organized as follows. An optical bus parallel
model is discussed in Section II. This model abstracts many
of the common features found in the models proposed in
the literature. The simulation is discussed in Section III.
Conclusions are given in Section IV.
II. AN OPTICAL BUS PARALLEL MODEL
The proposed optical bus model used for the simulation
is a generic model that captures many of the most common
characteristics of the optical buses described in the literature.
Table I lists the buses that have been proposed in the literature.
The folded bus using the coincident pulse addressing technique
is a common architecture found in many of the existing
A folded bus architecture consists of a linear array of pro-
cessors interconnectedvia one or more waveguides. Processors
are connected to a waveguide at two points: to the transmitting
segment via light injectors and to the receiving segment via
light detectors. The processor at one end of the array is
The optical bus parallel computing models are based on
the use of a waveguide to direct optical pulses between
processors. Since a common communication medium connects
all of the processors, the interconnect is a bus. Although
there are variations in the proposed architectures for these
models, the majority make use of the coincident pulse tech-
nique for processor addressing. The inherent pipelining of
communication due to the unidirectional property of light
renders the usual bus collision detect method ineffective for
multiple independent communications. It appears that much of
the algorithms developed under these optical bus models make
use of relatively simple communication patterns in a lock-step
sequence with computational steps. In order to enable more
general use of the optical medium, bus contention issues need
to be solved. The approach taken in this paper is to consider
In this paper, an analysis of the conditions necessary for
bus collision and the subsequent communications model that
represents the necessary information to determine potential bus
collisions are presented. A simulation based on this model has
been implemented. Results of the simulation indicate collision
patterns and trends in sequences of general communications.
We believe that this publication represents the first optical bus
communication simulation results published.
The model and simulation considered in this paper estab-
lishes a mechanism that enables future analysis. Existing algo-
rithms could be analyzed for sequences of unicasts, broadcasts
and multicasts and compared. However, a more useful study
would be to consider new algorithms which use more general
A limitation with our approach is the reliance on the global
reference clock. To do so nicely allows the checking for bus
collisions. However, in a truly MIMD environment, such a
global clock poses problems of synchronization together with
the issues of slower clock and control circuit speeds compared
with the optical bus network. Either implementation feasibility
of the global clock architecture or enhancements to our
approach to eliminate the global clock must be considered
in future work. We are considering two alternatives, first,
decouple the waveguide time from the processor time in the
conversion expressions, second, to consider the feasibility of
newer optical technology, for example, the use of an optical
 M. Beltran, “A safe communication model for optical buses,” Master’s
thesis, Dept. Computer Science, The University of Texas at El Paso,
 Y. Pan and M. Hamdi, “Quicksort on a linear array with a reconfigurable
pipelined bus system,” in Proc. of the IEEE International Symposium
on Parallel Architectures, Algorithms, and Networks, G.-J. Li, D. Hsu,
S. Horiguchi, and B. Maggs, Eds., Beijing, China, June 1996, pp. 313–
 Z. Guo, R. G. Melhem, R. W. Hall, D. M. Chiarulli, and S. P.
Levitan, “Array processors with pipelined optical busses,” in Proc.
3rd Symposium on Frontiers of Massively Parallel Computation (Cat.
No.90CH2908-2), J. Jaja, Ed., College Park, MD, USA, October 1990,
 C. Qiao and R. G. Melhem, “Time-division optical communications in
multiprocessor arrays,” IEEE Transactions on Computers, vol. 42, no. 5,
pp. 577–590, May 1993.
 Y. Pan, “Order statistics on optically interconnected multiprocessor
systems,” Optics and Laser Technology, vol. 26, no. 4, pp. 281–287,
 C. Qiao, “Efficient matrix operations in a reconfigurable array with span-
ning optical buses,” in Proceedings. Frontiers ’95. The Fifth Symposium
on the Frontiers of Massively Parallel Computation, February 1995, pp.
 S. Pavel and S. G. Akl, “On the power of arrays with reconfigurable op-
tical buses,” Technical Report No. 95-374, Queens University, Kingston,
Ontario, CANADA, February 1995.
 Y. Pan and K. Li, “Linear array with a reconfigurable pipelined bus
system — concepts and applications,” in Proc. of the International
Conference on Parallel and Distributed Processing Techniques and Ap-
plications (PDPTA’96), Vol. III, H. Arabnia, Ed., Sunnyvale, California,
USA, August 1996, pp. 1431–1441.
 Y. Li, Y. Pan, and S. Zheng, “A pipelined TDM optical bus with
conditional delays,” in Proceedings of the Fourth International Confer-
ence on Massively Parallel Processing Using Optical Interconnections,
J. Goodman, S. Hinton, T. Pinkston, and E. Schenfeld, Eds., Montreal,
Canada, June 1997, pp. 196–201.
 H. ElGindy, “An improved sorting algorithm for linear arrays with
optical buses (extended abstract),” (Manuscript), April 1998.
 J. L. Trahan, A. G. Bourgeois, and R. Vaidyanathan, “Tighter and
broader complexity results for reconfigurable models,” Parallel Process-
ing Letters, vol. 8, no. 3, pp. 271–282, 1998.