Since the Paris climate agreement was signed in 2016, there has been a noticeable shift in the number of companies that are actively pursuing their own sustainability goals and closely monitoring power use. These companies are increasingly coming to us and asking how we can support them in reducing their carbon footprint – something we only expect to increase as new, power-intensive technologies are developed and implemented.
In response to this demand we’ve seen a profound shift in the way data center companies operate as they attempt to manage the global surge in data traffic as efficiently and sustainably as possible. The internet now reaches a staggering amount of people – estimates come in as high as 56% of the world’s population [i]. This rapid growth of internet penetration is being driven by consumers and enterprises alike, demanding real-time connectivity to cloud services and apps via connected devices from various locations.
With the widespread adoption of 5G, the proliferation of connected devices, and download speeds of up to 10Gb/s on the horizon, the levels of data being consumed are set to grow even further. For these huge data sets to move around the world, there needs to be a robust, secure and scalable digital infrastructure that Interconnects dispersed locations, allowing this traffic to be transferred reliably and cost-effectively.
As a global leader in the data center and colocation market, we believe we have a duty to drive innovation and industry best practice. It’s our job to consider what the data center of the future will look like, and to continue innovating to facilitate the secure and sustainable transfer of ever-increasing global data.
World's population who has internet access
The view from above
Fluctuating customer demand and escalating energy costs can make managing infrastructure in real-time for the optimal efficiency of a data center a daunting task. But these complicating factors make the perfect use case for machine learning – an artificial intelligence (AI) approach that enables machines to independently learn, test and apply their knowledge. The knowledge generated can then be used for the benefit of data center optimization with minimum human input.
One of the major issues associated with machine learning is managing the huge amounts of data that need to be analyzed to gather usable optimization insights. To cater for this, we developed IBX SmartView – our own bespoke data center infrastructure management (DCIM), Software-as-a-Service (SaaS) tool that allows us to gather an unrivalled view into the operations of our data centers.
IBX SmartView monitors all the electrical and mechanical infrastructure within the data center and enables the rapid acquisition of good quality data for both our engineering team, and for machine learning. This is beneficial because it enables our data center engineers to comprehensively answer questions around the power consumption of specific utilities within our data centers; and this is an all-important task in this increasingly power-conscious age!
One of the key facets of IBX SmartView is the ability to develop machine learning capabilities, reports and other applications for one site, before being rapidly deployed across other locations. This significantly lowers development time as equipment templates created for one site can be utilized elsewhere.
Almost half of our entire data center footprint is available via IBX SmartView – a figure we expect to accelerate significantly in the future until our entire data center footprint is on the platform. We are already testing the opportunity to replace all the building management systems (BMS) and environment monitoring systems (EMS) within our data centers with IBX SmartView. This will enable us to take advantage of the next generation of machine learning algorithms so that we can continually optimize our platform for efficiency and availability.
Almost half of our entire data center footprint is available via IBX SmartView – a figure we expect to accelerate significantly in the future until our entire data center footprint is on the platform.
One of the key reasons for utilizing machine learning within our data centers is to keep equipment operating at its optimal level, around the clock, in a sustainable and efficient manner. In order to do this, we need to ensure the most mission-critical equipment is constantly monitored and adjusted, so the ambient surroundings of the data center are primed for the best functioning of every component. If we need more information that we can derive from IBX SmartView, we can use a series of battery-operated edge sensors that feed data into the cloud via the Equinix Cloud Exchange Fabric (ECX Fabric), our flagship cloud connectivity platform. Extensive data gathered from a family of machines, for example, can be analyzed and characterized by cloud-based AI algorithms and human experts.
In the initial phase of our power usage effectiveness (PUE) optimization models, we would simulate the IBX and create a custom network architecture with engineered features using a neural network algorithm. This model has since undergone a series of updates. Our latest version now combines asset-based models within a network into a single IBX model. It also allows for end-to-end training of the entire network that simplifies and speeds up model training. The auto-discovery of assets and tags can further be used to autonomously adapt model architecture for each IBX.
As we continue to develop and create more advanced physics models, capable of making assumptions about individual items of our infrastructure such as IT workloads, external ambient conditions and other variables, we can gradually train the model to predict PUE increasingly effectively. Ultimately, when we reach a high level of confidence about the model, we can enable our platform to make real time adjustment to our infrastructure without the need for human interaction at all.
The future is now
As a global leader in the data center and colocation market, we have a responsibility to demonstrate and shape industry best practice. Supported by a concerted effort from some of the other big players, our sector is now increasingly sharing knowledge about sustainable, dynamic and efficient solutions to ensure digital progress is achieved as sustainably and seamlessly as possible.
Human design optimization and machine learning can only achieve so much. We need to think about how we integrate data centers intelligently in cities, looking for opportunities to supply useful waste products such as heat to neighbors who can use it. This will help to reduce financial and environmental costs for all parties involved.
As renewable generation increases, we need to consider how to become better energy neighbors through leveraging next generation power storage technologies and our generation capacity to support the energy grid. It is crucial that data centers of the future are built not only to power the immediate global data demands, but also the sustainability demands of a world that will be driven by digital.
Learn more about how Equinix is championing sustainability in the industry by visiting our interactive sustainability report website.
[i] How Much Data Is Collected Every Minute Of The Day, Forbes, August 2019