A System dynamics approach to data center capacity planning - A case study
MetadataShow full item record
This thesis is an empirical study where the System Dynamics methodology is applied to help the Chief Technical Officer of a Norwegian IT company, operating in the cloud computing industry, in planning for future data center capacity. Put simply, cloud computing is the provisioning of centralized IT services and infrastructure to businesses in an on-demand, reliable, and inexpensive fashion, which is why it is sometimes loosely referred to as computing as a utility'. The client's main interest in this project is to gain an analysis tool that can help in estimating the point in time at which the capacity limit of the company's data center in Oslo will be reached. This is a critical question for the business since setting up a new data center has a lead time of around one year, and it is essential to start planning for such an effort well beforehand. In this thesis, a System Dynamics model is built for this purpose, with its structure based on empirical knowledge elicited from the client of the project. Rigorous testing is applied to build confidence in the reliability and usefulness of the model. The model structure successfully replicates historical behavior of important variables in the system. The established robustness of the model qualifies it as suitable to use for policy and scenario testing. A few examples of such tests are carried out and documented in this report, including various tests regarding the central question of when the data center's capacity limit will be reached. This model can eventually become the basis of a management flight simulator that the client could use to try out different policies to see their consequences before implementing them in the real world. This project has been carried out with two overarching purposes, one professional and one academic. The professional goal, as already mentioned, is to help the client in medium-term capacity planning. The academic aspiration of the thesis, however, is to establish the usefulness of the System Dynamics methodology in data center planning and cloud computing business fields. To the best of the author's knowledge, no previous System Dynamics works have been carried out in this area. Yet, being dominated by aging chains, co-flows, accumulations, delays, and feedbacks, data center management is in this thesis demonstrated to be a promising area for applying System Dynamics.