The Green Grid has suggested a metric known as Power Usage Efficiency (PUE) for determining datacentre efficiency, which can be calculated by dividing the Total Datacentre Power by the Information Communication Technologies (ICT) Power, i.e. total power supplied to the facility divided by useable IT computing power. However, PUE is a measurement which varies. It’s possible to take an average over a year to establish a datacentre’s PUE as a measure of its overall efficiency but, on a daily, weekly or monthly basis this will change due to a multitude of variables.
The main consumer of power in a datacentre is the load, other attributors are: Uninterruptible Power Supplies (UPS), Heating, Ventilation, and Cooling (HVAC). However, the reality is that everything can affect PUE including room layout, lighting, servers, transformers, cables, generators, system architecture and configuration. Even the seasons and the time of day can have a big impact, external temperatures and the need to control humidity levels will increase the energy used. Legacy equipment will also have a negative impact. As components age, they become less efficient, can generate more heat and so on.
It doesn’t take much in terms of change to one or more of these elements to have a considerable effect on a datacentre’s PUE. For example: something as simple as switching to LED lighting will use significantly less energy than conventional lighting. It also creates less heat, so air-conditioning is reduced.
So, how else can datacentres optimise their energy efficiency when it comes to the UPS element of their facility?
Variable Efficiency Optimisation
Advancements in the latest UPS technology means that losses in efficiency can be minimised to less than 3%. Selecting a UPS that automatically adapts to the load demand, to ensure that it is always hitting its optimum efficiency, can help to make long term gains. Over the lifecycle of the UPS, this will make a big difference, especially when you consider that datacentres are running continuosly for 365 days a year and need a lot of power.
Proving Your Systems Efficiency
When selecting a UPS, buyers need to be diligent. Interpretations of efficiency levels can vary, even in the specifications, and clever marketing can be misleading. How can you be sure that you are really getting the most efficient UPS available on the market?
Check the fine print. So often we see efficiency levels being promoted with caveats, for example: 99% efficiency (in eco-mode/static bypass mode). Running the UPS this way will provide energy savings. However, it means that you are bypassing the rectifier and inverter, defeating the purpose of having a true online double conversion system. A UPS’s true efficiency should only be evaluated based on its performance during true online double conversion.
There are various levels of testing available to prove the performance of a UPS system, for example: efficiency can be proven by the UPS manufacturer during a Factory Acceptance Test (FAT) ahead of installation. However, once you have purchased your UPS and witnessed its level of efficiency, how do you know it will continue to perform this well once installed? Best practice would be to analyse the input and the output power values of the UPS, which will illustrate the losses of the overall system. This type of montioring can be done on an adhoc basis or continuously, by building it into the infrastructure.
Sustainable Lifecycle Replacement
Closer monitoring and management of a system is certainly a key way to manage efficiency. It also enables decisions to be made about the best time to replace components. Due to technological advancements and aging components, we know that legacy UPS systems are not as efficient as new ones. So, total cost of ownership (TCO) calculations can be made to establish if it is better to replace systems sooner or later. It is possible to refresh individual UPS modules rather than an entire system. The infrastructure remains, while individual elements are replaced. The highest quality true modular UPS are so robust now, that parts can actually be re-conditioned, re-certified, re-stamped and re-used. This ensures the components within a system are always in optimal condition.
Some leading datacentres are now looking at sustainable lifecycle replacement programmes on a rolling basis. This is highly preferable than the ‘let’s see if we can get it to limp over the line’ approach employed by some organisations whichcan reduce CapEx, but allows OpEx to rocket. Yes, some UPS solutions can last 15-20 years but is it ideal to keep them going for that long, burning more and more energy than necessary?
Management of battery banks is a further way to optimise energy use. Electricity is expensive, therefore, one approach to reducing costs could be to switch to UPS rather than using energy from the grid. Depleting the battery bank in a controlled manner means that datacentres can cleverly and safely save on electricity. Datacentres will need to ensure that enough stored energy is available in the event of a mains failure to inact their power protection plan, i.e start the generator or conduct a graceful shutdown.
Traditional VRLA batteries will not be able to stand the quantity of discharge and recharge cycles, including the recharge time required to adopt this approach. However, advances in Li-ion battery technology offer this possibility and we are working with several datacentres experimenting with this innovative way of working. Li-ion batteries are capable of literally thousands of cycles. As well as occupying less space, they also work at much higher temperatures meaning less air conditioning is required to keep them cool.
One of our international clients has now chosen Li-ion to support its UPS systems across multiple sites. As a result, they have been able to switch the air conditioning off permanently. Although Li-ion is more expensive to purchase than traditional lead-acid batteries, this organisation has calculated the payback from reduced energy usage alone will be recouped within 4-5 years.
Modular Data Centre Approach
Other innovative approaches we are seeing to reduce energy use within leading datacentres include the introduction of a complete modular infrastructure concept . Rooms within rooms! Hot Aisle Containment Systems (HACS) and Cold Aisle Containment Systems (CACS) encase sections of the datacentre, generally within two rows and prevent the mixing of hot and cold air. The temperature within these systems are controlled separately to the rest of the datacentre room.
These systems are already proven as an effective way to reduce cooling, so why not build on this concept? Grouping together aspects of the infrastructure into managable rooms can significantly reduce or even remove the need for cooling in some areas. For example: UPS systems and associated switchgear could be run warmer (30 – 35 degrees). The batteries could also be in a separate room with an ambient temperature of 22 degrees.
Joined Up Approach
To optimise energy efficiency it’s important to talk to the right people. Joined up thinking, taking into consideration the latest developments and ideas around how to reduce energy is necessary from the outset. So many organisations request a 2D design, but manufacturers and consultants need to evaluate the site thoroughly on the ground, to understand the challenges and come up with workable, long term and flexible solutions.
Once a plan is created many fall in to the trap of ‘cutting costs’ by buying cheaper, inferior elements of the system. CapEx can be reduced but at the expense of escalating OpEx. A long term view is required includling reful TCO calculations to see how higher quality equipment which lasts longer and is more efficient, affects the overall financial and energy usage equation and therefore the PUE.
When working to reduce the energy use of a datacentre, there are many elements to consider. At CENTIEL our technical sales engineers are always available to discuss how to optimise efficiency when designing UPS systems and to help make the relevant calculations. We work in partnership with our clients to ensure they can always maximise system availability and reduce their PUE over the long term.