A potential problem with distributed queue dual bus (DQDB) is the achievement of fair sharing of bandwidth in terms of the location of the access node for queued arbitrated access. To compensate for this unfairness, the IEEE 802.6 standard recommends utilizing a bandwidth balancing (BWB) mechanism with the proper value of the system parameter named BWB-MOD. The authors study and analyze the DQDB performance and the fairness with both BWB-on and BWB-off for some influential parameters based on the simulation results. For light load condition, the performance approximation is established with remarkable accuracy. It is also shown that the BWB mechanism is effective on fairness only in overload condition. An effective BWB-MOD assignment method is examined. This method indicates the appropriate value of BWB-MOD to optimize the tradeoff between the average performance of the DQDB protocol and the fairness achieved by the BWB mechanism.>

A formal technique is presented for generating traffic matrices for distributed data networks. Traffic matrices are used by engineers to design new networks or, more typically, plan for changes in an existing network. The technique combines measured data from proposed design changes to calculate a set of traffic matrices. Network hardware and applications are defined by a weighted connection matrix. Traffic is estimated using a set of statistics for different types of applications.>

The problem of selecting minimum cost topologies for the interconnection of Local Area Networks (LANs) with transparent bridges is addressed. A graph theoretic model for a LAN internetwork is developed and an algorithm for calculating internetwork flows is presented. The flow algorithm accounts for O-D flows due to hosts and overhead flows due to the bridge learning algorithm. Internetwork delays are examined and standard queueing models are used to approximate the delay of a packet traversing multiple LANs and bridges. A mathematical program minimizing system cost subject to internetwork flow constraints and delay constraints is formulated. The resulting optimization problem is a complex nonlinear integer program; it is solved using a simulated annealing heuristic algorithm. Numerical experiments have showed the heuristic to be within 0.8% of the optimum for small problems. For large problems the annealing algorithm is evaluated using a statistical technique and results indicate excellent performance. INFORMS Journal on Computing, ISSN 1091-9856, was published as ORSA Journal on Computing from 1989 to 1995 under ISSN 0899-1499.