SPECIAL FEATURE cloud computing & virtualisation
There is currently a tendency to design , build and commission cloud data centres as fast as possible .
the past 12 months . Similarly , check for previous outages , the reasons for them and find out what measures are taken to mitigate reoccurrence .
According to a number of surveys , 75 to 80 per cent of unplanned downtime incidents involve human error . Many of these are due to out of date procedures and uncoordinated monitoring systems leading to the loss of control over the four major data centre constraints – space , power , cooling and IT connectivity . This highlights the importance of having an effective operations and maintenance ( O & M ) programme in cloud data centres .
Back in 2009 , a team of Facebook engineers dedicated themselves to tackling a challenge – how to scale the company ’ s data centre infrastructure in the most efficient ways possible . The aim was to design the world ’ s most energy efficient data centre . After designing and building a data centre from the ground up , the result now stands in Prineville , Oregon . It was 38 per cent more energy efficient to build and 24 per cent less expensive to run than the company ’ s previous facilities .
This led to the formation of the Open Compute Project ( OCP ), in which Facebook shares its designs with the wider data centre and IT communities , and this has created a seismic change in the way that cloud data centres are built . Indicative of this is the OCP rack . Traditional IT equipment cannot fit into an OCP rack and vice versa . Rather than having rails that are 19 inches apart and vertical space measured in a U , which is 1.75 inches ( 44.5mm ), an OCP rack maximises the horizontal space by placing the rails 21 inches apart and a rack OpenU is 48mm high . The use of server power supplies is another example . Rather than each server having redundant supplies and each supply being oversized by three times to the nominal power draw , OCP pulls this function out of the servers and utilises a common power supply unit ( PSU ) in the rack , which feeds 12V to servers via a busbar .
For many professionals the OCP has heralded a back to basics approach that is facilitating a move away from vendor lock-in based situations . Too many vendors have tried to lock-in customers , something that can lead to problems . For example , if there ’ s a disaster then there can be massive lead times on equipment . To counter this , some more forward thinking data centre providers have one vendor ’ s equipment at one site and a different vendor ’ s at another . This means that if there is a problem , they don ’ t have all their eggs in one basket . The OCP means that data centres of all sizes should be able to benefit from this type of flexibility .
Standard issue
There has traditionally been a lack of standardisation relating to data centre facilities and infrastructures , something that has been addressed with EN50600 , a new a European reference for all parties . Published by the European Committee for Electrotechnical Standardisation ( CENELEC ), it will help cloud data centre designers configure highly efficient and effective facilities . EN50600 comprises : EN50600-1:2012 . General concepts EN50600-2-1:2014 . Building construction EN50600-2-2:2014 . Power distribution
EN50600-2-3:2014 . Environmental control EN50600-2-4:2015 . Telecommunications cabling EN50600-2-5 2016 . Security systems
Furthermore , the standard has four levels or classes of operation to describe resilience and redundancy in the data centre for facilities and utility design and operation .
While EN50600 will improve how network infrastructures within cloud data centres are configured and the way that different devices operate , there ’ s no substitute for well thought out designs .
Once the cabinets and associated hardware such as structured cabling are in place , ideally they shouldn ’ t change . A good example of the consequences of poor planning is within organisations that extensively used top of rack ( ToR ) simply because it was flavour of the month . In some cases , this led to poor port switch utilisation and this can only be addressed by installing more cabling , with the resulting expense .
It is a fact that cabling is the smallest initial expense in a data centre , but a part of the infrastructure that can save the most and offer excellent return on investment ( ROI ). Therefore , it pays to think about it beforehand and make sure that the needs of the future are accounted for by getting input from various sources using an ‘ ecosystem ’ approach , rather than one based on silos . Unfortunately , in the cloud environment there is a scramble to get these data centres designed , built and commissioned as quickly as possible . Owners get so excited about hardware and electronics that they forget to apply a decent level of thought to the connectivity , the fundamental aspect that can save money .
Where next ?
There ’ s no doubt that the cloud offers a bright future for the development of data centres . While nothing new , software defined networking ( SDN ) and virtualisation are coming of age and shaping up to be massive game changers in this space . The ability to assign network resources through software means a more stable data centre where equipment can be decommissioned from one location and recommissioned into another . Network resources are committed on the fly in a matter of minutes rather than hours and the possibilities here are amazing . It really is a case of watching this space .
30