Beyond virtualisation’s undisputed IT benefits is a parallel story of substantial efficiency benefits which can be derived through optimisation of the data centre physical infrastructure (DCPI) that supports the servers.

By Paul Tyrer, VP UK & Ireland, APC by Schneider Electric

In addition to a reduction in space requirements, a major benefits of server virtualisation is reduced power consumption. However, it may come as a surprise that facility PUE will increase if the power and cooling supply is not addressed together with the server load. Further, availability takes a higher priority as the criticality of individual physical servers is raised.

From a physical infrastructure perspective, virtualisation creates changes in the datacentre which presents challenges for both its effectiveness (how well it safeguards the IT) and efficiency (how well it conserves power while performing its job). While an upgrade of power and cooling is not required to make virtualisation work, the greatest power efficiency will be achieved through power and cooling that responds to challenges characterised as follows:
1. Dynamic and migrating high-density loads
2. Underloading of power and cooling infrastructure
3. Ensuring power and cooling demand is met at row, rack and server levels

The shift towards virtualisation re-emphasises the need for integrated solutions and the adoption of an holistic approach – considering the IT load and the physical infrastructure as a single system - as identified in the EU Code of Conduct for Datacentre Efficiency.

1.  Dynamic and Migrating High Density Loads
While virtualisation may reduce overall IT power consumption in the room, virtualised servers may be organised in ways that creates localised hot spots. Not only does server and heat density increase, but virtualisation also allows applications to be dynamically moved, started and stopped, resulting in a thermal profile which can shift at different times, unseen and no visible changes in equipment.

In legacy datacentres, perimeter-based cooling could typically be configured to provide adequate cooling. However, virtualised, high density environments call for a cooling system which can detect and automatically respond to thermal variations. Row-based cooling meets the essential criteria to provide a short air path between cooling and load, and a dynamic response to changes in load conditions.

Row based cooling substantially increases the efficiency of the system by providing cool air only where needed, only when needed and only in the amount needed. In addition, it is an enabling technology for the deployment of high density zones and Hot Aisle Containment Solutions (HACS), as well as providing a scalable foundation to meet the second challenge.

2.  Underloading of Power and Cooling Infrastructure
With the onset of the virtualisation trend came the opportunity to downsize power and cooling equipment to avoid the wasted operational expense of infrastructure not being used. Since virtualisation can significantly reduce the IT load, oversizing becomes an important efficiency issue because DCPI equipment tends to be reduced in efficiency as the load reduces. Power and cooling equipment which is scalable provides an answer.
Power and cooling devices that can scale in capacity will reduce fixed losses and increase efficiency, facilitating not only downsizing to follow IT consolidation, but also upsizing as the load grows again.  However, in addition to efficiency benefits, optimally sized physical infrastructure can also protect against the effects of extreme underloading which can include safety shutdowns on compressors, short-cycling of compressors with a consequent reduction in equipment life as well as unwelcome issues for standby generators.

3.  Ensuring Power and Cooling Capacity
The dynamic nature of virtualised computing demands accurate, timely and actionable information about power and cooling capacities to ensure that both resources are in supply to support a rapidly changing load profile.

Capacity Management answers the need for a system approach the datacentre, providing instrumentation and software for real-time monitoring and analysis of three critical capacities; power, cooling and physical space. For any location where a new or reconfigured IT deployment is being considered, all three must be available in sufficient quantity. Capacity Management therefore enable resources to be utilised effectively and efficiently throughout the datacentre through continuous real-time visibility of capacities at rack and server level.

An effective capacity management system uses automated intelligence and modelling to suggest the optimum location for adding equipment, to predict the effect of proposed changes and to recognise conditions and trends in time for corrective action to be taken. Capacity management that comprehends server locations and loads, power and cooling capacity available to servers, temperature fluctuations and power consumption not only protects against downtime from localised shortages of resources, but also increases datacentre efficiency by optimising the use of power, cooling and space.