There is a perfect storm coming our way. Data is being generated at a pace we have never seen before. We contribute to 6.9 billion searches on Google every day and WhatsApp users exchange up to 65 billion messages in just a 24 hour period.
Everything is collecting data. Smart cities, smart cars, even smart doorbells. All these devices and sensors require processing and analysis to make them useful. When data sits unused it costs businesses between $9.7 million and $14.2 million annually.
This data explosion continues to provide challenges for the enterprise. Data, artificial intelligence (AI) and machine learning are all rapidly becoming intertwined. AI is fed by data and making sense of data requires AI. Gartner Research predicts that 75% of enterprise-generated data will be ‘created and processed outside a traditional centralised data centre or cloud’ by 2025.
To cope with that level of change, our industry must create the platform to support low latency, dense compute capabilities within edge data centres. These platforms need to offer at least the same server resiliency and serviceability as those in larger scale sites if the expansion at the edge is to be effective.
Finding solutions
A clear solution to edge server environment design is chassis-level precision immersion liquid cooling. There are several variant solutions that address these edge conditions and most offer a sealed chassis which creates a controlled environment that is impervious to dust, gases and humidity. These solutions are also able to maintain data centre compute density while offering improved energy efficiency; this allows high speed, higher processing power servers to be efficiently cooled by liquid compared to the inefficiency of air-cooling systems. Sealed chassis servers also ensure that external environmental factors do not affect the compute capability of the edge system.
Autonomous vehicles are often cited as examples for high performance compute at the edge. This is for a good reason as they are constantly generating data for predictive analytics and search patterns to keep drivers safe on the road. In a split second, the data needs to be filtered, analysed and moved. If you are trying to predict if someone is going to have a car accident, latency becomes a critical issue when moving data back to a centralised data centre. The infrastructure needs to support the speeds and feeds of the data being generated otherwise you can have a very serious problem on your hands.
Consider, as well, the retail environment. Real-time data is used to improve the in-store customer experience. However, the equipment and servers that are needed for that capability have to be in a form factor suited for a retail environment. Floor space is a premium asset and any computing device reducing floor displays or stock room footprint is costing the retailer money. Liquid cooled compute solutions are in form factors identical to air-cooled servers, with the benefit of greatly increased compute density in similar footprints, without the requirement for additional expensive air-cooling systems.
As the move to the IoT and edge computing continues, colocation is becoming an option for organisations who don’t want to manage hundreds of distribution points. However, it is also likely to be a greater point of disruption from AI. Colocation facilities were designed for legacy, traditional, non-compute intensive applications at 5 to 8 KW per rack. If multiple tenants are deploying AI and machine learning applications at 30 kWh per rack, power and cooling limitations within the data centre are quickly maxed out.
Energy efficiency at the edge
The good news is, as an industry, we have developed solutions to address these issues in the data centre itself. Over the last couple of decades, there have been many studies addressing data centre energy consumption. Our industry has made massive moves on energy savings by focusing on best practices for optimising energy and newer technologies to increase capabilities for the same energy use.
The shift to the edge will, however, disrupt these efforts. The economies of scale for infrastructure and solutions in a centralised data centre will not be easily reproduced at the edge, if at all. The question becomes how do you maintain data centre density and improved energy efficiency while bridging the need for ruggedised equipment required for the edge?
Edge locations contend with a variety of harsh IT environments. At one extreme you have the cold and damp in Scotland, on the other, the heat and humidity of India. There are also airborne contaminants, particles, and corrosive gases to be aware of. All of which need to be closely monitored to not impact the servers regardless of their location.
ASHRAE outlines key considerations for the reliable operation of servers and equipment in edge locations. These range from checking IT specifications in order to understand the impact to equipment warranties; servicing capability; corrosion limits; and the impact of air and temperature on equipment. New standards are likely to evolve as we see more deployments in unusual locations from utility towers, light poles or perhaps even in vaults beneath pavements.
‘Edge washing’
Until more solutions are developed, the industry runs the risk of ‘edge washing’. Being edge-ready will need to be about sustainability as much as being operationally resilient. New thinking and truly sustainable solutions will need to be developed and reengineered. It won’t suffice to take a solution developed for inside the data centre, tweak it and then place it at the edge.
Solutions will come to market to test the parameters, many will not be successful as they have not used the right type of electronics, or chips, or didn’t do something as simple as using conformal coating to protect the server boards.
Enterprises are at the centre of an unprecedented data explosion. Data, AI, and machine learning are becoming ubiquitous across multiple industries all over the world. Now more than ever, it is time for enterprises to have an edge transition plan in place. With the right preparation, organisations will be able to capitalise on the real-time insights and create greater value for their business.
