A measurement oriented approach to e-commerce distributed transactions based upon a mobile/intelligent agent framework

Internet based Business-to-Business (B2B) transactions are a rapidly increasing portion of many businesses' market place. However there are many challenges that must be overcome in order for a business to benefit from these types of transactions. Standard transaction techniques such as the linear-sequential method of transaction management and the preservation of the ACID properties have been the topic of much research and constant debate. These standard transaction techniques have been shown to be problematic because they potentially cause extended record locking. Extended record locking is noted as producing a higher degree of coupling in these Business-to-Business transactions.; The problems of extended record locking and high coupling are the focus of the research produced in this dissertation. The initial focus of the research was to develop a framework using a parallel approach to reduce the average record locking time in a Business-to-Business Transaction while preserving the transaction ACID properties. We have shown statistically that the parallel approach reduces average record locking time by 30%. Using our parallel technique allows developers to re-use the ACID-based techniques which provide transaction with the means to ensure consistency.; The final part of this research has focused on a suite of metrics that measure the amount of coupling in a Business-to-Business system. Our metrics meet all of the required theoretical criteria based on the Kitchenham, Pfleeger, and Fenton metrics validation framework. We also provide an empirical validation using statistical correlation to a set of evaluators.