The power of intelligence at the edge

By Daniel Quant

Vice President of Strategic Development

Multi-Tech Systems

October 20, 2017


The power of intelligence at the edge

To support the exponential growth in data collection, IoT networks of the future must adopt a decentralized architecture in which processing of raw data begins not in the cloud, but at the edge.

The Internet of Things (IoT) has set the IT world on fire with innovative technologies and astonishing growth levels. Forbes forecasts that the boom will continue, with the IoT market for products and services predicted to account for sales of $267 billion in 2020. The rate of data collection has also accelerated in recent years: an IBM Marketing Cloud study claims 90 percent of total data in the world was created in just the past two years at a rate of 2.5 quintillion bytes of data per day.

To support this exponential growth in data collection, new smart connected sensors and assets are being deployed every day. This, in turn, calls for innovations in the machine-to-machine (M2M) and IoT networks that carry vast data flows in order to mitigate the cost and inefficiency associated with handling data centrally in the cloud. IoT networks of the future must adopt a decentralized architecture in which processing of raw data begins not in the cloud, but at the network edge.

Opportunities for intelligence at the edge

Enterprises now recognize that the ever-increasing flow of data must be managed more efficiently in order to optimize information utilization, reduce costs, and improve business performance.

Today, many data management systems use the cloud as an all-purpose dump for every type of raw signal or data output generated by local sensors and assets. In this type of highly centralized data management architecture, huge amounts of laborious and costly data processing have to be performed centrally, either in the cloud or in an enterprise data center. This approach has various drawbacks:

  • Cost of backhauling data between edge nodes and the cloud or server – Traffic charges are levied at a cost per byte. These charges can mount rapidly, particularly when using cellular networks.
  • Cost of cloud services for processing raw field-bus data – This data is often from proprietary industrial assets, and must be converted into a format that can be handled more efficiently by Internet-based platforms
  • Cost of data storage – Storage charges are typically levied by cloud service providers at a cost per byte of data storage capacity. The more raw data is accumulated in the cloud, the higher the storage cost.
  • Latency introduced into business processes – It takes time to upload data to the cloud, process the data and produce a decision, and then transmit the decision back to the edge node. Simple decisions (such as whether to accept delivery of a shipment of frozen food or whether to perform a truck roll to deliver or inspect an asset) can often be made much more quickly if the processing is performed locally.
  • Risk of downtime – Downtime can disable or impair a local business process. Every decision or business process that relies on an application running in the cloud is vulnerable to a failure anywhere in the link between the edge and the cloud server. Internet connections are prone to downtime. If time-sensitive decisions (such as whether to accept that shipment of frozen food) do not occur because of a network failure, the economic cost can be considerable.

In reality, much of the raw data sent upstream has little or no value. For instance, small changes in temperature inside a cargo truck carrying perishable goods do not require any decision or action, so the truck’s logistics application does not need data about them. The application just needs to know when pre-set maximum and minimum temperature thresholds have been breached. Therefore it makes sense to capture and sift such information at the network edge and make decisions locally, securely transferring only actionable data and exceptions to the cloud.

Opportunities for intelligent data management

Implementing such a decentralized architecture requires intelligent data management systems at the edge of the network. Efficient data management at the edge of the network eliminates the need for time-consuming data backhauls, as well as the parsing required to prepare data for business decisions.

This raises the question of how to deliver a hardware and software platform enables intelligence at the edge of the IoT while maintaining built-in support for cloud connectivity, services, and applications.

One approach to this question, supported by products from Multi-Tech Systems, is to build intelligence into connectivity components at the edge. Gateways and embedded modems, when backed by programmable real-time operating systems (RTOSs) or application development toolchains such as ARM’s Mbed and IBM’s NodeRED, can:

  • Implement and deploy applications locally
  • Support rapid and intuitive application development in environments familiar to IT developers
  • Provide a complete suite of software-to-cloud connectivity that includes security and authentication capabilities, standards-based cloud middleware, and telecoms protocols required to scale deployments

Embedded application development for IT professionals

In the world of electronics engineering, the concept of local processing and intelligence at the edge is in fact nothing new. The microcontroller (MCU), the core component in the majority of embedded devices, has been steadily growing in capability over the past three decades while falling dramatically in price. Today's MCUs are capable of complex signal processing, logic processing, and graphics processing tasks, as well as handling multiple communications interfaces and hardware peripherals.

MCUs with powerful ARM Cortex-M processors cost as little as $1.50, but provide the ability to execute rules-based algorithms and are able to run sophisticated OSs like ARM Mbed OS. This in addition to all the application code required for processing sensor data at the edge and transmitting it to the cloud.

If the vision of an IoT containing billions of connected things is to be realized, however, it cannot be reliant on the relatively small pool of electronics design engineers who have lots of experience with M2M communications and telematics, and who are comfortable developing applications in a MCU’s integrated development environment (IDE). The rapid development of applications capable of controlling things locally and providing efficient data transfer management to and from the cloud must be accessible to the millions of web developers, systems developers, and computer scientists who understand enterprise IT environments. These developers are accustomed to programming using high-level API’s that provide direct access to services rather than AT commands that act as a mnemonic, proprietary language used to configure a modem and send data blocks.

There is an ecosystem in place to support this model of development for intelligence at the edge, comprised of hardware, software, and wireless connectivity. It includes the ARM Mbed OS-based programmable MultiConnect xDot and MultiConnect mDot LoRaWAN-certified system-on-modules (SoMs) that feature a range of up to 10 miles line-of-sight or 3 miles in buildings. The MultiConnect Dragonfly incorporates certified and mobile network operator-approved (MNO-approved) global cellular licensed band technology, including LTE-M and NB-IoT on the recently launched Dragonfly Nano.

The advantage of Mbed running on target devices such as the MultiConnect Dragonfly or LoRaWAN Dot modules is that it reduces that telematics complexity. By utilizing pre-developed MultiTech libraries that provide native wireless and telecoms middleware protocol stacks, sensor-to-cloud drivers, and remote device management, enterprise computing professionals can leverage a faster, less-complicated path to designing IoT edge applications. This approach does not require deep domain expertise in embedded electronics or LoRa/cellular wireless, and enables enterprise developers to focus solely on developing applications that solve business challenges.

In a similar way, MultiTech products such as the MultiConnect Conduit programmable gateway support IBM’s NodeRED visual development tool. Here, applications can be simply created and deployed with the click of a button. An intuitive graphical programming tool ideal for rapid prototyping, NodeRED is designed for IT professionals to help them optimize and scale behavior at the edge of their IoT network.

Opportunities for cost-effective expansion of IoT implementation

For a growing number of businesses today, resources and effort are devoted to data collection rather than to data processing. The highly centralized data management systems in place cause opportunities for business process improvement to be missed, and result in substantial and unnecessary operating expenses.

By incorporating intelligence into devices at the edge of IoT networks, businesses can rapidly accelerate processes, reduce data storage and data transfer costs, and improve system resilience.

Such an approach calls for the deployment of smart gateways and modems that support IT-friendly development ecosystems such as ARM Mbed OS. MultiTech products are available today to support such an approach. 

Passionate about creating innovative global wireless products & services, developing strategic alliances and driving industrial ecosystems. Over 20 years international experience disrupting markets with new technologies, improving brand recognition, growing market share and profitability.

More from Daniel