What Is Edge Computing?
The Internet of Things (IoT) has arrived, is expanding, and represents the future of some of our most innovative industries. Digital transformation and Industry 4.0 efforts depend on the rapid ingestion of data and the raw cloud computing power necessary to turn vast quantities of data into actionable insights.
Juniper Research projects the number of connected IoT sensors and IoT gateways will grow to more than 50 billion by 2022, from 21 billion devices in 2018. Another analysis from Cisco Systems predicts that cloud connectivity traffic data will have increased from 3.9 zettabytes (ZB) per year in 2015 to 14.1 ZB per year by 2020. That the two are connected should come as no surprise to those who have investigated or invested in an IoT solution.
But figuring out how to aggregate and process all the data these devices collect is one of the biggest challenges for today’s Internet of Things companies. Data is most valuable the instant it’s ingested, and the longer analysis takes, the less value a business can glean from it. And cloud computing, for all its ease of use and power, can also become the bottleneck in IoT and Industrial Internet of Things (IIoT) applications.
IoT technology is undergoing a rapid and expansive change to accommodate these changing needs, and quickly adopting what is now known as edge computing. For a large number of organizations, it will be invaluable to their future growth.
What is edge computing?
At its core, edge computing is the practice of deploying distributed devices that are capable of performing data processing, computation, and decision making across a network. Businesses place these computational devices as close to the “edge” – the physical location within the network where data is collected by IoT devices – as possible, to minimize the space and time between ingestion and analysis of data.
Edge devices come in a variety of forms depending on the application and the computational power necessary. They can be entire offshore wind turbines, industrial controllers, a smart light bulb, security cameras, gas turbines, autonomous vehicles, and much more. Sometimes, they’re as inconspicuous as a single industrial sensor on an oil rig.
They can also be small “data centers” of servers located in strategic areas around a business’ network, or embedded within distributed facilities, such as warehouses or distribution centers, rather than centralized either on-premises at company headquarters or in a cloud computing data center.
Regardless of the size, these IoT edge devices take on the responsibility of processing data themselves rather than relying on a centralized cloud computing service. They can perform analytics, make decisions, and even run artificial intelligence or deep learning models on the data, without additional support. An autonomous car is an excellent example of an edge architecture.
An autonomous vehicle relies heavily on sensors that monitor the vehicle’s current state and observe its surroundings. These sensors are connected, through the vehicle’s operating system, to essential mechanical functions like steering and braking. In order to respond rapidly to hazards, shifting lanes, other vehicles, pedestrians, cyclists, and stop signs, the vehicle needs to ingest data from sensors, analyze that against the current speed and other state data, make a decision of its next move, and apply that decision in a fraction of a second.
The raw power of edge devices is one of the primary reasons autonomous vehicles exist today. If the vehicle needed to send sensor data back to a cloud computing data center for analysis, it would have crashed by the time a decision was transferred back. By placing computational power in the vehicle itself, and thus the fleet’s physical edge, an entirely new industry is born.
Consumers also see edge computing in action in voice-recognition hardware, smart thermostats, internet-connected TVs, and even the smartphone in their pocket. At the same time, industrials are pushing edge infrastructure and edge analytics to empower their digital innovation and create smart factories using truly intelligent machines.
The 2018 Juniper Research report also predicts edge computing will be the primary engine behind the growth in IoT. As more IoT providers offer edge technology as part of their existing platforms, and more businesses see the benefit of smaller bandwidth requirements and dramatically reduced latency, deployments will begin to scale in size and into areas previously thought inhospitable to computing power.
By applying computational power, artificial intelligence, machine learning, and advanced analytics at the machine’s location, businesses can tap into yet-undiscovered value from their machine data.
Benefits of edge computing
- Dramatically reduced latency. Data transfer between an IoT device and a cloud computing infrastructure adds latency and reduces value, and isn’t fast enough for applications that require instant action, such as autonomous vehicles or industrial applications in a closed-loop system where insights from machine data directly affect its following actions.
- Value gained instantly from data. Data is most valuable at precisely the moment it is created, and its value diminishes with every passing second. Through reduced latency – the time gap between the acquisition and analysis of data – businesses can learn more from the same data and drive disruptive innovation.
- Streamlined analysis process. Traditional IoT+cloud infrastructures transfer all data from dozens, hundreds, or thousands of devices to and from a centralized computing environment. But, if edge devices are capable of making critical decisions locally, they can choose when and how to push that data to the cloud – if it’s necessary at all.
- Bandwidth reduction. Because high volumes of bandwidth incur a heavy cost for businesses, any cut can make a meaningful impact on the bottom line. Edge computing presents an agile opportunity to keep data close to home whenever possible.
- No one single point of failure. A centralized cloud computing environment, despite all its benefits, creates a single point of failure for making mission-critical decisions on IoT machine data. Edge computing spreads out the risk – an edge platform can still operate and make key analysis even if the central cloud computing experiences downtime.
- On-device encryption and security. By reducing the need for data in transfer and through on-device data encryption, edge computing has theoretical security benefits to industrial applications requiring the tightest security measures. By minimizing reliance on a single cloud environment to store all company data, there is less vulnerability if attackers compromise that environment.
- Operations maintained where data connectivity is difficult. Remote or challenging locations, such as oil rigs, often struggle to maintain strong network connectivity. Edge computing can allow these rough operations to function self-sufficiently using onboard decision-making capabilities until the network regains connection.
- Onboard data streaming more easily. Instant analysis of streaming data plays an important role in the future of smart manufacturing and challenges around IT/OT convergence across industries. Edge computing can open the door to complex analysis on high-velocity stream data without a significant investment in a centralized computing environment.
Will edge computing overtake cloud computing?
Cloud computing immediately presented an opportunity for businesses to reduce up-front IT costs, scale as needed, and manage infrastructure more efficiently. Cloud computing providers have consistently improved both the performance of their product and its ease-of-use, which has resulted in enormous innovation and economic growth. Today, it’s possible for even the smallest of businesses to leverage the power of a supercomputer-strength data center.
Edge computing presents a paradigm shift away from computing architectures of the past, but there is no indication it will displace cloud computing environments. In fact, businesses are gaining the most value out of bridging their cloud and edge environments together and utilizing each to their strengths.
Cloud computing will continue to dominate in applications that require massive computational power or managing large quantities of data, such as deep learning training. That said, edge computing is more than powerful enough for even complex AI applications. An autonomous vehicle’s training may happen on the cloud, but its instantaneous decisions happen with onboard edge architectures.
How is edge computing being used today?
Edge computing has already brought its low latency and high flexibility capabilities to a variety of industries:
Defense: Soldiers work in some of the most rugged environments imaginable, and require the ability to make split-second, informed decisions. Edge sensors can analyze the communication spectrums being used on the battlefield and then make real-time decisions to shift operations in the event of interference or jamming operations from the enemy.
Energy: An offshore oil rig might only send sensor data back to the cloud computing environment twice a day, and through a slow satellite connection. Edge devices eliminate latency and allow the most urgent decisions – such as shutting down operations due to a safety breach – to happen automatically, while also leveraging the raw power of the cloud to make wider analysis of efficiency and possible optimizations.
Transportation: Aside from autonomous vehicles, the transportation industry is embedding trains with edge platforms to help analyze the billions of data points per second that are already being collected by dozens of onboard sensors. Edge computing is being used to monitor the health of locomotive equipment in real time, and the industry is working toward autonomous operations applications, as well. For improved operational efficiencies, passenger recognition systems are being used to speed boarding for rail and air travel alike.
Healthcare: A wearable medical device, such as a wristwatch that monitors seizure activity, can now do more than collect motion and heart rate data. Using onboard computing power, the edge device can detect epilepsy episodes and notify caregivers or emergency services without the latency of cloud computing and without the risk of being momentarily disconnected from the network.
Smart manufacturing: Smart manufacturing facilities can improve safety, efficiency, and quality all while reducing costs through automated processes driven by decisions made on edge computing devices. Edge IoT, feeding off of machine-to-machine data, can detect when a machine is about to fail and automatically apply contingency plans, such as diverting product to alternative lines or notifying stakeholders.
Retail: Infrared sensors create physical heatmaps of where retail customers gather in the store and how they move about, which gives retailers up-to-the-minute insights on performance. They can optimize store layouts, ensure associates are in the right places, and deploy promotions with guaranteed results.
Smart cities: As municipalities upgrade their aging gas or water infrastructures, they can install edge devices capable of monitoring for faults or anomalies. An intelligent water meter, for example, could detect a water main leak and reduce or shut off supply to minimize the impact before services can respond.
Environmental monitoring: Warning the public about dangers from volcanoes and earthquakes is difficult because of the speed at which it happens, and the general remoteness of the locations where they’re most common. Seismic, air pressure, and temperature sensors, when placed in rugged conditions at the edge, can not only detect anomalies internally, but also know when it’s time to inform public safety organizations.