IBM Research has a distinguished history in the theory and practice of performance analysis, modeling and optimization. Fundamental and foundational contributions have been made in a number of important areas, including: product-form queueing networks; optimal control and scheduling in queueing networks; stochastic ordering and majorization; rare-event and parallel simulation; matrix-analytic analysis of stochastic models; polling systems; and performance modeling tools. These have played a critical role in understanding important problems in the design, development, management and planning of complex systems.
Performance modeling has been and continues to be of great practical and theoretical importance in research labs in the design, development and optimization of computer and communication systems and applications. With the advent of new technologies such as Autonomic Computing and new business models such as On Demand, performance modeling and analysis has become increasingly important for the delivery of high-quality and continuously available services. Researchers at IBM have carried out a broad spectrum of research activities from the application of more empirical methods (ranging from experimental tuning of simple existing models up to building and experimenting with prototype implementations) and simulation to more sophisticated mathematical methods.
From theoretical perspective, a key effort is on the understanding the causes and the impact of the long-range dependencies, heavy-tail distributions and non-stationarity of traffic. It is known that these complexities can significantly impact on performance (several orders of magnitude). IBM researchers have further established fundamental results for the mathematical analysis of stochastic models and queueing networks exhibiting these complex characteristics, which include approximate methods and asymptotic results. Another major on-going effort is the investigation of optimization and control of e-business infrastructures and applications. By deploying various mathematical studies using queueing theory, probabilistic modeling, stochastic scheduling and control theory etc, IBM researchers provide solutions to theoretical issues related to areas such as quality of service, scalability, dynamic scheduling algorithms, load balancing, admission control, buffer management, inventory management, profit and risk management, with service-level-agreements constraints.
Such theoretical results have been applied to a wide variety of problems and have allowed the development of practical solutions, ranging from hardware design to software tuning, from low-level system architecture to high-level infrastructure, from component performance to end-to-end quality of service, from applications to business process, from performance testing to service-level agreement, from system control to autonomic computing... These theoretical results have allowed IBM to provide state-of-the-art solutions in a number of areas including traffic generation and benchmarking, model validation, capacity planning, workload and performance forecasting, power-consumption models, generating and serving dynamic content, resource control and management, cooperative caching, dynamic offload, and network and server design. Such theoretical investigations have also guided tool development in the performance modeling area. Among these are a, a flexible framework and a comprehensive toolset developed for workload characterization, performance modeling and analysis, and on-line control. This toolset can be used to provide capacity planning, performance prediction and performance engineering solutions for computer systems as well as business processes.