Edge Computing and the Future of Data

edge computing future

Moving beyond the cloud, the secret to faster data processing lies on the ‘edge’

Cloud computing has seen a big push to centralise both data storage and processing on servers accessed over the internet. And while this approach has certainly been good for business (and we wouldn’t have services like Slack and Dropbox without it), a new model is starting to gather interest, one that at first seems at odds with the cloud concept.

It’s called ‘edge computing’ and it uses a distributed network of highly-local computers to process information rather than draw upon resources provided by an Amazon, Microsoft or Google server. This might sound like a technological step backwards. After all, almost all modern services – from email to Netflix – use the cloud to operate at scale. But edge computing has several key advantages and it promises to revolutionise the way that data is collected and used. Not only that, but it opens up the prospect of entirely new applications and use-cases.

Where edge computing meets fog computing

At its simplest, edge computing uses dedicated computers located at the physical source of data and these computers process that data locally, rather than offloading workloads to the cloud. This allows real-time processing to take place and, coupled with localised Edge AI, instant decisions to be made.

Edge computing works alongside so-called ‘fog computing’, where data processing occurs at the local network level. A typical edge/fog/cloud model, therefore, would see some data collected and used at the edge, giving very fast local results; fog servers processing the next level of data (that could overwhelm the resources of an edge computer); with cloud computing used for high-level business analysis and big data queries.

Why do we need edge computing?

Far from undoing what the cloud has achieved (or replacing it), edge computing is a complementary technology that delivers reduced latency and better application customisation across key market areas such as automotive, enterprise and smart city.

These areas are being transformed by data. In fact, data is often described as the oil of modern business, with analytics providing new insights into current operations and potentially uncovering new opportunities for growth. Thanks to the rise of Internet of Things (IoT) technology, companies have access to more data than ever. In fact, Gartner reports that by 2020 the number of IoT devices (sensors, cameras, etc.) will grow to more than 20 billion.

Edge computing nodes in a factory environment
IoT devices in factories can monitor every aspect of a manufacturing process, with analysis performed on the edge.

With so much potential data available to process, cloud computing systems typically face two main problems: speed and availability.

Speed is a significant issue when dealing with modern data sets. Latency becomes a real problem when you want to send information to the cloud, process it, and receive the results in real-time. Even the fastest internet connection and cloud platforms can introduce a delay.

With edge computing, processing can take place at the source of the data, dramatically reducing response times. This can prove useful for autonomous cars, for example, where real-time processing of situational sensor data is crucial.

Improved reliability and performance

Availability is the second big issue, as cloud computing requires a fast, always-on connection to be of any use, with only limited offline availability. A reliable internet connection isn’t always practical or available, whether due to geographic location or hardware limitations.

Again, one area where edge computing can shine is in the ongoing development of driverless cars. Collecting vast amounts of data from onboard sensors, driverless cars need to be able to make decisions instantly to get their occupants to their destination safely. If an autonomous vehicle needs to rely on an internet connection for its intelligence, it can’t operate reliably.

By processing data locally, cars can react to their immediate surroundings at the speeds required, something a cloud-based system just can’t match.

Working with the cloud

Far from edge systems sitting isolated from the cloud, they can form part of a broader cloud strategy. For starters, edge systems are data collectors, which can filter and optimise information before it’s transmitted to the cloud. This filtering process can save on bandwidth and storage costs, delivering clean and relevant data for later analysis.

Autonomous cars on the streets
Driverless cars can’t progress towards full autonomy without edge computing technology and onboard AI.

Secondly, edge devices can be fine-tuned and operated based on larger datasets, both external and collected by the edge device itself. For example, the International Finance Center Mall in Seoul incorporates digital signage with built-in edge facial recognition systems. These can detect a person’s gender and age, then display customised adverts that are likely to appeal.

Edge computing makes the process of identifying shoppers and showing them ads much faster. Imagine this system powered by a larger central cloud system, using much larger data sets to inform the edge system about which ads to show on any particular day. That’s the power of edge and the cloud when they work together.

Improved security

Security is a big concern when collecting and processing data. Edge computing can add another level of protection into the mix by processing captured data locally and keeping private information from being transmitted to the cloud.

Consider a facial recognition system. With edge computing hardware in place, all data capture can be handled locally. General data, such as a person’s age and gender may be stored in the cloud for subsequent big data analysis, but actual private information remains in the edge system.

By distributing workloads, rather than relying on one centralised repository, security also becomes distributed, cutting down on the availability and access to personal data.

Customisation is the key to unlocking power

For edge computing to be successful, it requires robust, slim and reliable systems that can be installed at the required data collection point or embedded into products. There also needs to be a high level of customisation, so that edge systems can be programmed to act correctly based on data flow, learning and adapting on the fly.

VIA has a growing range of highly-customizable Edge AI solutions powered by the Qualcomm® Snapdragon 820E Embedded Platform, including the VIA SOM-9X20 module, the VIA Edge AI Developer Kit, and the VIA ALTA DS 3 system. Find out how they can work for you here.

VIA Technologies, Inc.