Abstract
This paper discusses some aspects of the decisional challenges implied by the management of modern information technology (IT) systems. These rapidly evolving complex systems are crucial to the overall performance of industrial companies. Two challenges are delineated: (1) the complexity of modern systems requires a system-thinking approach; (2) the critical role of IT in business performance implies taking into account these technological aspects within any business plan. Some IT management frameworks are designed to face these challenges but seem not to provide enough comprehensible decisional support to IT managers in real time. That is why this paper proposes innovative decision-aid methods, designed for some of the activities related to capacity management. Attention is paid to modelling and monitoring activities, through methods that combine qualitative and quantitative analyses. These approaches can provide operational support to IT managers. The added value of the methods developed is demonstrated by practical examples implemented in a semiconductor manufacturing firm. These results are an encouraging first step towards more comprehensive and more effective solutions.
Acknowledgements
The authors would like to thank STMicroelectronics for supporting this research project, Marie-Agnès Girard for her advice about exploratory statistics and system thinking, and Chris Yukna for his help in proofreading.
Funding
This work was supported by STMicroelectronics.
Notes
1. Generally, these last two authors have dealt with information systems complexity and therefore also consider further sources of complexity, like social or organizational aspects. These are not within the scope of this paper, which is focused on the technical dimensions of IT architecture.
2. For more details about the way this survey has been conducted, interested readers may refer to Lutz (Citation2013).
3. For an overview of the full range of IT capacity planning techniques, refer for instance to Jain (Citation1991).
4. When the data contain exceptional fluctuations, standard deviations are overestimated, sometimes dramatically, and the control limits are biased. As a result, control loses efficiency. The Qn robust estimator was used. It offers good properties: 50% breakdown point, high Gaussian efficiency. It is defined proportionally to the .25 quantile of the distances {|xi - xj|;i<j}, where x1, …, xn denote the data (Rousseeuw & Croux, Citation1993).