A fter 40 years of under-investment in the electricity supply, reliability of mains power and the cost of electricity are major concerns for IT departments. Now investment is underway, increasing power costs, more efficient designs are required to reduce operating costs.
Construction has started on the North Island Grid Upgrade which, with an estimated cost of $824 million, will provide 196km of primarily overhead transmission lines by 2012. The first new upgrade since the 1960s, this new transmission line feeds a population that has doubled, and an electricity use that has trebled.
The cost of electricity is likely to increase substantially in the near future, driven by a number of factors. These include a proposed Emissions Trading Scheme, expected to increase the cost of electricity by 0.8 cents per kilowatt hour during the 2010-2012 transition phase; the costs of upgrading the grid; and investment in renewable energy plants such as wind and tidal power.
The data centre factor
As we know, the total power used by data centres has grown dramatically over the past 10 years, and data centres are now estimated to consume 2% of the total power generated in the USA. This is something that is now being actively addressed by the industry through organisations such as the Green Grid (www.thegreengrid.org). This is an organisation of IT professionals who “seek to define best practices for optimising the efficient consumption of power at the IT equipment and facility levels, as well as the manner in which cooling is delivered at these levels”.
One of the ways of doing this is to understand Power Usage Effectiveness (PUE). This is the ratio of total electrical power used in a data centre to power used by the IT equipment alone. A perfectly efficient data centre would have a PUE of 1. The Uptime Institute has shown traditional data centres to have a PUE of 2 to 2.5 – as much power used by the IT equipment as is used to keep the data centre cool and support losses in the power distribution. There have, however, been recent developments for improving PUE and therefore energy efficiency. These include:
Widening temperature and humidity ranges.
Improving air conditioning efficiency, such as containment of hot exhaust air.
Chilled water for transporting heat out of the data centre instead of less efficient refrigeration DX systems.
Using ‘free cooling’ techniques.
Improved monitoring, management and control.
Effective power usage is very important, but we also need to improve electrical resilience and reliability in the data centre. Many data centre managers have already felt the effect of the unreliable electricity grid, and dual power feeds from different parts of the grid are not always possible in New Zealand.
Data centre facilities departments need to engineer a design that removes single points of failure. Although generators with automatic transfer switches (ATSs) ensure that, when mains power fails, the data centre will remain powered, a UPS is required for ride through, and generators need to be maintained and tested once per month.
UPSs should have 10-30 minutes’ standby, selected to match the time your data centre takes to reach maximum operating temperature without air conditioning. Beware, however, as high-density data centres can take as little as five minutes to reach their maximum operating temperature. On sites where chilled water cooling is deployed, a buffer tank of chilled water can be used to keep the temperature of the data centre within temperature limits for the duration of the UPS battery. In anticipation of the event of an extended power failure, you should deploy software to shut down servers – this is available for most platforms including VMware and Hyper-V. Monitored dual power feeds to racks can also help because dual corded servers can then be fed from different circuit breakers. You should also incorporate reliable rack PDUs (power distribution units) with IEC sockets, rather than hardware store power strips.
Overarching all of these improvements, you should strive to improve your management for fault and maintenance control and get your IT staff to take a greater interest in the facility area, once the preserve of the properties departments.
Resilience in the data centre requires fault tolerance and redundancy, but this is achieved at the expense of PUE. A trade-off between resilience and efficiency must be made in line with your business objectives, to provide the best balance of green computing with data centre uptime.