Niklas Lindqvist, Nordic Common Supervisor at Onnec, explores why knowledge centre operators want a brand new definition of modularity to adapt to rising calls for pushed by rising expertise like AI.
Rising workloads in AI, cloud computing, and streaming have all created an enormous demand for knowledge centre compute capability. Operators are racing to fulfill this demand, however to take action, they have to determine the way to prolong the life of knowledge centres to be extra versatile.
This requires a brand new type of modularity – knowledge centres which are designed with flexibility and interchangeability from the outset. However solely with a holistic strategy to knowledge centre design and planning can operators create modular knowledge centres that may preserve tempo with altering workload and consumer calls for.
How has the idea of knowledge centre modularity modified lately?
Modularity has historically meant constructing knowledge centres from prefabricated modules or containers that included every part from electronics to cooling. This was meant to hurry up development of the websites.
Nevertheless, these containers restricted operators’ means to scale and get forward of issues that hinder interchangeability, equivalent to dangerous cabling, poor set up, and useful resource shortages. These challenges noticed high-profile hyperscalers, together with Google and Microsoft, transfer away from utilizing containers to scale knowledge centre capability.
As an alternative, operators should embrace a brand new type of modularity, the place knowledge centres are designed with flexibility and adaptableness from the outset. These websites will utilise plug-and-play {hardware} to allow frequent and straightforward upgrades to make sure knowledge centres sustain with the demand for AI. This strategy may also make knowledge centres extra environment friendly, adaptable, and sustainable, serving to firms obtain their ESG targets.
What pressures do operators face which are driving the demand for modular knowledge centres?
Evolving expertise traits, together with AI, the IoT, and edge computing, have introduced surging workloads and challenges to knowledge centre operators. This modification is obvious in rack energy and density traits, as rack energy densities have already risen from 4-5 kW over a decade in the past, to 8-10 kW in 2020. The JLL North America Knowledge Centre Report additionally predicts rack densities might want to improve additional, to 50-100 kW per rack.
Surging demand has created three clear challenges for operators: a speedy improve within the want for AI-related knowledge centre capability, a development slowdown, and the extended operational lifespan of {hardware}. To beat these challenges, operators must conduct speedy {hardware} refreshes extra steadily, particularly to undertake and combine GPU-powered racks.
How are cabling limitations holding again modular knowledge centres?
Defective and dangerous cabling can vastly influence the effectivity of knowledge centres by inflicting poor connectivity, knowledge corruption, and troubleshooting points, which all result in downtime and outages. These issues are attributable to badly deliberate cabling that limits interchangeability. This begins within the design part, the place operators are sometimes focussed on energy and cooling, quite than fascinated about cabling or anticipating future modifications in demand.
Too usually, the cabling spine put in isn’t designed to assist knowledge centres scale. As demand grows, the cabling infrastructure can’t speed up to fulfill increased bandwidth wants. This implies cabling might turn out to be out of date inside three years, as a substitute of a extra acceptable 10-15-year lifetime. Not solely is that this expensive, in comparison with greatest follow set up from the outset, nevertheless it additionally will increase environmental issues by way of the manufacturing of e-waste. For modularity to turn out to be a actuality, right, high quality cabling should turn out to be a way more vital facet of knowledge centre design.
Why is cabling so vital for enabling modularity?
Cabling is vital for modularity – it’s the muse that offers knowledge centres the pliability that’s wanted over their lifespan. When greatest follow is adopted, operators can seamlessly improve {hardware} as wanted to ship AI-compute and cater to evolving workloads. This implies websites can keep aggressive for longer with much less waste and extra effectivity. Ongoing upkeep can also be simpler, and websites might be repurposed as calls for change.
However to undertake greatest follow, a mindset change is required – operators should take into consideration websites as built-in engineering purpose-built initiatives, quite than simply one other industrial property growth. This mindset shift helps operators take into account how the information centre might evolve as workloads change and develop cabling infrastructure that helps interchangeability. A constant strategy to cabling, led to by limiting the variety of cabling suppliers chosen, retains operators centered on the information centre imaginative and prescient. This creates a standardised structure that permits seamless interchange of {hardware} to attain modularity.
How can cabling assist knowledge centres be constructed with modularity at their core?
Knowledge centre design can not occur in a silo. Operators should take a extra holistic strategy to knowledge centre design and planning, and guarantee cabling greatest practices are adopted. This can enable operators to undertake a contemporary strategy to constructing a modular IT infrastructure the place interchangeability and connectivity are foundational.
To create modular knowledge centres, it’s helpful to start by defining the information centre’s strategic path and give attention to business-driven selections up entrance, quite than technical necessities. Groups also needs to take into account how AI may also help the on-site group optimise knowledge centre operations, in addition to the way to make investments the correct amount of capital within the early design phases to allow long run efficiencies quite than short-term price financial savings.
Earlier than operators take into consideration cabling set up, they have to create an in depth cabling venture plan to put the foundations for the infrastructure required, which can minimise set up delays. A meticulous set up information also needs to be created that outlines how cabling ought to be put in, offering a working plan for set up, in addition to detailed directions for seamlessly interchanging {hardware}. Operators should additionally guarantee they’ve the proper expertise within the group to put in, keep, and substitute gear.
For present websites, it may be tough to create an interchangeable infrastructure. A superb start line is making a digital twin of the information centre to grasp infrastructure, cabling, and complexities at play. With the dual, operators can evaluation infrastructure to determine the way to standardise and scale back complexity by introducing shorter cabling and scale back the variety of connections. This additionally permits operators to enhance {hardware} upgrades, scale, and adapt to altering calls for extra successfully and makes technical groups extra environment friendly.
