James Rix, associate director for data centres at Arcadis, outlines some important do’s and don’ts when it comes to keeping things cool in your facility.
Margins in colocation are now lower than they used to be, and many clients are now wanting to pay via a solely PUE mechanism, with PUE as a multiplier on the utility cost. Therefore, more efficient operations are essential to be able to maintain a higher level of profitability.
Significant efficiencies can be gained from CaPEx neutral changes to the data centre or from minimal CaPEx expenditure and best practice primarily around air flow management, containment and supply/return temperatures.
Invariably it can be seen that fans operate on fixed speed rather than a variable speed and temperature set points that are too low. Air flow management is often poor with a lack of blanking panels, too many or too few ventilation floor tiles and a limited understanding of either hot/cold aisle containment allied with negligible or non-existent instrumentation.
Long rows of racks, while seeming efficient, are often not; if an operator only has CRAHs (Computer Room Air Handlers) at one end of the row, the engineers will have to over-pressurise the under-floor plenum to make sure cool air reaches the furthest point. The ‘air thow’ distances of CRAH’s are finite and constrained by the physics of the fans.
One inadvisable solution that has been seen, is the use of a fan assisted floor tiles for ‘hot spot’ cooling. Why use a fan to suck air out of a pressurised underfloor plenum to assist cooling – this only exacerbates the existing hot spots and has the potential to create more elsewhere in the floor and depressurise the floor. Often such devices are single corded and thereby constitute a single point of failure at a critical point within the floor.
Alternate cooling methods abound, from ‘free air’ cooling (it still costs money and equipment to cool the hall so it’s not essentially free – just the air is colder to start with), adiabatic cooling, the use of dry/hybrid coolers, sea water cooling, full immersive cooling, in rack cooling and the list goes on. However, what is right for one facility and/or client is not necessarily right for another.
Co-location for example will be constrained by SLA’s and expensive breach clauses if temperature and/or humidity exceeds the stated figures. It may be that certain client’s equipment has not been ‘hardened’ to be able to run at higher temperatures. Therefore, the whole hall is constrained by the equipment that only performs at lower temperature limits.
The ability to increase ‘free cooling’ hours is vital to reduce cost and improve efficiency.
All these little adjustments can decrease a facility’s energy consumption. On a large facility, in excess of 30mW, the ability to demonstrate over USD$350,000 per year OpEx savings on utility costs would not be at all unusual.
If upper temperature limits are used this will also have the effect, in temperate zones, of increasing the amount of ‘free cooling’ hours that can be used. How though can you understand what the effect will be when changing the set points within a cooling system to an ASHRAE upper limit or changing the UPS to a more efficient type?
This, however, cannot be done with the zeal of one who has just attended a course and is now on a mission to change the world. A baseline first needs to be drawn up to understand how a facility is performing.
Many facilities are sadly lacking in appropriate instrumentation/metering, or that is correctly calibrated and installed and the data generated is being verified as correct.
Once a baseline has been produced on the efficiency of a facility, work can then be undertaken to understand what the most effective strategy will be to reduce utility cost. Often this is by building a virtual energy model from site drawings and using historic data and live feeds from the sites BMS/EPMS,
Other solutions to achieve the same ends could be to supply an independent data gathering solution, not requiring BMS and EPMS connections which will give an accurate interactive picture of a data hall and the ability for the data centre owner to run their own ‘what if’ simulations as needed.
Whilst a raft of measures can be proposed, change needs to take place incrementally to understand what the effects of change are. A good change control process will understand many of these effects, however, just altering the position of ‘knobs and dials’ without any real understanding of the effect is a sure route to calamity.
Continual monitoring of the measurements received from monitoring devices and the interpretation of results is key to increasing the efficiency. Some solutions can produce actionable results within a very short timescale, days rather than months from the point of engagement.