At the infrastructure level, a cloud is simply a lot of data centres interconnected and working together as a giant distributed system. The biggest cloud networks are built by hyperscale providers of cloud and content services with household names: Amazon, Facebook, Google, Microsoft and a few others. Sometimes lumped into a category of Internet Content Providers (ICPs), these fast growing giants share two common infrastructure problems: how to scale up data centres in locations around the world to provide the best customer experience to customers far and wide, and how to interconnect them. These providers and their challenges push the envelope of data centre interconnect technology.
The how of it
So what are they buying? How are they doing DCI? The answer is not so simple. To connect the largest hyperscale data centres, often built in remote areas with low power costs and efficient cooling options, to the metro data centres that are close to customers, ICPs use long-haul optical transport equipment that is similar to the same ones traditional service providers use.
They are increasingly the biggest buyers of new capacity on new subsea fibre-optic systems connecting continents, driving demand for ever higher combinations of capacity plus long distance reach.
These demands prompt new innovations in the sophisticated optical engines at the core of long haul and subsea transport systems, including integrated photonics and extremely powerful digital signal processing electronics implementing advanced algorithms to squeeze as much capacity as possible out of very expensive fibre-optic links.
Easy in the city
In the metro, it’s a different story. ICPs tend to have many smaller data centres in any given metro area and DCI links between them are shorter, typically less than 150km. While ICPs will use a variety of technologies for metro cloud DCI, in most cases and most places the name of the game is efficient scaling: adding DCI capacity quickly and simply with minimum space, power and operational support costs. Optical transport products geared to this market are built more like simple appliances, with attention paid to ease of use and automation.
To ICPs that means several things. It starts with easy installation and provisioning, minimising steps and automating them. When all those pieces are in place, building scalable multi-terabit networks becomes routine and no special optical expertise is required.
The automation word
Automation also extends to other functions required throughout the life of the DCI system. Topology discovery allows DCI systems to be “found” so no manual effort is required to track how the network is connected. Streaming telemetry changes the paradigm of network management so that copious operational data continuously flows from the network to management systems, which track, analyse and store whatever is needed to run the network efficiently. And programmability through open, standard application programming interfaces (APIs) enables software-defined networking (SDN) and data centre automation tools to add capacity or change the behaviour of the DCI network on the fly.
The whole world in your hand
The fact that these metro DCI systems enable simple, automated operations should not be mistaken to imply that the systems themselves are unremarkable. On the contrary, the technology inside is highly sophisticated. At the core of the most-advanced metro DCI systems are optical engines specialised for the task and scaling to incredible capacities.
The common trend across metro, long-haul and subsea DCI networks is that ICPs are the primary drivers of DCI technology innovation.
As these innovations mature, however, everyone else who needs powerful DCI systems that are efficient, secure and extremely simple to use will benefit as well.