Datacentre and rack space is the foundation of any colocation service. Power and cooling are perhaps the second most important components.
A reliable and well regulated source of power in a datacentre is important to ensure the continued operation and availability of your ICT equipment. In addition to its power demands, as a result of its operation, ICT equipment generates heat. Due to this heat output, constant and reliable cooling is needed to ensure the temperature and humidity in a datacentre are regulated.
As a general rule, as the power density and the amount of equipment increases, then so do the power and cooling demands. When opting for a colocation service, the price you would pay would be made up of the rack cost, power cost and network charges.
As power utilisation increases, so do the power charges and ultimately these costs are borne by the customer. Software and equipment vendors are constantly trying to identify more energy efficient designs and modes of operation, to improve efficiencies and reduce overheads relating to power and cooling.
Cost savings are only one of the reasons: there are also environmental justifications to conserve power and therefore limit the impact on the environment. It was recently reported that 10% of the worlds electricity consumption is from IT systems. Datacentre power growth is set to continue at a rate of 12% per annum.
The exact power specification a hosting operator can provide will vary between datacentres, with different levels of resilience and service options available. Diverse A+B supplies at the rack-level would be the most common configuration when considering a resilient solution. Each supply would have a diverse logical and physical path through the power infrastructure, using separate distribution routes, UPS systems and diesel generators.
Power / Electricity
A datacentre will supply a protected mains supply to the rack . The electricity supply would be within the bounds of a defined range of acceptable characteristics (e.g. voltage, frequency etc). A datacentre provider will have invested in a range of equipment to ensure the highest quality supply and unparalleled levels of availability and resilience. This investment is to ensure there is adequate protection against black outs, brown outs and similar power anomalies.
To achieve the highest levels of power provision, a datacentre operator will employ a number of mechanisms to deliver this reliable, robust and highly available electricity supply. The most common components a datacentre customer should be concerned about are outlined below :
The first component is the electricity supply from the mains grid. This is perhaps the most important consideration when selecting a jurisdiction or country to host your ICT equipment; the reliability, scalability and availability of a high quality mains supply. A mains supply is typically delivered through a high-voltage network (from the power stations/grid), into a sub-station (and transformer) which then converts the HV power into an LV supply.
A robust and reliable mains supply is a key factor that will ultimately form part of the reason an operator has selected a particular area to operate their datacentre. Heywood House, our IOM datacentre is based in the Isle of Man. The Island has a world-class electricity infrastructure with its own self generating power plants and a further interconncet to the UK national grid. As a body, the Manx Electricity Authority is a leading electricity supplier, benefiting from constant investment and ensuring the highest quality electricity supplies are available to Isle of Man datacentres such as our own.
I had explained in an earlier article the concerns some companies have with regard to the UK’s future power capacity. It is for these reasons the Isle of Man is considered a leading destination for disaster recovery, e-gaming and hosting operations.
UPS (Uninterruptable Power Supply)
A UPS system has two main tasks: (i) to regulate and clean the mains supply and (ii) to provide short-term backup in the event of a mains failure. The backup system is facilitated through the use of batteries whilst the mains optimisation process is accomplished through a range of components and complex circuitry.
In the event of a mains failure, the amount of time a UPS can sustain the total load is typically limited to around several minutes (rather than hours). It is for this reason that UPS systems are commonly used to provide “ride-through” power until the larger diesel generators fire up and take over the load.
It would be common for a datacentre to have “banks” of UPS systems to ensure there are several devices in a fault-tolerant setup.
Wi-Manx’s Isle of Man datacentre features banks made up of multiple UPS systems, on both the “A” and “B” supplies.
I have described how UPS systems provide a clean source of power and a “ride-through” supply in the event of a mains failure. I also explained how a UPS system can provide several minutes of backup power when the mains supply is lost. Clearly, several minutes of backup power is not sufficient and it is for this reason a datacentre facility should be backed up by one or more diesel generator systems. A diesel generator is in effect a personal power station. Running on diesel, the generators can generate sufficient electricity to power the facility out-right, for hours or days, in the absence of a mains supply. The generators have a diesel supply so provided the generators are serviced with a readily available diesel supply, a generator can safeguard against a mains failure for several days.
In the event of a power failure, a diesel generator may take several seconds to start – this is where it becomes apparent that the UPS and generator pairing are critical. When the mains fails, the UPS immediately takes over and the equipment runs on battery until the diesel generators start. After a successful start, the UPS reverts back to its “mains” supply, which is in fact now being supplied by the generators.
Diesel generators are the most common form of generator although there are several alternatives including Flywheel or Kinetic generators, which often combine the function of a UPS and Diesel generator into one.
In the next article I will describe how power is distributed within a datacentre and how the power consumption is typically monitored.
PDU (Power Distribution Units)
The PDUs are devices that distribute power from the UPS protected supplies to the datacentre racks. Best practice would normally dictate the need for multiple PDUs to distribute power within a datacentre. This follows on with the A+B paradigm, ensuring there is no single point of failure throughout the chain.
The PDU takes a high capacity supply from the UPS systems and from that, distributes multiple smaller supplies to the individual equipment racks. The datacentre PDUs are akin to the consumer units you would have at home, naturally built to much higher tolerances and capacities and with a wealth of high end features.
In our next article, I will outline the power distribution options and methods within a datacentre rack. I will also outline a number of useful tools that can enable you to estimate the amount of power you may consume. Finally, I will explain why technology companies (and server vendors) are constantly striving to develop more energy efficient server equipment and what this means to our datacentre tenants.