The Case for Medium-Sized Regional Data Centres

Cloud computing is widely associated with majorcapital investment in mega data centres, housing expensive bladeservers and storage area networks. In this paper we argue that amodular approach to building local or regional data centres usingcommodity hardware and open source hardware can produce acost effective solution that better addresses the goals of cloudcomputing, and provides a scalable architecture that meets theservice requirements of a high quality data centre.In support of this goal, we provide data that supports threeresearch hypotheses:1. that central processor unit (CPU) resources are notnormally limiting;2. that disk I/O transactions (TPS) are more oftenlimiting, but this can be mitigated by maximizing theTPS-CPU ratio;3. that customer CPU loads are generally static andsmall.Our results indicate that the modular, commodity hardwarebased architecture is near optimal. This is a very significantresult, as it opens the door to alternative business models for theprovision of data centres that significantly reduce the need formajor up-front capital investment