‘Spine-and-leaf’ may have developed into an increasingly popular term in cabling circles, but there is more behind this fresh architecture that is influencing infrastructure decisions in large modern data centres.
Usual expectations of achieving connectivity using patch cable, MPO/MTP trunks and patch panels collapse when supporting hyperscale, big data, and cloud data centres.
Rather than simply representing the number of ports in a network, the number of connections is a combination of transmission type, redundancy, subscription rates and switch configuration. Here, we are going to explore the main contributors to high fibre count trunk selection and new options and skills available to satisfy the connectivity needs of hyperscale deployments.
An increase in bandwidth demands has fuelled the growth of data centre build outs, which has resulted in the continued compression of time-to-market needs for the data centre production environment.
Recent research revealed that 10GbE switching was dwarfed by 40G and 100G by the world’s leading internet content providers (ICPs), namely Amazon, Facebook, Google and Microsoft.
These companies employ various forms of massively scalable spine-leaf topologies, with varying options able to extend beyond simple spine-leaf to super spines and computer and infrastructure port counts of 100s of thousands in the leaf. Consequently, delivering this capacity as a physical fibre link is no longer an IT administration task.
With different sources contributing to this demand, the need to offer services quickly and improve time-to-market is critical, and so must be the efficient deployment of the infrastructure, enabling quick movement into production for large networks.
There are many cause and effect implications attached to any networking architecture decisions, from the speed and architecture of a network to oversubscription and switch configurations.
In terms of network speeds, a small form factor pluggable (SFP) transceiver with a duplex LC connector interface needs just two fibres to operate and is most commonly available in 10G formats up to 40G. But the quad small form factor pluggable (QSFP) - its more powerful cousin - offers a greater efficiency and flexibility of aggregation and breakout.
Focusing on oversubscription, the oversubscription rate in east-to-west traffic for spine-leaf is by necessity 1:1 with every leaf via the spine needing the ability to negotiate to every other leaf without any delay. Heading north through to the point of delivery generates rates often cited between 3:1 and 2:1, diluting the fibre count requirement for core switching, but introducing the need for fibre cable trunks in the 1000s of fibres due to the sheer number of leaves in a network.
The impact of application, environment, and installation methods required to provide High Fibre Count (HFC) connectivity should be an important consideration as deployments of infrastructure designs with higher fibre counts come into play.
Different MTB solutions present advantages dependent on the application environment and requirements of the design and installation.
Qualities to consider when evaluating these solutions include speed of deployment, product quality and performance, and migration path to 400G and future-ready technology support with parallel optics.
Other considerations could be fibre count, pathway type and ability to plan for or measure cable routes.