WavebreakMediaMicro - Fotolia

Industrial edge requires a focus on machines

After the office edge and the mobile edge comes the industrial edge. We look at a shift in computing paradigms

Edge computing means that applications are written for devices, not people – so IT and operations must also recalibrate their relationship.

In 2018, McKinsey estimated that up to $500bn in value could be generated if industrial companies drove transformational change in their operations using digital technologies.

This transformational change is at the heart of Industry 4.0 and digital manufacturing and it is being powered by edge computing devices that use sensors to provide real-time feedback loops from machinery that they monitor and control.

The Industrial digital transformation report from Freeform Dynamics for Fujitsu notes that the challenge for the CIO and COO is that operational technology is no longer discrete. “The smart factories of Industry 4.0 are connected factories – their smart machinery is on the industrial internet of things, and they must communicate with the rest of your organisation and its supply chain – and potentially with the world beyond,” says the report.

“The same goes for your vehicle fleet, the medical devices around your hospital, your inventory management systems, and much more. All this distributed yet intrinsically connected technology gives your IT network yet another edge: after the office edge and the mobile edge comes the industrial edge.”

But this edge is very different from the office and mobile computing edges, where IT provided employees with PC desktop and laptop devices, desktop applications and smartphones. Satyam Vaghani, vice-president and general manager for internet of things (IoT) and artificial intelligence (AI) at Nutanix, says: “The rise of the edge will require a redistribution of the centre of gravity.”

Vaghani says this means there will be a major shift in computing paradigm from one that is mainly focused on delivering applications that are used by people, such as email, web browsing and social media, to applications that are predominantly machine-oriented. Such machine-to-machine interfacing enables the processing of sensor data and the use of AI and analytics techniques to convert raw data into business insights.

There are a number of architectural and technical concerns that must be addressed when looking to incorporate edge computing into IT. Nelson Petracek, chief technology officer at Tibco, recommends organisations that are looking to harness edge computing to begin by considering the capabilities of the device itself, such as its processor and installed memory, battery life and available storage. On the software side, edge computing considerations include the operating system, the reliability and performance of the connected network, and the need for the device to operate when not connected.

But Petracek believes businesses also need to think about both upload and download network speeds, system monitoring requirements, security, the location of the device in terms of its distance from the primary operations control centre, how software will be deployed and how upgrades will be performed.

Read more about edge computing

Edge computing is regarded as the next frontier for data, with rapid growth expected in the market. We assess the prospects.

Edge computing is currently one of the most important trends in IT that is likely to complement cloud by supporting new and emerging workloads. we explore how edge devices live alongside cloud computing.

“Maintenance and continuous improvement are interesting issues as we now have a situation where a [data] model is not necessarily deployed in a datacentre or in the cloud, but is potentially running on hundreds of thousands of devices,” he says. “That in itself presents an interesting problem. Organisations can, of course, containerise that logic and use different technologies to push logic out to the edge. However, one of the main issues – other than considerations around security – is how to make sure that if they do have to push out an update to 100,000 devices, they all receive it, and all update themselves.”

According to Petracek, rolling back an update that has been installed on 100,000 distributed devices is a very different challenge to rolling back software on a centralised cloud platform.

“In this scenario, organisations may decide to put transformation and simple streaming or rule logic at the edge,” he says. Some kind of gateway then acts as an intermediary, which handles the core functionality required by the edge computing application. Petracek says this means organisations no longer have to directly manage hundreds of thousands of endpoints; instead, they manage dozens of gateways.

Architecturally, says Petracek, at least some or all of the raw IoT information generated at the edge, and the results produced by the edge data model, need to be forwarded up to the cloud. “This ensures there is a complete understanding of what happened at the edge, as opposed to just a summary, and also enables the evaluation of the models to try to identify cases where the model is not performing correctly,” he says.

Petracek says it is important to determine whether model drift is occurring and whether the model is performing as accurately as it should be. One way this can be achieved is to run periodic snapshots where, every so often, the edge device sends more data to the cloud so that it can be recorded, and evaluated to check that the data model being run is still valid, he says.

Electrical power is another architectural consideration. Over the last decade, datacentre computing has become more energy efficient. Sue Daley, associate director, technology and innovation at TechUK, says: “The increased convergence of emerging technologies such as 5G and edge could represent a significant growth in what is effectively a new form of distributed IT.”

AI, cloud and edge

Daley says 5G and edge servers could provide local, time-critical, low-latency data processing that work together to enable IoT devices to become a reality, transforming traditional sectors, such as manufacturing. “We could see a future where the convergence of AI, cloud and edge computing provides the vehicle and roadside digital infrastructure needed to make driverless cars a reality,” she says.

But for Daley, a key issue is that industry must account for the energy impact of an increase in data processing at the edge. “During the last decade, we have encouraged the consolidation of IT functions, including data collection and processing, into larger, purpose-built facilities where electricity consumption is transparent, where energy stewardship is scrutinised and where there are strong incentives for efficiency,” she says.

“If we see increased demand for more data processing at the local level by edge computing, before data is sent to the cloud, how will we aggregate that new energy use and how can we ensure that it is transparent and accountable?”

But there is also potential for energy to be reduced through the use of edge computing. Thanks to factors such as Moore’s Law and virtualisation, the energy needed to process a given amount of data has decreased by over six orders of magnitude in 30 years. Daley says: “Some people also consider that increasing use of edge datacentres can be far more autonomous in terms of electricity supply, and that a much wider range of power sources will be available to them because they are less monolithic and resilience can be built in by duplication and overlap rather than by continuity of supply to individual units.”

McKinsey’s article The next horizon for industrial manufacturing, published in 2018, offered a number of suggestions on how industrial companies can make a success of next-generation manufacturing that uses Industry 4.0 techniques such as edge computing. While companies need to set clear business objectives as to what they want to achieve, among the key stumbling blocks, its research found, was in the maturity of IT. McKinsey reported that almost half (44%) of the 700 respondents it surveyed for its 2018 report Digital manufacturing – escaping pilot purgatory, regarded IT deficiencies as a major challenge in implementing digital manufacturing.

Dale Vile, distinguished analyst at Freeform Dynamics, says: “In order to implement a relatively future-proof edge computing environment, you have to start with sound architectural design. This not only means thinking through the dependencies and flows across the various tiers from the edge to the cloud, but also considering how provisioning, configuration, monitoring and administration of systems and components will be handled on an ongoing basis. The phrase ‘design for operations’ sums up the mindset to adopt here.”

For Freeform Dynamics, edge computing provides a way to link IT with the operational technology within manufacturing and industrial systems. But beyond the technical requirements of connectivity, data exchange and protocols that enable machines to communicate across an IT network, Freeform warns that CIOs and COOs must also consider how they share and allocate budget for devices that bridge the divide between operations and IT.

Read more on IT architecture

Data Center
Data Management