5 ways AI and IoT will change the data centre and the edge as you know it

5 ways AI and IoT will change the data centre and the edge as you know it

The world is heading for big. Not just big data, but big transformation which will introduce new ways of designing and running data centres. In Dublin, the industry heard from LinkedIn’s Zaid Ali Kahn on how data centres will transform. João Marques Lima reports.

The Internet of Things, which is today connecting more than 20 billion devices and is expected to connect as many as 70 billion by 2020, according to the highest estimates, will drive a challenging need to build IT infrastructure capable of handling the 24×7 ongoing demand for applications such as video, social media, enterprise business, financial systems, and so on.

However, the data centre industry can sometimes be reluctant to change, especially with the introduction of technologies which, if faulty, could cut off revenue streams.

Nevertheless, although reluctance might exist, the reality is that today, most outages are caused by humans, not technology.

The big transformation heading towards data centres is not only targeting the death of ‘lights out scenario’ but improving computing efficiency overall.

It was based on that that Zaid Ali Kahn, Senior Director, Infrastructure Engineering, at LinkedIn, took to the stage at the inaugural Datacloud Ireland congress in Dublin.

Addressing an audience of circa 300 people, Kahn delved into what the company sees as the future of data centre design and what technological breakthroughs will transform server farms.

“The future is going to be driven by two things: AI and IoT,” he said. “Those will be two key drivers on how we will build our data centres moving forward.

“We really have to think differently looking five to ten years from now.”

Here’s the top five design considerations for data centres of the future.


LinkedIn’s Zaid Ali Kahn

Coming at number one, Kahn spoke of decoupling software and hardware in order to open up data centres to innovation.

Sometimes, data centre architecture is built in silos, which can hinder innovation and the development of new hardware or software technologies.

The future cannot be built by one person, and therefore data centre operators and investors need to open up to a new reality that is today the digital sector.



Secondly, as the IoT expands and computing power booms with highintensive AI-hungry applications, that is set to radically force the way data centres cool down their infrastructure.

Kahn said: “Component level cooling inside racks, such as GPUs and CPUs, will become very mainstream.

“[AI] changes the way we do computing. It definitely changes the way we do computing in the data centre and how we work with the rack.”

He continued to say that the new rack operations will generate higher amounts of heat which will require new stronger cooling techniques, which also take into account the take up in power usage.

“[The new rack] is going to generate a lot of heat. These things will consume a lot of power, more power means a new way of cooling. You cannot rely on your ongoing cooling.

“I predict liquid cooling is really going to take off. It is no longer about ‘I am going to go to the coldest place on theplanet and use that’, this is really going to change.

“We brought cooling to the rack at LinkedIn. You have to think about bringing cooling right down to the CPU level.”



Eventually, demand will force data centre  technologies to be re-thought and redesigned. “We will have to think how to be more efficient, how you get things to integrate faster,” Kahn said.

He firstly took on storage, saying the concept will have to “significantly change” to accommodate what is coming towards the data centre. In the hyperscaler world, Kahn said storage is built much flatter. “Storage has to significantly change.

It is moving to fabric. Before you had a big machine, stored in a rack. But today it is all going to be distributed.

“You really have to distribute it; storage is going to change the data centre, especially as we move to the edge.” With edge computing becoming a growing focus to many of those operating in the data centre markets, Kahn labelled edge as a key driver within the wider IoT ecosystem.

“You have to have very low latency to get all the data from a nearest point, such as a cell tower [when looking at driverless cars, for example]. “How you look at the edge data centre is something people will have to think about. And there’s not enough power to power this so people will have to think about this. In next five to ten years, we will see massive change of edge computing.”



Up next is self-healing infrastructure. These sort of deployments are not completely new and have been used in other environments such as industrial. A self-healing environment can, as the name suggests, prevent any sort of business disruption at the hardware and software layers.

This could take many shapes and forms such as software that can write or rewrite itself to prevent, amend or stop any faults, or hardware that can be replaced without human labour.

Nanotechnology will be one to watch here, and the development of materials that can reproduce themselves in case of physical damage. Self-healing infrastructure will not only bring CAPEX and OPEX down, but will also push the five nines availability to potentially 100% or at least more nines.

Kahn said: “The data centre we will want to live in the future will be selfhealing. At LinkedIn, we are already looking at this. “Networks and systems should be able to heal themselves, we cannot be going around and do this ourselves [as humans] for ever.”



Lastly, just like what has been happening in manufacturing plants and industrial sites all over the world for some years, data centres will open their doors to robots.

The need to manually install a rack could be just a few years away from becoming an obsolete task, as robots will navigate the data centre and carry out the works. And this is not just down to the rack or cabinet. Robots will be able to work cables and other components of the centre.

At the edge, we could see these robots taking the shape of driverless cars or drones, that would take the needed equipment to an edge site and repair any issues – with potentially the help of a second robot – minimising the need for human interference with the deployment.

Kahn said: “Cables, for example, will be fixed by robots, moving cables around will be a robotic task, and we are not far from that. I think in about five years we will get there.”

Follow by Email