Rob Shore
Head of Portfolio Marketing, Optical Networks, Nokia
Rob Shore, Head of Portfolio Marketing, Optical Networks at Nokia, says that the rise of AI presents unique — but solvable — data centre connectivity challenges for hyperscalers, cloud providers and enterprises.
What is behind the extraordinary growth of data centres?
AI is the primary driver. AI requires far more processing and distributed communications — about 10 times more than traditional internet traffic. At the same time, today’s data centres consume so much electricity that individual local power grids often struggle.
Consequently, operators are selecting locations to build new data centres largely based on where power is available. These data centres must then be interconnected with high-speed, low-latency networks to effectively communicate with each other.
How can operators scale these inter-data centre networks sustainably?
Over the last 30 years, optical technology has largely centred on powerful lasers that maximised the amount of information that can be transmitted over a single fibre cable. However, if operators adopt a multi-fibre strategy to interconnect these data centres and spread the data across multiple fibre optic cables, they can use lower-cost, lower-power optics and still keep up with bandwidth demands.
While these optics provide roughly 30% less capacity per fibre, they leverage the latest technology to minimise cost and power per bit of transmission, resulting in a more cost-effective and efficient method of keeping up with the relentless growth in AI traffic.
Power consumption of compute elements
is rising at an unprecedented rate.
What’s the best way to ensure security and compliance as data sovereignty rules tighten?
Standard encryption methodologies have been used for many years to protect data as it travels through fibre optic cables. However, there is concern that quantum computing could crack standard encryption methods. This has given rise to innovations in transmission security called quantum safe networking — new capabilities that increase encryption complexity, secure encryption keys and safeguard data.
Which hurdles still block true ‘zero-touch’ automation in optical networks?
Despite certain advances in the area of agility, today’s optical networks remain largely reactive. To further reduce human effort, networks must become more predictive. This is where AI’s predictive capabilities play a critical role, enabling networks to prepare for shifting data patterns rather than simply attempting to react once they happen.
With greater predictability, operators can reduce the need to overbuild network infrastructure to accommodate peak traffic and shifting traffic patterns, resulting in a more efficient and cost-effective network.
What are the challenges of moving high volumes of data within data centres?
As AI workloads surge, the volume of data moving between server racks and storage elements within data centres is growing exponentially. Moreover, power consumption of compute elements is rising at an unprecedented rate. One of the biggest challenges is how to scale intra and inter-rack connectivity capacity to meet these demands while minimising the amount of power required.
What’s the answer to that problem?
Innovation will continue to be key. For high-speed transmission applications, optical solutions require both front-end optical components that generate and modulate the light and back-end processors to help in converting data from analogue to digital and enabling error-free transmissions at these high data rates.
Today’s optical solutions package the optics and the processors together into a single product. This approach is called a ‘retimed’ optical solution. One key innovation that Nokia is helping to pioneer is to leverage the processor that already exists on the servers or network interface cards to process the optical signals, eliminating the need for a separate dedicated processor on the optics module. This approach is called ‘linear pluggable optics’ and can reduce the power requirements for high-speed transmission solutions — 1.6Tb/s and greater — by as much as 70%.
How can optical options be implemented?
There are two capabilities critical to enabling solution providers to deliver the breadth of optical solutions necessary to cost-effectively address numerous applications associated with data centre connectivity: (1) sufficient scale to both keep up with the massive demands and keep costs down; and (2) a high level of vertical integration, which is required to drive pioneering innovation.
