Maintenance and energy management needs for data centers
Mike Proctor—March 2008
WHEN MANAGING A DATA CENTRE is part of a facilities manager’s remit, the requirements for maintenance are likely to be very different from other types of buildings. In particular, it needs to be tailored to match the criticality of the data centre to the organisation’s operations.
A further complication is the fact that one organisation’s ‘data centre’ may be a purpose built facility dedicated to the processing and secure storage of data; while another’s might be a room that houses the central servers, located in a general office building. To further complicate matters, if the latter arrangement isn’t purposedesigned it may share services with other, less critical, areas of the building.
Similarly, the level of criticality can vary considerably. For example, a ‘data centre’ may be a repository for information that the organisation can manage without for a day or so. Or the data may be absolutely critical to maintaining operations. A typical financial institution, for instance, can lose between £40,000 and £180,000 per minute if its data centre is disrupted.
Given this variation, it’s clear that there can be no ‘off the shelf’ approach to maintaining the services that keep data centres operational. The maintenance needs to be planned and implemented on the basis of a proper understanding of every aspect of each system and the potential impact of every action. This is in contrast to traditional, some might say ‘old fashioned’, building services maintenance contracts, where each item of plant is serviced to a pre-set ‘recipe’ that will suit the average requirements of the average building.
While all data centres offer a challenge to the maintenance contractor, it’s very often the nondedicated facilities that require a more innovative approach to mitigate the vulnerabilities of the system.
For example, in a purpose-designed data centre the central plant will, or should, have N+1 resilience. N is the plant capacity that should be sufficient for normal operation and the +1 represents additional standby plant that is there as a safeguard against catastrophic failure. So if two 500kW chillers are needed for anticipated normal operation, three 500kW chillers will be installed.
Maintaining this level of resilience is also a critical element of the maintenance regime. So the maintenance should be backed by condition monitoring, using sophisticated techniques such as thermography, vibration analysis and ultrasonic diagnosis to build a picture of the condition of each item of plant. Equally importantly, the maintenance technicians should be fully trained in the use of these diagnostic procedures. This approach is invaluable in identifying problems as they first emerge so that prompt action can be taken to nip them in the bud.
Additional peace of mind may come from 24 hour monitoring of critical alarms, either on site or from a remote bureau location.
All of these principles also apply to data rooms within buildings that haven’t been purpose designed. If that data room is served by dedicated building services plant this is relatively straightforward within the criteria described above. But if the services are shared with other areas the expertise and innovation of the maintenance contractor really comes into its own.
Again, monitoring the condition of the plant will be just as important, if not more so. If the services haven’t been designed to protect data, they are unlikely to have N+1 resilience, so understanding the plant’s condition and predicting problems becomes even more important.
An experienced contractor will also be able to advise on the feasibility of installing additional back-up plant, or it may be possible to fine tune the zoning of services. In this way, if there is a problem, the available cooling capacity or standby power can be diverted to the areas where it’s most needed.
These decisions will often be based on a balance between the investment required and the criticality of the data to the enterprise. The important thing is to have meaningful information on which to base those decisions.
Energy hungry data
A recent study at the Lawrence Berkeley National Laboratory in California found that worldwide power consumption by servers has doubled between 2000 and 2005. This means that information and communications technology now accounts for around 2 per cent of the world’s global CO2 emissions — on a par with the aviation industry.
Clearly this has implications for the services. For example, servers using more power will generate more heat, so there are greater demands on cooling systems as well as on the building’s power infrastructure.
For this reason, we believe very strongly that M&E maintenance contractors should take much more responsibility for managing energy efficiency in parallel with maintaining the plant. A good way of achieving this is to make the management of energy consumption, and associated risk, a contractual requirement. This may also include installation of more energyefficient plant and extend to renewable energy sources — perhaps in relation to adding extra standby capacity where needed, as discussed above.
Underlying all of these issues is the need to have a flexible maintenance arrangement, that is tailored to your needs and is easily adapted as those needs change.
Mike Proctor is London Regional Manager at Cofathec www.cofathec.co.uk
