The performance evaluation tool (PET) is a package to aid the designers of architecture and protocols of computer networks. It is specially targeted, but not limited, to IBM's systems network architecture (SNA) networks. The tool allows its users the capability to easily construct simulation models of networks and conduct traffic related performance studies. PET is based on the research queueing package (RESQ)-version 2, which is a package for constructing and analyzing extended queueing network models. The two key features of PET are ease of use and flexibility. The user specifies the network configuration with the aid of an interactive "networking language." This includes a description of the type of representation he/she desires for the various protocols at different layers of the architecture. Based on these specifications, PET builds a RESQ simulation model for the network by piecing together many "submodels." Each submodel is a model of a certain network component (e.g., communication link) or a protocol (e.g., data link control). The submodels may either be prebuilt (stored in the PET library) or built dynamically according to user specifications. There can be several submodels for the same function capturing different levels of detail. Thus, a variety of network models with different types of components and with varying levels of detail can be constructed. Facilities to perform parametric studies and provide user specified output statistics (such as link utilizations and buffer usage statistics) are also available in PET.

Algorithms for effectively routing messages from a source to multiple destination nodes in a store-and-forward computer network are studied. The focus is on minimizing the network cost (NC), which is the sum of weights of the links in the routing path. Several heuristic algorithms are studied for finding the NC minimum path (which is an NP-complete problem). Among them are variations of a minimum spanning tree (MST) heuristic and heuristics for the traveling salesman problem, both of which use global network information. Another set of heuristics examined are based on using only the shortest paths to the destinations. While the MST algorithm has the best worst case performance among all algorithms, a detailed, empirical study of the "average" performance of the algorithms on typical, randomly chosen networks reveals that simpler heuristics are almost as effective. The NC cost measure is also compared to the destination cost (DC), which is the sum of weights of the shortest path distances to all destinations. A scheme of algorithms is given which trades off between NC and DC.

Flow control is proposed as a means of obtaining an "optimal tradeoff" between low delay and high throughput in computer networks. Several versions of "optimal tradeoff" are defined based on network power. A class of algorithms which attempt to optimize network performance are investigated. These algorithms operate on the design principles of dynamic, distributed execution and use of local information. These design principles force the algorithms to be suboptimal, and we thus investigate the relative performance of each in different network configurations. Several properties of power as a network performance objective function are examined. In certain configurations, two variations of network power are unfair to certain users by not permitting them to send any messages. A version of network power ("product of powers") corrects this deficiency. Other properties discussed include the nonconvexity of the generalized power function.

Queueing systems in descrete time that can model certain computer communication systems are considered. First, a single-resource model consisting of a geometric server with feedback facility, to which packets arrive in independent bulks, is analyzed. Then, it is shown that a network of such geometric resources with packets arriving in independent Bernoulli streams (which, in a sense, is a discrete analog of Jackson's network) does not possess the independence property. A byproduct of this result is the fact that output of a geometric resource fed by two or more independent Bernoulli streams is not an independent Bernoulli stream.