Intelligent Data Centres Issue 65 | Page 24

TIMES ARE CHANGING , DRIVEN BY THE POWER AND CRITICAL INFRASTRUCTURE NEEDED FOR SUPPORTING THE GROWTH IN DEMAND FOR IT DELIVERY MODELS AND ENVIRONMENTS .
I N D U S T R Y I N T E L L I G E N C E
loud repatriation

C is a growing trend . Increasingly , many businesses are finding it beneficial to remove workloads from public clouds , evidenced by a slew of market reports over the past year or so , such as Barclays ’ recent survey which found 83 % of enterprise CIOs looking to repatriate some or all of their workloads to on-premises and colocation data centres – almost doubling from 2020 .

The reason for this shift , following the stampede into the public cloud a decade ago , is that many CIOs are finding themselves coping with a complex and often costly cocktail of public cloud environments . This has been exacerbated by issues of interoperability , compliance and governance .
Additionally , the growing focus on AIenabled applications is a disruptor , many of which thrive on vast volumes of data which often perform better and more cost-effectively when compute , storage and data centre infrastructure are under the direct control of enterprises .
Of course , not all have followed the cloud-first trend . IBM Consulting , for example , recently identified among its enterprise clients that 70 % of missioncritical workloads had never been moved to the cloud . This was down to such factors as compliance and data residency requirements , costs and the innovation available in the private cloud sector .
Either way , maintaining workloads onpremises is not without challenges for any enterprise , not least , building and overseeing infrastructure . And for many , the procurement of hardware servers , storage , network equipment and running of data centres has , until now , been the domain of their public cloud providers .
Moving to hybrid cloud
From what we are seeing – as a colocation provider – is many enterprise organisations are becoming much more circumspect on which workloads to keep under their direct control in their data centres and what to retain in the public cloud . The key is ensuring both environments interoperate seamlessly within a ‘ hybrid cloud ’ model .
Hybrid cloud integrates applications that run across potentially multiple clouds , moving data securely across the cloud estate and improve business processes and workflows . This has the potential to simplify and integrate diverse elements of a fragmented cloud estate into a single , unified IT environment .
Modern data centres – owned directly or colocation – offer the benefits of security and cost control , which are also prerequisites for supporting hybrid clouds . For many enterprise organisations faced with everdecreasing IT planning windows and the escalating costs , risks and complexities involved in designing , building and managing their own facilities , colocation will be the preferred choice .
It offers the advantages of renting space and the associated power and infrastructure on a pay-as-you-go basis , funded from OPEX rather than CAPEX .
But this is no time for complacency for the colocation sector and it doesn ’ t necessarily guarantee an easy ride for operators . Times are changing , driven by the power and critical infrastructure needed for supporting the growth in demand for IT delivery models and environments such as hybrid cloud and HPC , and new AI technologies .
There is a growing focus by prospective and existing enterprise customers on levels of resilience , security , cooling , connectivity , as well as space , sustainability , energy efficiency and compliance .
Those select few colocation providers which can expedite the delivery of a hybrid cloud strategy by virtue of direct on-ramp connections to multiple public cloud platforms and connectivity partners , will be in a stronger position by ticking more boxes for CIOs in cloud repatriation mode .
Many may still be wary of lifting and shifting cloud workloads unless they can be assured of the same high-quality , high-speed user experience that they ’ ve enjoyed with Microsoft Azure and AWS , for example . Their respective low latency dedicated networks , ExpressRoute and Direct Connect , are only available as a direct ‘ trunk ’ connection to certain colocation and connectivity operators for ensuring core data centre speeds to largely eliminate latency issues and optimise bandwidth .
Key considerations
Along with high levels of connectedness , data centre power-to-space ratio will be increasingly critical to CIOs . Converged , AI and HPC deployments , for example , are already driving up rack densities ; 15 – 20kW is becoming more the norm , and we are now seeing densities rise to 40 , 50 , even up to 100kWs .
This has called into question a facility ’ s immediate and forward power availability and ability to provide cooling solutions . But it varies considerably , as does available space and levels of M & E redundancy .
Security credentials and the operational regime are also important . However , there ’ s a huge variation in the thoroughness and regularity of relevant testing , planned preventative maintenance and reinvestment . The more comparable a colocation facility appears in this respect to the current or potential public cloud provider , the more

TIMES ARE CHANGING , DRIVEN BY THE POWER AND CRITICAL INFRASTRUCTURE NEEDED FOR SUPPORTING THE GROWTH IN DEMAND FOR IT DELIVERY MODELS AND ENVIRONMENTS .
24 www . intelligentdatacentres . com