Traditional Optical Burst Switching (OBS) protocols assume that data bursts arrive at a given node asynchronously and that burst size varies, which causes a relatively inefficient utilization of the outgoing wavelengths. Time-Synchronized Optical Burst Switching (SynOBS) was proposed to overcome this ineffective resource utilization by utilizing a synchronized timeslot-based mechanism in the OBS network. However, while synchronization improves resource utilization over traditional OBS, the system must be carefully designed in order to achieve the best performance possible. This paper investigates the effect of timeslot size on the performance of SynOBS and proposes an analytical framework for approximating the optimal solution, based on the network configuration and input traffic characteristics.

In this paper, the adaptive overlay multicast (AOM) protocol is presented, which is specially beneficial to long-lasting, large data distribution applications with large receiver sets. AOM distinguishes itself from the previous research with the following characteristics: (1) scalable overlay multicast tree construction method and a scoping technique to decrease overhead, (2) active adaptation to dynamic network conditions, and (3) efficient loss measurement and adaptation algorithms. We present extensive simulations of AOM including real Internet data and compare it with a HMTP-like protocol. The results show that AOM builds high-quality trees and adapts well to the network conditions while incurring low overhead.

We propose to model network traffic using a probabilistic context-free grammar, which is based on the multi-type branching process. Since this research is in its very early stages, the purpose of this paper is merely to suggest a justification for this model. This paper demonstrates how the lengths of the strings generated by one simple example of a probabilistic context-free grammar have first-order statistics with the characteristic "long tail" that is observed in real network traffic. The paper also shows that the lengths of the strings generated by corresponding Poisson or Markov models fall short of having this long tailed distribution.

The performance of circuit-switched and of packet-switched networks are portrayed as surfaces in a three-dimensional evaluation space. The intersection of these surfaces defines two operational domains-one optimal for circuit-switching and one for packet-switching. This result is applied to network planning in future operational domains.

In three-divisional photonic switching networks, data is interchanged among channels that are formed in space, time, and wavelength. This paper presents several techniques for designing the architectures of small photonic switching modules that switch in three divisions. In one technique, Marcus' (1970) classical transformation from a space-only Clos network to a time-space-time architecture is generalized to the wavelength division and extended to three divisions. In another technique, Hunter's (see IEEE J. Lightwave Technol., vol.11, no.3, p.495-511, 1993) space-time transformation is also generalized and extended to three divisions. It is illustrated on several classical network architectures, including the most elementary case of a three-division photonic switching module based on the Benes (1965) architecture: an eight-by-eight network in which each side has two fibers and each fiber has two wavelengths and two timeslots.

Photonic switching elements are designed from semiconductor optical amplifiers and passive couplers with fiber-to-fiber unity gain and low crosstalk. Designs for a 2*2 and an asymmetric 2*3 element, and several designs for 4*4 elements, are presented. While most amplifier analyses have stressed the importance of ultralow facet reflectivities for high-gain operation, with protection against external reflections with optical isolators, modest facet reflectivities are satisfactory for these elements. It is also shown that substantial amounts of external reflection can be tolerated. The various architectures are compared according to amplifier count, blocking characteristic, broadcast potential, noise power (amplified spontaneous emission), and fault tolerance.>

In a class of photonic systems, the intensity of the noise component caused by crosstalk from internal signals is investigated. The relevant systems are those that use loops of optical fiber as reentrant delay-line memory elements. Three architectural variations are shown to improve the ratio of the intensity of an internal signal to the intensity of this noise component. These variations include providing one extra memory element, minimizing the number of switches in a memory element that a photonic signal must traverse, and designing large systems hierarchically from smaller ones. These architectures, and the detailed formulas used to evaluate them, pertain specifically to a photonic time-slot interchanger, but the concepts are believed to generalize to similar photonic systems with applications in computing and signal processing.>

The interface between a machine, which provides a variety of services, and the user (the one served) is described. These experimental telecommunications services include enhanced calling, personal assistance, and electronic messaging. They are felt to be mutually consistent and are perceived at a user's instrument, called a teleterminal. A tree-like structure for the access of services, which is customizable to the point of being user-alterable, is presented and justified. This access software and the physical teleterminal combine the concepts of selection from a menu, traversing a tree, and the relabeling of buttons. An illustration is presented and several generalizations are discussed. The research nature of the project is emphasized.

An experimental access method has been designed to provide a "friendly" interface between a variety of telecommunications services and the user (the one served). The access software and an experimental teleterminal combine the concepts of selection from a menu, traversing a tree, and the relabeling of buttons. The casual observation over many initial users has produced a valuable, if not biased, experience. The purpose of this paper is to share this experience. A comparison is made between access of service by oral communication and by a typical interaction at a computer terminal. A number of conjectures and observations are presented in the areas of the organization and functionality of the access tree and the sense of "climbing" through it, utilization of the teleterminal's display screen, and a variety of user aids. The emotional nature of such human-machine systems and several emotional characteristics which can diminish the subject-user's objectivity are discussed.

A research model of a system for multiple-channel subrate circuit-switched digital communications is described. Three Components of a local communications system (residential or business) are the switching machine, the line, and the user's equipment. Each is briefly described and overall system operation is presented. The user's communications controller, which interfaces the digital line to the user's "devices," is described in detail. The hypothetical user is allowed a multiplicity of devices, each running at a different data rate and connected to a different "party" through the switch. Any device may itself be subclassified into a multiplicity of "virtual" devices, each with a potentially different switched connection. Research objectives include the optimal utilization of system resources, a high degree of user programmability, a friendly interface to the user (or at least a test-bed for developing one), a high degree Of system flexibility so that a variety of implementations can be tested, and a test-bed for experimental services.

Cellular spaces are shown to possess properties favorable to reconfiguration. As a first step in the direction of reconfigurable cellular spaces, this paper demonstrates the implementation of arbitrary finite-state machines in self-diagnosing cellular spaces. The results cover single cell failures caused by erroneous state transitions or by erroneous outputs. One of the attractive features of the implementation is the absence of any hard core components. As an intermediate step, it is shown how to transform any given cellular space into a self-diagnosing cellular space. The results demonstrate that the control structure (which is a finite-state machine) of any computing device can be implemented as a self-diagnosing entity without hard core.

Diagnosis of stuck-at faults (s-a-f's) in modular trees is studied. Detection conditions for each distinguishable s-a-f in a module are derived. For single s-a-f's, the detection conditions are easily partitioned to achieve fault location by performing a small number of additional tests. A multiple s-a-f that produces the same test result as a single s-a-f can be located by applying additional tests whose number grows with the tree depth. All other multiple s-a-f's are detected but cannot be located. In this paper location algorithms for combinational modular trees are presented in detail. They are then modified to locate faults in modular trees which realize arbitrary definite machines. Since a pair of these tree structures can be connected to realize arbitrary sequential machines, the results derived here are useful in diagnosing sequential machines. The ability to diagnose faults, combined with the fact that the function of the tree is easily altered, makes this structure attractive in reconfiguration applications. In particular, application to array processors is suggested.

A program which performs adaptive optimal data compression has been implemented using Jackson's Structured Design Methodology. The design process is discussed at the level of partitioning the program into tractable blocks and using a typical block, at the level of comparing data structures and program structures. Ideally the structure at these levels of generality is independent of execution and machine considerations. However, such considerations do affect the high levels of structure in this program. This "structure corruption" is detailed.

Sufficient conditions are derived that make multidimensional bilateral arrays of combinational cells easily testable for single faults. These conditions are easily implemented during the initial design of the arrays. No restrictions are made on the interconnection patterns or directions of signal flow in the arrays. The conditions are equally applicable to synchronous and asynchronous arrays. No assumptions of stability are made. Any functional fault in a single cell will be detected as long as the failed cell is still combinational. The test sequence is preset and grows linearly with the size of the array.

The MSI decoder chip has had significant impact on heuristic logic design. The intention of this correspondence is to begin to open the classical theory of combinational logic to include design with decoders. Several new kinds of implicants are introduced and a variation on the Karnaugh map is shown to be a useful tool. Examples are given (admittedly carefully chosen) to show the possible cost reduction by using design with decoders instead of with classical design.

An alternative design procedure is presented for the implementation of finite-state Moore machines as universal modular trees (UMT's). A machine property is shown to provide a design goal in the tree depth. Design specification is made heuristically by labeling transitions on a shift register state diagram. Programming of feedback and output trees is accomplished with a template tree. The procedure is felt to be more prone to reaching minimal tree depth and more expository of design choices (an aid in fault reconfiguration) than previous design procedures.

Error correction of programming languages has been effected in a heuristic fashion; error correction in the information-theoretic sense is very precise. The missing link is provided through probabilistic grammars. This paper provides the theoretical foundation for the precise construction of an error correcting compiler. The concept of code distance is extended to account for syntax in language. Grammar modifications are demonstrated so that a probabilistic parsing algorithm corrects various kinds of linguistic errors using an ideal observer rule. A generalized error correcting algorithm is described.

A class of universal finite-state machine structures employing arrays of identical modules is shown to be capable of dynamic repair by the process of reconfiguration. Coverage on the order of 98 percent of all stuck-at faults is demonstrated with no redesign of the array being necessary. After specifying a formal model for the array structure, a complete theory is developed capable of determining the minimal number of configurations necessary to obtain maximal coverage of stuck-at faults without redesigning the network. Inherent redundancy in the array structure results in this unusual capability. Many faults are also tolerated in each system configuration.

The nature of probability-measure languages (pm-languages) has been investigated [2], [3], [7], [8], in particular, those languages generated by given probabilistic grammars (p-grammars). However, the determination of a p-grammar that can generate some given language has been an open question. Since languages are infinite in general, the specification of a pm-language is vague. In this paper, it is assumed that some finite representation exists for the set of words (this can be a nonprobabilistic grammar) and that the probability of each word in the language is computable by some word function whose domain is the language.

The problem of assigning a probability to each word of a language is considered. Two methods are discussed. One method assigns a probability to a word on the basis of particular measurable features of the language. The second method is applied to languages L(G) generated by a grammar G. A probability is associated with each production of G. These in turn define the word probabilities of each word in the language. The conditions for this assignment to be a probabilistic measure are derived.