Capacity Planning for
Data Centres

Traditionally, Capacity planning has been a performance engineering activity that determines the necessary capacity (processing power, storage, network, etc..)  for new applications. Capacity planning has been an off -line activity, part of pre-sales, sales and deployment activities. In general, a capacity planning tool used simple heuristics and knowledge about the workload of a new application and knowledge about the performance characteristics of the hardware. Prior knowledge and data about similar deployments played also a big role in a successful capacity planning.


Recently, capacity planning started to be associated with on-line activities. The emergence of clouds and large data centres adds new requirements on the capacity planning activity.  If in the a service used its own dedicated hardware, now,  a new service will share the infrastructure with existing ones, competing for the same resources. Virtualization makes possible the isolation of the virtual machines and therefore the sharing of the same resources. Multi-tenancy in the web applications deployed in a cloud makes the sharing even more fine granular and pervasive. All these make the accurate estimation of the performance and capacity more difficult because one has to account for interferences among services. However, there are new opportunities  that the on-line activities bring, the availability of more data, the possibility to query the availability of resources as well as the opportunity to learn from repetitive deployments.


This project proposes to develop a capacity planning and estimation tool that satisfies the requirements of the new IT infrastructures. A special attention will be given to cloud infrastructures. The tool will evaluate the performance characteristics of the IT infrastructure, the available capacity, estimate the workload conditions, and calculate new deployments plans for new and existing services.