Pricing, resource allocation, and incentive issues in IT infrastructure services: A focus on service level agreements.

摘要

Maintenance and management of Information Technology (IT) infrastructure is increasingly taking a services-orientation. Despite its appeal, high failure rates are cited due to conflicts between customer and provider organizations arising from complexities in service demand and interrelationships among different services. Such conflicts call for research efforts to provide insights into demand management and performance evaluation processes to formulate effective Service Level Agreements (SLA). To address these issues, this dissertation focuses on the interrelationship of service price, resource allocation, performance, and incentives through a series of three interrelated studies.; The first study investigates heterogeneous demand within an organization and its effect on resource allocation decisions. An analytical model for SLA formulation as an alternative to the current industry practice is developed. Comparisons using computational experiments, while indicate the superiority of the proposed mechanism from a systems perspective, also point to the opposing effects of resource utilization as a primary source of conflict. Subsequently, the second study explores the role of sharing demand forecast in reducing such conflicts. A resource allocation mechanism incorporating periodic information sharing is developed and tested under varying levels of information accuracy and granularity. Results establish the efficacy of the proposed mechanism in reducing conflicts, even when forecast is noisy. Incentive alignment is the objective of the third study. In an IT Infrastructure service arrangement, activities supporting a service may impact the efficiency, effectiveness, and/or performance of other related services. This may lead to misaligning provider's effort allocation decisions and customer organization's interests. A formal process-modeling framework is adopted to identify patterns of interdependence. This knowledge is subsequently utilized in deriving optimal reward/penalty schemes to provide theoretical insights to incentive structures.; Taken together, this dissertation research provides significant insights into the practice of IT Infrastructure service provisioning that may benefit both customer and provider organizations. The dynamic approach to SLA formulation is an attractive alternative to consider instead of traditional fixed-price single-quality level SLAs. Exchanging demand information, even if noisy, is shown to be useful in maintaining service levels efficiently. Finally, the notion of service interdependence significantly changes the way incentives should be designed in practice. A process approach to characterize IT services can help in getting better insights on relative importance of different performance measures---a crucial element of SLAs.