Virtualisation is an undisputed leap forward in data centre evolution – it saves energy, increases computing throughput, frees up floor space and facilitates load migration and disaster recovery. However, less well known is the extent to which virtualisation’s entitlement can be multiplied if power and cooling infrastructure is optimised to align with a new, leaner IT profile.
While the virtualisation challenges of dynamic and migrating high-density loads, underloading of power and cooling systems and the need to ensure that capacity meets demand at the row, rack and server level are not new or unique to virtualised data centres, the combined simultaneous effects of virtualisation are focusing attention on them with a new urgency, especially in light of the growing interest in energy efficiency.
There are three key points to understand about virtualisation as it relates to the data centre’s power and cooling infrastructure:
- Power and cooling technology is available today to safeguard availability and meet the challenges of density and dynamic power that often accompany virtualisation and consolidation.
- Power consumption will always be lower after virtualising as a result of computing consolidation and physical reduction of the amount of IT equipment. With optimised power and cooling to minimise unused capacity, power consumption will typically be less.
- Data centre infrastructure efficiency, measured as Power Usage Effectiveness (PUE), will worsen after virtualising, due to fixed losses in unused power and cooling capacity. With optimised power and cooling to minimise unused capacity, power and cooling efficiency PUE can be brought back to nearly pre-virtualisation levels – sometimes even better, depending on the nature of improvements to the cooling architecture.
The comparison of pre- and post-virtualisation power consumption involves two concepts relatively new to data centre cost analysis. As mentioned above, the first is “fixed loss” – the amount of power consumed by devices and systems regardless of load – which is responsible for the often surprising inefficiency of underloaded systems.
The second is the distinction between energy consumption and energy efficiency, which can confuse a comparison of energy savings. Even without a parallel upgrade to power and cooling, virtualisation will always lower the electrical bill, but not usually as much as might be expected, because the presence of fixed loss in power and cooling systems and in spite of a reduction in power consumption by the data centre, the data centre’s PUE is typically worse after virtualising due to the inefficiency of underloaded power and cooling systems.
This lowered efficiency indicates room for improvement in power and cooling systems – it is, in effect, a measure of the potential for extracting even more value per energy Shilling.When virtualising, a parallel upgrade of power and cooling infrastructure will optimise both architecture and operation in a number of ways that safeguard availability, enhance manageability, lower power consumption and increase efficiency.
In fact, row-based cooling, scalable power, variable capacity cooling and capacity management tools specifically are essential elements in realising virtualisation’s full potential in cost reduction, efficiency and reliability. These solutions are based on design principles that simultaneously resolve functional challenges, reduce power consumption and increase efficiency.
In addition, the shift towards virtualisation, with its new challenges for physical infrastructure, re-emphasises the need for integrated solutions using a holistic approach – that will consider everything together and make it work as a system!
By Jonathan Duncan, director for the Central and North East Africa region at APC by Schneider Electric