In this insightful article, Jürgen Hatheier, International CTO at Ciena, delves into the impending surge in AI training that is set to overshadow cloud traffic. This shift will compel operators to reassess their capacity, automation, and fibre strategies, particularly as they aim to cater to demands exceeding 800 Gb/s and beyond.
For the past two decades, the global broadband network has experienced a consistent increase in traffic. However, the rise of AI workloads, coupled with ambitious initiatives by governments and businesses to excel in the AI realm, is poised to disrupt this pattern. The UK and the US, for instance, are undertaking significant efforts to bolster their compute capacity to stay competitive in the AI landscape.
As AI data centres expand, the implications on data centre interconnect (DCI) are profound. A recent survey by Ciena revealed that AI is driving a transformation in data centre network infrastructure. Experts anticipate that AI workloads will soon surpass those of cloud computing and big data analytics, necessitating substantial investments in data centre facilities dedicated to AI.
To meet the escalating demands of AI compute, data centre operators are investing in scalable, high-capacity infrastructure. The advent of smarter, more adaptive networks will be crucial in efficiently managing the diverse traffic types and dynamic patterns associated with AI workloads. Intelligent automation platforms will play a pivotal role in prioritizing and routing AI traffic effectively.
Managed Optical Fibre Networks (MOFN) are poised to play a key role in supporting the burgeoning demands of data centres as they expand to accommodate AI initiatives. A significant portion of global survey respondents intends to leverage MOFN to scale their networks. While there is no one-size-fits-all approach to optimizing networks for AI workloads, the emphasis on high-performance DCI connectivity remains paramount for cloud and communication service providers.
