| Three essays in this dissertation study several problems in Information Systems design. Each essay uses methodologies such as dynamic optimization, numerical simulation, and game theory to tackle these issues.; The first essay addresses the problem of determining the optimal coordination policy when the system is under construction. The basic model in this essay studies the coordination of activities between coding and testing in order to determine how best to complete a project which has a fixed duration. For a relaxed project, project managers should gradually shift focus to testing as more coding work has been done. For a tight one, the testing effort shouldn't be pushed until the end of the project. The extended model jointly considers the development and release decisions, and integrates the customer and market factors into the early decisions concerning allocating resources during the development phase. For future work, a discrete version of the dynamic optimization model is needed with finite cost of reallocating resources.; The second essay studies how software companies use software upgrades to deal with future demand uncertainty by offering the base product in the beginning and the upgrade in the future if it's profitable to do so. We obtain some seemingly counter-intuitive results and relate them with insights and guidelines of software project management. Under demand uncertainty, producers will offer more features in the beginning in some situations, even if the component integration cost turns out to be higher or the network externality turns out to be higher. Our monopolistic model can be extended to study competition between software producers. New behaviors in software production are expected to arise from the competition among different companies. Also, how can the company design the right incentive for the current project team to invest upgrade effort now while a possible different future team will collect the future benefit? Incentive-compatible sharing mechanism should be designed to maximize the value of real options for the firm as a whole.; The third essay proposes and evaluates three hierarchical caching polices in the corporate network environment: the base, GLRU and L-policies. L-policy controls the degree of duplication in the browsers and the other two are shown to be the two extremes of L-policy: full duplication for base policy and no duplication for GLRU. Numerical simulation is carried out to verify that results derived from somewhat restrictive assumptions still hold when those assumptions are relaxed. So far we only consider the demand of each document is correlated not only over time. When different users have correlated demand for a set of documents, how will the performances of caching policies change?... |
