14 MODULAR DATA CENTRES
14 MODULAR DATA CENTRES
Integrated Modularity
measured in weeks not months ; otherwise the benefits of matching supply and demand are clearly lost . Furthermore , when an additional module is deployed to site , it must be done with absolutely zero impact on the day-to-day operations of the existing facility , and then there is the question of ongoing management . A data centre built in different phases can fall into the trap of deploying a variety of different management platforms or subsystems over time . Good planning and standardisation from the start is therefore critical .
At ICTroom we often talk about multi-level modularity and it can be applied in at least three important dimensions : firstly by ensuring that modularity continues to deliver intrinsic benefits , next with the practicality ( both technical and economic ) of dividing a large , complex development into bite-sized chunks , and lastly by matching as accurately as possible an increasingly diverse range of end-user requirements .
As well as providing flexibility and scalability , modularity can also deliver another data centre essential : efficiency . Whether triggered by environmental concerns or cost reduction , a modular approach is important in not wasting resources including space , energy , water , time , skills and the list goes on . In particular , modularity can ensure that a design PUE ( Power Usage Effectiveness ) is achieved at even low utilisation levels as only the M & E ( Mechanical & Electrical ) equipment that is required to support an actual IT load need be deployed at any given moment . Modules make super-efficient use of space : racks only needing 3kW do not waste infrastructure capable of supporting twice that ; high power racks do not orphan huge areas of white space ; highly secure areas can be used solely for highly secure applications . Aside from the technical benefits , finance is only required when it is needed to support expansion and operations can benefit from high quality , standardised facilities being simpler and more robust .
At a technical and economic level there are practical realities governing the desirable granularity of the sub-parts of any system . For example , within the data centre itself , modular UPS systems might come in blocks of say 20kW or 50kW . If a load is already known to be 250kW then buying a UPS able to carry that will be more cost effective than five 50kW devices . The same is true at the larger infrastructure level . Even though a 1MW facility can be built out of 50kW modules , economics dictate that a better starting point might be 250kW or 500kW sub-modules . Logic therefore suggests that a range of modular form factors , rather than a single model , is needed to address realistic life-time builds .
Bringing all this together , modularity does not mean offthe-shelf . We use an 80:10:10 rule . 80 per cent of any build is standard , using building blocks from globally-significant M & E providers : 20 per cent is custom-built to suit , 10 per cent is dictated by physical constraints , climate or local laws and finally , 10 per cent is use-case specific including special security arrangements , availability of hybrid forms of energy or the need to manage a dark site remotely .
But the most important reason why modular data centres are so critical to the data centre industry concerns their ability to address the myriad of use cases we encounter . From hyper-scale cloud factories to edge collocation ; from high-density university compute platforms to secure pods holding hospital patient records . Just take the example of Microsoft ’ s recent under-sea adventures to see how the form factor of data centres is continually changing .
One size clearly does not fit all . Having an extensive tool kit of modular platforms seems to be the optimum compromise between having to settle for what the local collocation provider decided to build ten years ago or alternatively , building everything from scratch . n
www . networkseuropemagazine . com