Feature interaction, in the context of telecommunications, is the phenomenon where a user (end-user or system operator) observes services failing to perform as expected, and where the failures are due to the presence of other services (or multiple instances of a single service) in the network. We propose an architectural approach to minimize the problem of feature interaction while increasing the reusability of software for services. Our approach recognizes that currently many assumptions about service operations and system capabilities are implicitly built into the service software architecture. We claim that many nontrivial feature interactions arise when services or technologies with new capabilities, based on conflicting assumptions, are introduced. We propose an architecture based on software agents and separation of concerns. A processing model for service-independent interaction management within the agent architecture is discussed. This architecture effectively removes many common assumptions about roles and technology from the service software architecture, thus reducing the need to provide ad hoc solutions to particular feature interactions. We demonstrate the power of such an architecture by showing how it deals with examples of interactions among existing services, interactions introduced by Universal Personal Telecommunications (UPT), and its applicability to emerging broadband and multimedia service requirements. We also consider the architecture in the context of the intelligent network.

The static and centralized routing of voice and data traffic in burst switched networks is addressed. It is assumed that the routing allows random bifurcation in voice and data paths and preemptive priorities for voice requirements. A study is made of routing of voice only, by using a multicommodity flow model with linearized link losses and average network loss as a minimization objective. Solving the resulting linear program, it is observed that optimal routing strategies prefer to freeze a requirement at an early stage of its path rather than those requirements that are close to their destinations. A study is made of the voice-data interaction at the link level using an available fluid-flow model, and the combined link performance is translated as a maximum flow constraint on a link. This constraint may have undesirable effects on the voice, such as introducing routes with flow absorbing loops, and unfair freezing of some requirements. All conflicting multiple objectives and constraints are included in a linear programming formulation, and it is shown how parameters can be tuned to produce desirable voice and data paths.>

It is argued despite its designation as the long-term target transport mode, parameter specifications for ATM (asynchronous transfer mode) have been driven by near-term considerations: in Europe, by the question of compatibility with the existing public network, although voice communications would clearly be a minor service in the presence of most video services: and in North America, by the apparent desire for B-ISDNs to be carriers of connectionless data between local and metropolitan area networks. If it is accepted that for better or for worse, ATM is here to stay, at least in the standards arena, then it must be ensured that the following occur: an international consensus must be reached on ATM system parameters (i.e. cell size, header functionality, etc.), on admission and congestion control techniques, etc.; long-term interests of the telecommunications industry must be reflected in today's standards' developments; and migration strategies must maintain grade-of-service requirements for existing services such as voice.>

The performance of data in burst switching has been analyzed in previous work with a fluid approximation of the data traffic. This study extends the previous model to the case where the silence interval between talkspurts has a hyperexponential, rather than an exponential, distribution. It is shown that data performance is extremely sensitive to the variance of the silence interval, and that, for empirical talkspurt and silence distributions, this model provides a vast improvement on models which assume that both types of intervals are exponentially distributed.

The performance of interactive data traffic in burst switching is analyzed with a model that allows a number of the channels to be reserved for data messages, assumes that voice talkspurts have preemptive priority over data messages in the shared channels, and uses a fluid-flow approximation for the data traffic. Extensive validations are provided for the model.

A new simulation technique is presented for performance studies of local area networks which use CSMA/CD as the access protocol. The method requires partitioning the network stations into a few primary stations and the remainder as background stations. The background stations are modeled as a group by a background algorithm based on a dynamic CSMA/CD model which uses a set of probabilities and sampling to produce, in a recurrent manner, a stochastic sequence of busy/idle periods. The new method is more efficient than conventional discrete-event simulations when the number of network stations is reasonably large. The algorithm and the overall simulation technique are comprehensively validated. The technique is illustrated with a performance study of interactive transactions on a metropolitan CATV network.