“Edge computing,” like most new IT developments, isn’t a revolution; rather, it’s more of an evolution. Edge computing’s roots are in the content delivery and peer-to-peer networks of the early millennium as well as in grid computing. Yet a combination of improved technological capabilities in networking, computing and analytics coupled with demand based on huge projected data increases means that where computing takes place will be more and more critical for IT managers.
As the anticipation of massive amounts of data being sent to and through networks grows, firms are developing computing capabilities closer to the network edge, where data is generated. The disruption that edge computing creates allows local users to produce and perform analytics with data in real time. But despite the increasing momentum behind this trend, the jury is still out regarding precisely how and when edge computing will be deployed. Will edge technology drive industry business decisions over the next few years, or will its wider deployment require further advancement and holistic long-term planning? Which use cases will drive implementation and what challenges remain?
What Is Edge Computing?
For the sake of definition, the broad term edge computing seems to have originated nearly 20 years ago when edge servers was a term referring to servers in content-delivery networks (CDNs). It has more recently appeared in the context of processing, analyzing and applying knowledge from data produced by sources at the edge of the network, versus transmitting that data to a “core” processing unit.
The term edge is based on the proximity of processing to the data source as well as where analytics take place, so it covers a range of possible uses. Volumes of data from Internet of Things (IoT) devices, mobile devices, and networks with different processing requirements and priorities are and will continue to be generated. This shift will emphasize the importance of where computing takes place, requiring smaller and more-agile processing units closer to the user. It may take the form of distributed micro data centers with local network access and interconnection points, essentially forming a distributed cloud.
What’s Pushing Us to the Edge?
The overarching force behind the impending data-transmission growth is a result of increases in computer-to-computer communications and IoT devices. Owing to a wide variety of sensors and processors creating and transmitting massive amounts of information—as well as greater investment in and development of artificial and augmented reality, drones, and automated transportation systems—the push to the edge is coming from a number of sources.
The 2015 Cisco Cloud Index found that about 90 percent of the data created globally was generated in the two years preceding that company’s report. The index also forecasted that the amount of Internet Protocol (IP) traffic per month would grow at a compound annual growth rate (CAGR) of more than 100 percent between 2014 and 2019. At the same time, the number of Internet users is projected to increase at a 7 percent CAGR, and the number of connected devices will grow faster, reaching an 11.4 percent CAGR.
Video-data transmission was projected to increase by 80 percent in the same five-year span. To put these numbers in perspective, an estimated 3.4 billion people were on the Internet as of 2016. YouTube users downloaded 400 hours of new video every day, Instagram users liked 2.5 million posts every minute and Facebook users shared 3 million posts per minute and liked more than 4 million posts per minute. Additionally, about 4 million Google searches are conducted every minute of every day, over 200 million emails are sent every minute, over 400,000 Apple apps are downloaded and 277,000 tweets are sent. In a financial context, Amazon sells an estimated $80,000 of merchandise and services—every minute of every day!
This combination of consumption and technological factors creates a more complex set of drivers that vary widely by industry and geography. Generally, the three main drivers pushing us towards edge networks are the following:
- Changing consumer and business expectations, as well as data usage
- Emerging technologies, particularly in networking, processing, software and protocol areas that make edge computing possible
- Applications of edge processing, such as the opportunity to integrate IoT device data, greater efficiency in processing and transmission across the network, lower latency, the delivery of a better customer experience, and data security
What Are the Business Uses for Edge Processing?
Like any IT trend, business and technology professionals want to know how advances can help them optimize their operations. Edge computing, however, may not be for every business—at least for now. Adoption of edge computing and applications will ultimately be determined by how well these technologies mesh with business goals and will depend on whether an organization has the resources to effectively implement, manage and monetize them.
Several industries are particularly well poised to benefit from edge computing:
- Smart cities. Edge computing can apply broadly in a smart community or city. As the number of sensors and sources grow (citizens, traffic systems, health-care systems, utilities and security programs), storing and analyzing data in a central location becomes less feasible. Edge computing also reduces latency delays in community services where action must be quick, such as in medical emergencies, law enforcement, traffic patterns and public transportation. It also allows for geographic precision, so information relevant to a particular street, block or suburb can be shared instantaneously with users in that area. The applications and technology will ultimately determine whether the edge extends from traffic sensors and streetlights to pumps, turbines and other traditionally unconnected utility devices. How could smart-city edge networks collect and distribute information in the event of natural disasters? How could supply-chain impacts for resources such as water and gasoline be communicated and mitigated using smart-city and edge technologies?
- Smart commercial and public transportation. Edge computing already performs a number of functions for commercial and public transportation. For complex vehicles such as airplanes, ships and spacecraft, accelerated-processing requirements and compute/analytics at the edge mean that only the most vital information is transmitted for further analysis; most is locally stored. Edge technology allows traffic and environmental sensors to process and deliver the most relevant information to vehicles, including self-driving cars. The second function of this edge network is informational: feeding local data patterns into larger network systems that are more widely accessible to increase travel efficiency and safety. Smart transit systems are also a natural part of smart-city development.
- Smart homes. A number of data center OEMs claim that every home in the U.S. will soon become its own data center, and that claim is coming closer to reality. Edge computing, however, links smart-home systems back to the core production center rather than creating an independent data center as data moves toward the edge. The role of “batch and send” versus real-time connected devices in smart homes continues to discussed and developed.
- Automated vehicles, drones and remotely operated machinery. Perhaps the most well-known example of edge technology, a self-driving car will require an estimated 200 or more CPUs and is a “data center on wheels” according to Peter Levine of Andreessen Horowitz. Self-driving cars can process live video and stream photos to make immediate decisions based on data input. They highlight the need to share collaborative information through a smart transportation network. This concept can extend to drones for agriculture, mining, oil and gas and other industries that must react in real time to the data they collect.
- Media and other content. CDNs are already bringing content closer to the user, and edge computing is a logical next step to deliver additional operational applications to customers. They’ll also play a part in future content delivery, enabling providers to expand geographical reach and maximize the efficiency of delivery networks, particularly as they introduce more value-added and interactive services.
- Manufacturing and Industry 4.0. Robotics, artificial intelligence and machine learning have seen adoption by many industrial organizations, all of which are optimal uses for edge computing. A guiding principle for edge computing in manufacturing is to streamline production into a standard process from demand to production, delivery and consumption. This effort requires the exact type of collaboration between data sources across a range of locations that edge computing provides. Industrial IoT (IIoT) continues to increase efficiency and reduce cost with predictive maintenance, improved safety and other operational efficiencies. The edge is essentially a data supply chain for Industry 4.0.
Developing an Edge Strategy
Some businesses have publicly adopted an “all-in” approach to IoT and the edge, though they should remain mindful that the rush to the cloud is still occurring for slow-adopter industries. Trends away from centralization and the cloud to distributed edge computing require careful consideration in order to determine the optimal balance between aggregation and disaggregation for each use case.
For businesses adopting the edge, the five main components of developing a comprehensive strategy are as follows:
- Define objectives and requirements including business goals/drivers as well as brand, customer and return-on-investment requirements.
- Map the network topology from the edge back to the core.
- Define the systems, protocols and programs that constitute the edge processing, abstraction and communications capabilities.
- Define the networks that link edge-processing units to data sources and back to core processing facilities.
- Develop strategies for supervision, maintenance and security of edge-computing systems.
Edge Operational Considerations
Building an edge computing capability is only the first step in implementing this technology. Long-term success will rely on comprehensive development of operational, security and maintenance requirements, as well as integration with IT and telecommunications providers. As demand for edge computing grows, roles must be clearly defined for all parties and data prioritized across the network. Despite the clear benefits of edge computing, a number of considerations should be taken into account.
Security and privacy are major challenges that need addressing. How information is generated and processed—and its ownership—must be defined. The concept of criminal or civil responsibility could get complicated. For example, when a self-driving-car accident occurs because of a programming flaw, who will be held liable? What impacts will data-protection requirements have? Will the edge innovate, imitate or regulate?
Analysts also wonder when the growing number of data-producing devices will create a breaking point and overload networks. Software-defined networking, 5G technology and edge-to-edge communications will continue to evolve. Interference and bandwidth requirements in the radio environment will place tough demands on transport networks. Traditional macro networks will need complementing through the addition of small cells.
Resilience also must be considered when discussing the edge. It raises a variety of questions: What’s the impact of the failure of a unit, source or cell in an edge-processing unit? How will it affect connected units and shared programs? How can a system entirely dependent on a fluid network protect itself against cyberattack? How will capacity and latency coexist, and how will network priority be determined? The answer appears to be that resiliency will be determined on the basis of what’s being delivered, where it’s regulated, the end-user experience and cost/revenue metrics.
As edge computing evolves, the shared language that enables multiple-edge systems and a variety of platforms and run times will also increase and need advancement. Protocols that enable the systems and the network to work together will require development. Industry leaders are already in the process of developing IoT gateways and routers that can support edge computing, while software such as Apache Spark is taking the industry by storm by clustering and running a variety of modes while writing to disk when it needs to. Where is the tipping point when this combination of smaller milestones will have made a big difference?
Understanding and preparing for edge computing will need to take many considerations into account, including mobility and security. Holistic planning can ensure that system requirements are met and that the system can identify and communicate with devices and users on the edge. The conversation has only just begun.
About the Author
Rob Nash-Boulden is a thought leader in cloud and edge computing. He is a director in the data center group at Black & Veatch, a global engineering, consulting and construction company that provides specialized data center solutions.