News in the Channel - issue #45

AI AND DATA CENTRES

CONTINUED

plans look like and where the potential risks lie. “Do they need to keep sensitive data within the UK? Are they planning to introduce AI into time-critical operations? Could their current infrastructure support that growth, or will it become a constraint? “The answers will help determine the right mix of cloud, colocation, private infrastructure and edge computing. These conversations shouldn’t happen just once. Requirements will change as AI develops, workloads grow and new services emerge.” Slawomir adds that customers need support in understanding how their AI ambitions translate into physical infrastructure requirements. “This means assessing existing capabilities, identifying potential constraints and developing roadmaps that allow businesses to scale without making costly decisions or investments in infrastructure that may not meet their evolving requirements,” he says. Ian says partners need to help customers decide where workloads should run, how data should be managed and how the infrastructure can scale securely and economically. “That could mean cloud, on-premises or increasingly a hybrid approach,” he says. “FinOps, intelligent capacity planning, automation and workload optimisation should all form part of that conversation. “Partners should not focus on how much infrastructure is needed first. The emphasis must be on what the business is trying to achieve with AI. Partners can then work backwards from the business outcome and design an architecture that considers performance, cost, security, resilience and sustainability together.” Mark Russell, business unit director at Exclusive Networks, agrees. “Customers are looking for guidance on AI strategy and data readiness,” he says. “The organisations that can translate business outcomes into technology roadmaps

will create far greater value than those focused solely on product sales.”

Contributors

Balancing demands AI and data centres require increasing amounts of power and balancing this with competitiveness is a tightrope that resellers must help customers to tread. Slawomir notes that power availability will remain a significant challenge as AI drives greater demand. “Addressing this requires a holistic approach to infrastructure, with efficiency considered across the entire power train and thermal chain,” he says. “This can help businesses maximise available power, support higher-density deployments and scale their infrastructure while maintaining operational performance and competitiveness.” Dennis Frank, VP, Strategic Partners & Alliances EMEA at Hitachi Vantara, says available power should be treated as something to get more value from. “Then competitiveness and sustainability become the same question: how much useful work can the business do within its energy budget?” he says “Start with the workload. A narrow task might run perfectly well on a smaller model, and some requests can be handled by conventional automation. After that, look across the facility. Better utilisation, efficient storage, the right cooling and retiring redundant equipment can all free up capacity. “Measure total consumption as well as efficiency, because an application can get more efficient per request while its overall energy use climbs as demand grows. Tracking energy and cost against a business measure, such as documents processed or customer enquiries resolved, makes the trade-offs much easier to see.” Ross says that extending the useful life of infrastructure can help organisations reduce unnecessary replacement cycles and make more considered decisions

Slawomir Dziedziula

vertiv.com

Ross Teague

nebulaglobalservices.com

Mark Russell

exclusive-networks.com

Customers are looking for guidance on AI strategy and data readiness. “ ”

46

Powered by