Now Reading
Connectivity, intelligence and automation mean things won’t be the same again
0

Connectivity, intelligence and automation mean things won’t be the same again

Posted by IoT Now MagazineMarch 8, 2018

Jim Douglas, Wind River

Jim Douglas, the president of Wind River, tells George Malim the concept of the Internet of Things (IoT) is reliant on the introduction of fluid computing, machine learning and changed mindsets in traditional business domains if it is to deliver on its potential.

Although the Internet of Things (IoT) has now been high on organisations’ agendas for several years since its emergence from the machine-to machine (M2M) industry, there remains a long

road ahead for the IoT concept to reach maturity. That road involves bringing technology, society and economics together to create a new digitised arena in which we live, work and entertain ourselves. That is a distant prospect and we are still at the start of the journey.

“Most companies talk about IoT as a continuum. The general concepts are to first connect the devices, then make them intelligent and then to make them autonomous,” says Douglas, who joined Wind River, an Intel subsidiary focused on software for IoT, in 2010. “That has been the drumbeat since the early days of connecting things, but there are underlying technology and business needs that are required to complete that continuum.”

All the ingredients aren’t quite in place and several technologies are yet to find the form in which they will enable the intelligence and automation that completion of the IoT vision requires. The approach to computing resources is one area that Douglas singles out.

“One big topic of discussion is where compute is going to reside,” he says. “Computing power has oscillated over the past forty years between centralised and distributed, with the latest incarnation being cloud – a centralised approach. There is a school of thought in the market that suggests we are moving back to a distributed computing model, but I don’t think that’s necessarily going to happen. Yes, we need to move more compute out to the edge, but what really needs to happen to fully enable the promise of IoT is fluid computing.”

“For IoT to reach its full potential, systems need to be able to access computing resources on a fluid basis,” he adds. “In the cloud, enterprise class virtualisation has enabled elastic computing, where you can orchestrate applications to run on available resources regardless of location. We’ve been working with Wind River customers to map out their future and show them how they can take that concept and extend it. The idea is to orchestrate workloads between the edge, fog and cloud. Creating a topology that allows you to use compute at all levels on a fluid basis such that you can drive workloads to the best compute resource as required.”

Ultimately this will create an environment in which compute resources are available to meet the demands of applications as required with efficiencies generated from maximised utilisation of resources. However, this is complex to achieve and, for many, represents an entirely new IT landscape.

“Embedded system developers have been consolidating workloads on the east/west axis for some time. They have been using embedded virtualisation technology to consolidate federated systems to drive down costs,” explains Douglas. “But, the next challenge is how you consolidate and orchestrate workloads on the north/south axis.”

“One of the keys is to utilise cloud architectures rather than continuing to build custom built embedded solutions,” he adds. “The challenge lies in how to do this effectively while preserving the levels of integrity, performance and determinism that are required in embedded systems deployed in critical infrastructure.”

Douglas says that while Wind River has been very focused on the compute challenges that fulfilment of the IoT concept involves, it is also looking ahead to the business models that the market will create. “There are substantial business model considerations to take into account as we accelerate along the IoT continuum,” he adds. “The drive in cloud and telco data centres towards software-defined networking (SDN) and network functions virtualisation (NFV) has had a profound impact on the business models of traditional equipment suppliers. Their monetisation strategies in the past were tied to the value of purpose built solutions. The push by end users to decouple compute, software infrastructure, and applications and orchestration has created a major disruption in their supply chain. It’s forced their suppliers to rethink how they create and capture value. The transition we are witnessing Cisco make right now is a prime example of this evolution.”

“We’re now seeing this trend in traditional embedded domains such as the industrial control market,” he adds. “Factory and processing plant owners are trying to combat the age old problem of how to introduce innovation into their facilities without disrupting throughput/output. They are exploring how to make use of virtualisation and topology decoupling to enable flexible deployment of new technologies.”

Douglas acknowledges that the plug and ‘pray’ approach that has served industrial organisations well for the past 50 years is broken. “Plugging in a piece of equipment and praying it doesn’t break for 20 years is no longer an acceptable model,” he confirms. “With how fast innovation occurs today, incumbent industrial organisations will be at a huge competitive disadvantage if they can’t figure out how to accelerate the speed and drive down the cost to deploy new technologies. This realisation is what is driving them to evaluate the topology changes occurring in cloud and telco data centres to determine what they can use in their domain. ”

“These approaches potentially create scenarios that are at odds with the business models of their traditional suppliers much like we have seen in the cloud and telco markets. Industrial control companies are trying to learn from adjacent industries and evolving their value propositions while rethinking their monetisation strategies in anticipation of this transition. They’re beginning to understand that more and more of their value is going to reside at the application and orchestration level.”

“We – technology companies – are rapidly addressing the technology challenges to connect disparate devices, make them intelligent, and enable them to learn and act autonomously. To truly achieve the results promised by the vision of IoT, business models need to begin to evolve equally fast such that the economic imperatives of consumers and suppliers are not incongruous. This is very different and a radical transition for many companies and a lot of people are terrified of breaking their monetisation engine.”

Of course, the nature of the compute continuum is only one aspect of the IoT road forward. The other obvious topic to address is data. Data has been equated to the new oil. “Although we’ve made some strides harnessing and making use of data, we have yet to experience the transformational results promised by IoT,” says Douglas. “An analogous situation I read about recently was the advent of the industrial revolution. The big breakthrough was harnessing steam to power the belts and pulleys that drove machinery in factories. Soon after, large DC electric engines displaced steam engines. However, the shift if power generation didn’t fundamentally change manufacturing. It just provided a slightly more efficient way to drive the belts and pulleys. Truly transformation change didn’t occur until the focus shifted to really understand work flow and materials flow and small electric engines we distributed through the factories. We’ve done essentially the same thing with data. Data isn’t new. It’s always existed. We’ve just figured out ways to more efficiently access it, store it, and analysis it. We’re not doing anything radically different with it. We’ve just deployed the DC electric motor.”

A key technology that is increasingly discussed as the transformational agent is machine learning, which Douglas sees as a key for achieving thelofty expectations of IoT.

“Machine learning has arguably been around  since the 1940s and Alan Turing,” explains Douglas. “Early approaches utilised symbolic programming and relied on providing machines rules as a basis of learning. Algorithm development has shifted toward pattern recognition, using supervised and refined learning techniques. We’ve seen a massive acceleration in the use and effectiveness of machine learning as the approach shifted in this direction. It’s still fairly nascent, but the speed of development is going to be phenomenal and the impact will be transformational.”

However, Douglas returns to the importance of compute capability to enable machine learning. “The challenge is pushing compute down to the far edge where data originates such that you can make time critical decisions without dealing with the latency of moving data to the cloud to achieve results. Ultimately, you want to achieve a fluid compute continuum so you can have intelligence where you need it when you need it. This will really let us fully utilise the power of machine learning.”

The fact that Douglas has identified “the will to do it” as a key enabler or inhibitor is significant. Organisations see IoT and machine learning as a new and unfamiliar territory that requires them to upskill. However, Douglas thinks this can’t be overstated.

“If organisations are thinking they solely rely on their domain knowledge to insulate them from new competition from adjacent markets, they’re wrong,” he says. “With access to the right data sets and adequate compute resources, programmers with limited domain expertise can develop algorithms that can achieve remarkable results. It’s imperative that companies recognise the need to move down the machine learning path at an accelerated rate or they are going to be seeing new competitors approaching rapidly in their rear view mirrors.”

This is not a prize for today, but it is the goal for the near future. “It’s not a year away,” acknowledges Douglas, “but it will accelerate massively over the next decade. Fluid computing, IT scalability and machine learning are fascinating and from a strategic point of view, Wind River is looking at how to accommodate all that.”

“This resonates exceedingly well with engineers and terrifies business professionals because they have to rethink their strategies for value creation and value capture,” he says. “Incumbency is not always the asset it once was. It’s going to be a refreshing time for technology and very disruptive from a business point of view.”

www.windriver.com

About The Author
IoT Now Magazine

Leave a Response