IT CAPACITY FORECASTING: STATISTICAL MODELLING PROCESS AND APPLICATIONS FOR SEMICONDUCTOR INDUSTRY. M. Lutz* **. X. Boucher.* O. Roustant*. X. Ambrosioni** *ENSMSE-G2I, 158 cours Fauriel, 42100 Saint Etienne, France (Tel: + 33-4-38 92-21-98; e-mail: lutz@emse.fr). ** STMicroelectronics, Crolles, France Abstract: The objective of this paper is to present a modeling procedure, based on the use of statistical methods, to develop predictive models for Information Technology (IT) capacity planning. The goal of these models is to be able to SUHGLFW ,7 YDULDEOHV¶ EHKDYLRXU, using initial inputs provided by high-level business explanatory variables. The modelling process consists in two main sequences, exploration and explanation, divided into modelling steps, which mix qualitative and quantitative analyses. The modelling process builds a progressive understanding of relationships among the different levels of the IT architecture. The final outcome is a statistical model, linking IT resource utilization to manufacturing activities variables. An application of the modelling process is developed in the context of the STMicroelectronics information system. As a result, two predictive models which focus on subsystems of the whole information system, link the physical seUYHUV¶ GDLO\ use to manufacturing variables such as level of production, manufacturing processes deployment and fabrication throughput. Keywords: information system, IT capacity planning, statistical analysis, semiconductor industry 1. INTRODUCTION AND RESEARCH BACKGROUND An Information System is a critical factor when bettering modern manufacturing systems efficiency (Cardinali, 1992; Gowan and Mathieu, 1996). This is notably true within the current application field: semiconductor industry (Leachman and HRGJHV  ,QIRUPDWLRQ 6\VWHP¶V HIILFLHQF\ LWVHOI LV based on an adequate specification of firmsInformation Technologies (IT) architectures (Earl, 1989). Consequently, IT managers must basically anticipate to provide without interruption enough IT resRXUFHV WR VXSSRUW ILUP¶V RSHUDWLRQV plans and forecasts so that they function without interruption (Kloesterboer, 2011). Therefore, quantitative IT capacity models are of great LQWHUHVW WR VXSSRUW ,7 PDQDJHUV¶ GHFLVLRQ SURFHVVes. The ITIL (Information Technology Infrastructure Library) framework recommends three main approaches for IT capacity modelling: analytical & simulation modelling and as well as trend analysis (Kloesterboer, 2011). The purpose of such models is to quantify the level of IT resources needed for a given level of workload, forecasted by IT managers on the basis of hypothetical scenarios (Jain, 1991; Gunther, 2007). However, usual IT capacity planning studies in the field of computer systems engineering are still struggling to build such models, with an operational link to high-level business or industrial variables. For instance, in a recent publication it was stated³You may not be able to make a direct correlation between the business trends and your IT capacity, but you should at least look at whatever your organization publishes and see how it compare your component and service trends´ (Kloesterboer, 2011). The purpose of our paper is to present an answer to such a limitation. Indeed, we intend to demonstrate that 1) links between IT resources and high-level industrial activities (here silicon wafer manufacture) can be established; 2) a structured modelling process can formalize rigorously such links. This paper presents some results of PhD research in collaboration with the Crolles 300 IT Department of the company STMicroelectronics. From an industrial perspective, the objective is to build a model with good predictive features, based on the link among IT resources variables and high-level industrial variables. The impact of such model would be not only to ease communication among IT and production managers but also to compare and analyse IT capacity hypothetical scenarios, based on industrial input variables. Academically, many research papers deal with the link between semiconductor manufacturing and production equipment capacity (Geng and Jiang, 2009). However, to the best of our knowledge, few if any publications link semiconductor manufacturing system activity to the capacity of its IT architecture. More generally, it can be observed that IT capacity planning has little been explored by industrial engineering as of yet. The results presented in this paper only constitute an important initial step for much further research. The final objective will be to completely design a decision- making tool, which will encompass the use of quantitative models in a wider industrial scope. However, this decision tool is not the purpose of this paper and will be addressed in future publications. The focus of the paper is to present the modelling process used to build the quantitative models that will be required. Proceedings of the 14th IFAC Symposium on Information Control Problems in Manufacturing Bucharest, Romania, May 23-25, 2012 978-3-902661-98-2/12/$20.00 © 2012 IFAC 1568 10.3182/20120523-3-RO-2023.00043