Louis McGarry, Sales & Marketing Director examines the role of modular uninterruptible power supplies (UPS) in decentralising colocation data centre infrastructure.
What do you do when you see a space man? Park the car man! I’ve always loved that joke! But seriously, what relevance has it got to colocation data centre infrastructure?
The fact is that the processing capacity of servers is rapidly increasing. Ten processors used to do the job that one can now do. Moor’s law states computer processing power doubles every two years. Yet we also know that the speed at which we create data is growing at an astonishing rate and the situation is only set to escalate.
Datacentres are right to respond to anticipated future demand. However, what also needs to be considered is the equally rapid reduction in IT power consumption. With real-estate prices at a premium and set to continue to rise, plus the high costs of power consumption and maintenance of an oversized legacy UPS, being wise to maximise the use of existing infrastructures and constantly right-sizing appropriate to the load, can minimise running costs and maximise returns for the colocation datacentre.
Colocation datacentres are generally assessed by customers on their PUE (Power Usage Effectiveness), level of resilience and the cost of space. One option to optimise all three elements, is to look at deploying a decentralised infrastructure. We are currently seeing around half of the colocation datacentres we work with moving in this direction.
What is a decentralised infrastructure?
Traditionally datacentres have tended to house oversized critical power infrastructure, in anticipation for future demand. Those that have moved towards a decentralised infrastructure, fence off areas within the facility to create smaller rooms with UPS systems dedicated to individual clients. Advances in modular UPS technology have enabled this approach because datacentre managers have rightly started to ask: why install 1 MW of UPS from day one when you haven’t sold the space?
Installing the full equipment from day one (and paying the capital upfront) often means we see facilities that have never grown to their full potential. Resulting in everything being oversized.
For example: I often see UPS systems designed to run at N+1 that in reality are running at N+3 and sometimes N+4. This happens when the actual load is far less than what was anticipated. Keeping that many UPS ‘awake’ is a significant financial burden not to mention the environmental impact. I believe there is a case to make for education here when it comes to rightsizing and the general management of equipment using ‘pay as you grow’ options. Re-using, re-thinking, re-sizing and re-managing existing infrastructure could enable data centre owners to utilise existing equipment, reducing waste and overall expenditure.
Pay as you grow options enable close control of costs. It means colocation datacentres can scale as they sell the space, which offsets the purchase price of the new UPS modules.
Optimising this approach is made easy with a truly modular UPS. With truly modular, several individual UPS modules are contained within a frame. All the individual modules are effectively UPSs in their own right, all containing a rectifier, inverter, and static switch and all operating online in parallel with each other. For example, eight 20kW UPS modules may typically be contained within a single frame offering a resilient configuration of 140KWs N+1.
With a truly modular system, where the static switch is included in each module, the rest of the modules in the UPS frame continue to protect the load until it can be replaced. A loss of resilience in an isolated area of the datacentre is easier to address than with a standalone UPS where any sort of repair leaves the system vulnerable. If required, it takes moments (around 30 seconds) to ‘hot-swap’ a module while the rest of the modules continue to protect the critical load. At no point does the system need to be transferred to maintenance bypass which puts the load at risk. This increases the level of availability dramatically. In this way, truly modular UPS offer the highest level of resilience in the market. With no single points of failure, availability is maximised, and clients remain protected.
There are further advantages: the assets or modules from a truly modular UPS can also be moved around for re-use within the infrastructure. If one client moves on and another arrives and the module ratings are standardised (say for example all 20kW), then they can simply be re-used elsewhere. Also, if a client is looking to expand by 3 kW it’s far more cost effective to add a 20 kW module than another 50 kW one. Considering module standardisation as part of capacity planning could enable data centre managers to determine the appropriate module rating for the average load step.
The decentralised approach taking advantage of a truly modular UPS, also offers flexibility so the resilience level can be tailored to the client’s needs and budget. Any tier rating can be achieved in any area of the datacentre depending on requirements. This client centred approach means a flexible product and range of pricing which can appeal to a broader range of customers.
In order for Tier 4 datacentres to achieve the required high levels of availability, they attempt to remove all single points of power failure by duplicating UPS systems (2N) and by adding parallel redundancy into each UPS (2N+1). The problem is that such duplication causes the UPS to be operated at a low point on their efficiency curve and hence adversely affect the datacentre’s PUE. The most modern, truly modular UPS, if correctly configured, are able to combine the benefits of high availability with very high efficiency even at low loads because they operate at a lower point on their efficency curve, and therefore offer a very low datacentre PUE.
In this way, by selecting a truly modular UPS and a decentralised infrastructure proper ulitlastion of the UPS and fine edge of efficiency can be achieved.
If a client moves on and chooses a different avenue to protect its data, it’s easy to ramp down, reallocate the asset or isolate the space.
However, a decentralised approach taking advantage of truly modular UPS systems, requires input from experienced engineers who really know how much energy a server rack burns, suppliers with flexible product solutions and end-users and designers who understand the future requirements. Creating a community of experts at the project’s feasibility stage could play a significant role in right sizing and optimising the cost versus resilience of a UPS.
Advancement of knowledge and connected thinking at the design stage and aiming to achieve the best use of infrastructure, will reduce costs and optimise profits for the colocation datacentre. So next time you see a space man, don’t just park the car (man) because you can: think about how it may be best utilised! Borrowed from the words of one great spaceman: it may be one small step for man, but one giant leap for data centre kind.
Article originally featured in Inside Networks Magazine January 2020