Performance of information infrastructures: A concept comparison framework with a QoS mapping model

This research presents a framework for the evaluation of the end-to-end performance of information systems with respect to the real-time data transmission. While the evaluation of performance of the packet-switched communication networks at the network level has been defined by several standards (e.g., ITU-T Rec. Y.1540), the problem of achieving a similar definition for the end-to-end data transport in information systems remains open. Therefore, the main focus of this research is to investigate the specification of end-to-end performance characteristics. To this end, a model has been developed that integrates the network-level performance characteristics with the end-to-end performance requirements.;The research approach involves three steps. First, various terminologies describe the performance of "information infrastructure" are clarified by developing a systematic comparison framework. Second, a set of the end-to-end performance indicators is defined. Third, a QoS mapping model that translates the objective packet transmission performance characteristics (Network QoS) into the frame performance characteristics (Application QoS) is developed.;Following this approach, a critical analysis of the terminology used in the concepts of dependability, fault-tolerance, reliability, security, and survivability has been done. It has been concluded that these concepts can be used almost interchangeably. To measure end-to-end performance, the quality of application layer frames transmitted in real time has been specified in terms of two parameters; one is the ratio of the number and distribution of lost packets per frame to the total number of packets in a frame and the other is the frame delay. These two parameters have been expressed in terms of packet losses and packet delay which are modeled in an analogous way to the real traffic.;The effects of packet loss and delay on the quality of frames were investigated using Monte Carlo simulation. As a result of this investigation a linear relationship with a varying degree of statistical dependencies is found between both layers. The conclusion drawn is that under a specific network environment the lower layer performance parameters can be quantifiably mapped into the corresponding end-to-end upper layer characteristics. These statistics are of significant interests to information system engineers trying to design or optimize QoS support equipments.