by Suzanne Niles & Patrick Donovan
IT virtualization, the engine behind cloud computing, can have significant consequences on the data center physical infrastructure (DCPI). Higher power densities that often result can challenge the cooling capabilities of an existing system. Reduced overall energy consumption that typically results from physical server consolidation may actually worsen the data center’s power usage effectiveness (PUE). Dynamic loads that vary in time and location may heighten the risk of downtime if rack-level power and cooling health are not understood and considered. Finally, the fault-tolerant nature of a highly virtualized environment could raise questions about the level of redundancy required in the physical infrastructure. These particular effects of virtualization are discussed and possible solutions
Virtualizing a data center’s IT resources can have certain consequences related to the physical infrastructure. If these impacts and consequences are ignored, the broad benefits of virtualization and cloud computing can be limited or compromised, and in some cases, severely so. Areas of high density can develop after server consolidation takes place which may result in hot spots that can then lead to hardware failure. Various methods exist to ensure the cooling system has the means and capacity to reliably cool high density equipment. PUE can significantly worsen after consolidation occurs. By optimizing the power and cooling systems to better match this now reduced IT load, PUE can be significantly improved. This optimization is made much easier if scalable and modular systems are used. Dynamic IT loads which can change automatically in both time and location may unintentionally be put at risk if power and cooling status is not first considered at an individual rack level. Careful planning and on-going management is required to ensure VMs are only placed where healthy power and cooling exists. By constructing sound VM policies and by integrating DCIM software with the VM manager, this on-going management can be automated. Finally, the high level of fault tolerance that is possible with today’s VM manager software makes it possible to employ a less redundant power and cooling infrastructure. Such a strategy can save time, space, energy and significantly lower capital costs. Implementing the solutions described in this paper will keep a highly virtualized data center running with greater reliability, efficiency, and with expanded flexibility to meet highly dynamic compute power demand.