A Study on the Methods for Performance Design and Improvement in Collaborative Systems

In the conventional system development life cycle (SDLC), system performance and evaluation phase comes after the implementation phase. The aim of this research is to project system performance estimate at the requirement analysis and design phase itself, much before the implementation phase. This study introduces novel tools for modelling the system, evaluating the performance of the system and for improving and enhancing its performance. The types of systems that have been chosen as the object of this study are collaborative systems. In particular, wide-scale inter-collaboration, intra-collaboration and micro-level collaboration within the subsystem engaged in collaboration are analyzed, modelled and their performance is evaluated and improved. Different scenarios of collaborative activity are being examined in detail. At the centre of each scenario is a server model that seeks to represent a particular type of service in collaborative systems. Single and parallel servers model the general type of collaborative activity. Distributed type of service occurs when the same service providing unit (SPU) has to devote itself to a series of tasks. The shifting of the SPUs from server to server, offering service, is best represented visually by Petri nets. The performance of the system represented by Petri nets can also be evaluated and improved qualitatively. Collaborative systems, being client-server in nature, are modelled by using Multi-Context Map (MCM) technique. System evaluation is through GPSS simulation and improvement is by the Expert System reasoning with qualitative rules. MCM captures the workflow in a collaborative system wherein collaborators interact with each other through the exchange of Token, Material and Information (TMI). This triple-input-triple-output is what distinguishes contexts in MCM from ordinary single-input-single-output servers in queueing networks. These three additional interactions present a formidable challenge to improve the system quantitatively. To overcome the computational complexity, we make use of Qualitative Reasoning (QR) in our expert system. At the system performance evaluation stage, the expert system displays to the user the inherent bottlenecks in the system upon analyzing the performance knowledge acquired from GPSS simulation. At the performance improvement stage, the expert system inference engine refers to the structural-knowledge of the system and by applying the heuristics from the expert’s knowledge, draws the parameter-tuning plan to resolve the bottlenecks in the system resulting from underflow, overflow or non-uniformity in TMI flow. The integrated environment of model building, performance evaluation and performance improvement is semi-automatic by design, and as demonstrated by the successful application to the performance design and