1660343497668 Jasonandersenstratus2017

Driving processing power to the network edge

Feb. 6, 2019

A logical, stepwise approach is critical for edge computing.

The trend of making normal consumer products into digital devices now extends to hairbrushes, forks, water bottles and other mundane objects advertised as “smart.”

Most, if not all, of these supposedly smart consumer goods don’t have the right kind of intelligence to provide significant user benefits. However, it’s a different story when it comes to smart commercial and industrial products, which typically can provide substantial benefits if implemented correctly.

Automation systems originally required physical centralization and significant installation. However, recent developments have resulted in all sorts of powerful computing and sensing devices with far smaller packaging, minimal power consumption, and the ability to be cost-effectively installed almost anywhere and connected with wireless networking.

Stratus Technologies' Jason Andersen

Commercial buildings, laboratories and other such facilities are ideal candidates for environmental monitoring. Today’s connected cities have many remote sites and lightly manned infrastructure locations which can benefit from early warning systems. Automated transportation requires extensive telemetry to stay on track. These and other industrial and commercial IoT applications are powered by smart sensors.

Smart sensors exist in the field at the extreme edge of what traditional networking methods can access. Proactive users are looking to transport valuable IoT information up to the cloud where they can easily access and act upon it.

Of course, logical steps and best practices are required for establishing an edge computing architecture suited for processing and analyzing raw IoT data to produce useful results. Let's explore...

Smart device data challenges

Industry literature touts the recent development of IIoT products, but the reality is that industrially-deployed devices and systems have been becoming smarter for quite a while now. In fact, many skilled end-users have created their own IIoT systems by combining programmable logic controllers (PLCs) and other computing devices for collecting local sensor data and distributing it to higher level systems.

Addressing this market, vendors have developed multi-function smart sensors. A condition-monitoring device for rotating equipment may report three axes of vibration and temperature. Power monitor devices track energy usage and can detect deteriorating conditions. Temperature, pressure, level and flow detectors deliver early warning of trouble in unstaffed locations.

Traditional methods required costly conduit and wire installation. A major benefit of modern IIoT solutions is the ubiquitous adoption of wireless technologies and battery operation. These advancements let users rapidly and economically deploy sensors.

Likewise, users can find value in IIoT technologies by transferring the data they produce up to cloud-based supervisory and analytical systems and establishing preventative or predictive-maintenance programs.

But there is a catch with smart devices: they provide vast amounts of raw data without corresponding context. Some form of edge computing is needed to pre-process this data and boil it down to the essential information. The latest edge-computing methods are ideally positioned to take on this task, which protects upstream systems from being overwhelmed, among other benefits.

The case for edge computing

Dozens of IIoT devices blasting out hundreds of data points on a continual basis would lead to an indecipherable overload. Performance-wise, pushing all the data up to the cloud is technically possible, but obscures value. The key to harnessing raw data produced by intelligent devices is to first turn it into useful information.

Effective edge computing is the best way to achieve this, providing these and other benefits:

  • Cleaner data (pre-processing raw data results in better information)
  • Better security (guarding smart devices against cyber-threats)
  • Lower costs (streamlining traffic and network management tasks closer to the edge)

Pre-processing takes several forms. Noisy raw signals can be smoothed, values can be selectively monitored, and intermittent or fast-acting events can be counted locally without timing concerns. Cleansed data means supervisory systems receive the best-possible information.

Smart devices represent a cyber-security concern because each can be an entry point for attackers. An improved security model isolates smart devices behind edge-computing elements better suited for defending against cyber-attacks.

Significant cost savings are possible by incorporating edge computing into an IIoT-to-cloud system. The biggest driver is streamlined data, reducing upstream storage, bandwidth and computational resources because only essential and cleansed data is handled.

Steps to success

Edge-computing solutions must provide clean data in a secure real-time manner to support mission-critical demands enabling long-term analytics and even artificial intelligence. A methodical approach encompasses these steps:

  • Develop an architectural strategy
  • Establish infrastructure
  • Get securely connected

An overall architectural strategy provides a roadmap for connecting smart devices through edge computing and up to the cloud, addressing four levels: interoperating with field devices, capturing data, controlling information and cloud intelligence.

The edge-computing infrastructure must be simple, protected and autonomous. It must be easy to deploy quickly, readily restored if disaster strikes, all while being usable and manageable by remote-site personnel. Logical concentrations of smart devices must be identified so edge-computing hardware can be located geographically near these areas.

Of course, connectivity is often at odds with cybersecurity. Devices that communicate easily are often not very secure, while implementing rigid security makes system configuration much more involved.

A purpose-built answer

Standard PC hardware is not recommended for infrastructure data-processing systems and edge-computing applications for several reasons. Common PCs aren’t hardened for edge environments, don’t offer the best redundancy options, and can require significant hands-on support.

Purpose-built edge computing hardware platforms for industrial service are a better option, optimized for the edge-located role and tailored to the operational personnel who will support them. Integrated redundancy with no single point of failure means these systems will run indefinitely, while autonomous self-monitoring and remote management minimize maintenance effort.

The need for edge computing is clear. A large and growing variety of data is available from IIoT and intelligent field devices, but it requires pre-processing at the edge for cloud services to be most efficient and cost effective. Purpose-built edge computing solutions provide the most reliable way to deliver upon the promise of the IoT.

Jason Andersen is vice president of business line management with Stratus Technologies.