A Practical Guide to Artificial Intelligence for the Data Center

David Hall
A Practical Guide to Artificial Intelligence for the Data Center

The modern world continues to digitize at unprecedented rates. Cisco predicts that by 2021, every person will have 3.5 connected devices generating 35 gigabytes of traffic per month. Business traffic is also expected to grow to more than 45 exabytes per month in the same time period.[i] To adapt to this exploding demand, organizations are adopting new approaches such as moving workloads to the cloud and tapping into ecosystems of partners to deliver faster and better digital experiences. This proliferation of interconnected services is changing the IT landscape – driving demand for more datacenters, in more places, with greater and greater levels of efficiency and availability.

Data center efficiency is about optimizing the physical world equipment to match the digital demand with the highest degree of accuracy possible. That means having the right number of servers, power, cooling needed to support demand in that location at that time. Predicting that demand across a distributed infrastructure will be far more challenging as the number of variables that impact it increase. Differing climates, changing weather patterns, natural disasters or demand spikes for events like Cyber Monday make it difficult to plan ahead for optimal efficiency. It’s an ideal use case for machine learning (ML), an AI approach that enables a machine to learn and self-improve over time without human input. To understand how that works, we need to understand how this process works in the human brain.

The power of patterns

Have you ever wondered why almost anyone can play “Name that Tune”? Or how easy it is to detect a wrong or out-of-tune note?

Research studies have shown that humans are capable of sophisticated music processing, even without any explicit musical training. For example, we can easily recognize the “Happy Birthday” tune whether it’s played on a piano or tuba, or even sung in a different language. Through exposure to music in everyday life, people acquire knowledge about the musical system of their culture.[ii] It might even be said that music is the soul of culture, representing the rhythm of people’s lives – a universal communication of stories, history, traditions and emotion.

Why is it so easy for us to recognize which region of the world a piece of music comes from, or whether a song is happy or sad? It turns out the answer may lie in the evolution of our brain. Generally speaking, the brains of all animals are designed to process inputs from the outside world and generate outputs in the form of adaptive responses such as seeking food or running from danger. Sensory information is rapidly encoded as patterns which can then be recalled later in ways that enable comparisons of different patterns.[iii] In short, survival by pattern recognition.

In humans, this capacity for pattern recognition and processing became increasingly sophisticated, forming the basis of most, if not all, of the unique features of the human brain including intelligence, imagination, invention, belief systems.iii Our brains, in fact, are highly tuned pattern searching machines; and much of what we think of as ‘intelligence’ is, in fact, an expression of our ability to recognise complex patterns in the endless streams of data that we get from our senses.

Pattern recognition in neural circuits

The human brain has as many as 100 trillion synapses that form patterns when activated. When a person thinks about a specific thing, remembers something, or experiences something with one of their senses, it’s thought that specific neural patterns “light up” inside the brain. When you first learn or experience something, your brain has to figure out how to respond. Once you’ve learned the response, you begin to access a part of the brain more associated with memory than problem-solving. Thus a different set of synapses fire because you’ve trained your biological neural network what the response is.[iv]

Synaptic cleft, or synapse, in the human brain. Source: TheissCare v


Going back to the music example, we can see how the auditory cortex of the brain is responding to tones played at different frequencies. This provides an interesting clue about what’s going on in the brain – our response to patterns (or inputs) starts with basic feature analysis (or pattern recognition). These individual responses give way to the “lighting up” of the brain when a person listens to a whole piece of music such as Beethoven’s 5th. That’s because clusters of neurons responsible for feature analysis or pattern recognition begin to activate intermediate groups of processing neurons- before finally activating the parts of the brain that deal with conscious thought.

Source: Magnetic Resonance Imaging vi


Artificial neural networks

So how does this work with machine learning? An artificial neural network is represented by a series of layers that work like a living brain’s synapses. In a neuron, the electrical signal travels from one neuron to the other through the synapse. If the electrical impulse is of a certain strength, the synapse fires, sending the signal to the next neuron. In the simplified “engineer’s neuron” below, called a peceptron, there is a node (y) with inputs (x1, x2, x3) going into it. Each input has a weight (w1, w2, w3). The node y takes each input and multiplies its value by its weight, sending the signal on to the next node, depending on the weighting.


Source: Towards Data Sciencevii


Groups of neurons interconnected together to perform a specific function is a neural circuit. A neural network operates on a similar principle with clusters of interconnected nodes. The figure below shows how a neural network works for recognizing the image of a dog.


Source: Quanta Magazine[v]iii


AI in the data center

Just like the human sensory system, Equinix data centers generate large volumes of data on a continuous basis. Equinix monitors everything from external weather conditions, the performance of our infrastructure, and of course, customer demand. By taking inspiration from the pattern matching abilities of the brain, Equinix, in collaboration with key industry partners, is augmenting Platform Equinix with intelligence to help us make better decisions.

In Part 2 of this series, I’ll cover some basics on how neural networks are trained, as well as specific AI applications that are fueling data center innovations such as improved power usage efficiencies, higher availability and reduced faults. We’ll also go behind the scenes to see how AI is being used in our Equinix International Business Exchange™ (IBX®) data centers; and share some insight into our journey to build a truly intelligent data center.

To learn more, check out our Platform Equinix Vision paper.

You also may want to read these blog posts on artificial intelligence:

[i] Cisco, The Zettabyte Era: Trends and Analysis, June 2017, Doc ID 1465272001812119; Cisco Visual Networking Index: Forecast and Methodology, 2016-2021, June 2017

[ii] Handbook of Clinical Neurology, The Human Auditory System, Vol 129, 2015 and Comparative Music Cognition, The Psychology of Music (Third Edition), 2013 (as cited in Science Direct, Music Cognition).

[iii] US National Library of Medicine National Institutes of Health, Superior pattern processing is the essence of the evolved human brain, Frontiers in Neuroscience, Aug 2014.

[iv] The Next Web (TNW), A beginner’s guide to AI: Neural networks, July 2018.

v TheissCare, The Sodium Potassium Pump – Nutrition and the Brain, July 2018.

vi Magnetic Resonance Imaging, Development of sound measurement systems for auditory functional magnetic resonance imaging, 26. 715-20. 10.1016/j.mri.2008.01.020, July 2008.

vii Towards Data Science, A Beginner’s Guide to Neural Networks (Deep Learning), September 2018.

[v]iii Quanta Magazine, New Theory Cracks Open the Black Box of Deep Learning, September 2017.


Avatar photo
David Hall Former Fellow focused on Technology and Architecture in the Office of the CTO
Subscribe to the Equinix Blog