One of the most interesting and dynamic areas of technology today is also one of the hardest to define: that is, the space between the massively scalable infrastructure of the cloud, and the sensor-based devices that are proliferating all around us. To understand this better, I spoke to Steve Jennis, corporate VP and Head of Global Marketing at hardware platform and connectivity provider ADLINK.
1. What do you mean by Edge Computing, with respect to Cloud and Fog Computing?
“Edge Computing” is still in the eye of the beholder. Cloud Service Providers use the Edge as a reference to any computing resources on their users’ premises, the telecoms industry sees it as the end-node of their proprietary network (as in Multi-access Edge Computing), and corporate users often think of the term as the boundary between operational technology (OT) and information technology (IT). All are valid: Edge Computing is computing at the edge of a network, but the user’s context is obviously important.
By comparison, the term “Fog Computing” represents the continuum between data sources, Edge Computing and Cloud Services. As such, Edge Computing is a subset of Fog Computing, ‘the meat in the sandwich’ between the IoT’s “Things” and Cloud-based Services. Fog Computing is about exploiting compute resources anywhere in an end-to-end system, to add the most (business) value overall.
Meanwhile we have “Cloud Computing”: despite its enormous growth it has some well-understood limitations. For example, the cost of exclusively using Cloud Services can be unacceptable if you are generating terabytes of data every day. That drives IT people – who tend to think top-down – to consider Edge Computing as a complement to Cloud Computing.
Simultaneously, we have the “hard-hat” guys in their OT world, working at the sharp-end of data collection, analysis and control, with a need for real-time systems that require 24/7 fault-tolerant operation. These guys think bottom-up, they often look suspiciously at the corporate IT world and feel that IT people don’t understand their production computing needs; they see Edge Computing as the boundary between themselves and their counterparts in IT, a boundary where their ‘northbound’ data is processed (ingested, normalized, aggregated, analyzed, abstracted, etc.) before it enters the traditional corporate IT domain.
2. If it’s all about enabling end-to-end IoT solutions, how do you see the opportunity?
A few years ago we were in the proof-of-concept era: the main question around IoT was, would the technology work end-to-end? That question is answered today (although concerns around security remain, see below), as are questions around business value: the business case for exploiting the IoT ‘tool kit’ has become a nod, whether for operational excellence reasons (e.g. predictive maintenance), or to support new business models (e.g. post-sale connectivity, services and subscription models).
Therefore, the focus has now shifted to, how do I get started, how do I deploy, and how do I manage the new risks? End-to-end IoT solutions are, by their nature, multi-technology, multi-vendor and multi-standard. They also almost always include both greenfield (new) and brownfield (legacy) data sources and data sinks. All of which is moving the focus to IoT systems integration. The bigger vendors (the traditional leaders in IT or OT) have a single vendor, one-stop shop culture, but IoT solutions aren’t like that. So, who do you turn-to to take end-to-end responsibility for these heterogeneous IoT solutions – an in-house multi-disciplinary team, your trusted local services supplier or a major SI? A lack of good options is the biggest bottleneck in IoT solutions deployment today, across vertical markets.
We’re also seeing two models of IoT deployment. Bottom-up, equipment providers are adding IoT technologies into their product lines, enabling evolutionary adoption by end users within normal technology cycles. And top-down, strategic digitization opportunities and threats are getting a lot of attention, so turning IoT threats into opportunities is a growing concern for company boards, CEOs and CIOs. Both models lead to greater enterprise digitization in pursuit of operational excellence (the cost line) and support for new revenue opportunities (the top line).
Putting the two together, we see machine providers increasingly supporting as-a-service models, marrying IT and OT worlds to optimize post-sales performance and provide customers with new services. Right now, the bottom-up (evolutionary adoption) model is the most prevalent, but five years from now, it will become more balanced as supporting new business models increasingly drives IoT investments. i.e. as one vendor introduces a new service, its competitors will have to quickly follow-suit to remain competitive.
3. How is ADLINK responding to this market evolution?
20 years ago, ADLINK was a traditional electronics manufacturer, building boards and modules, often to customer designs. Soon thereafter, ADLINK started designing its own innovative (analog and digital, hence AD-LINK) embedded computing products and building its own brand globally. This business has been consistently successful over more than two decades. Then, a few years ago, Jim Liu our CEO, added to our corporate strategic vision and ADLINK entered the emerging market for industrialized IoT products and solutions, essentially enabling “connected embedded computing”.
As we started to think about the elements of end-to-end IoT solutions, we quickly realized how much of an opportunity existed at the Edge. Edge computing really is virgin territory in computing, where no incumbent vendors dominate, making it a great growth opportunity. But, as mentioned, the channels-to-market for deploying IoT solutions are relatively immature. As a result, we offer what we call “Digital Experiments as-a-Service” (DXS) — where we partner with customers who are looking to improve their operations or prove-out new business models and revenue streams.
As we help our customers and learn more about the best opportunities-of-scale (both in terms of the size of deployments and the number of potential customers) this solutions view also helps drive the way we embed IoT tech into all our enabling products: platforms, data connectivity and advanced application enablement (e.g. AI-at-the-Edge). Through this top-down (DXS IoT solutions) and bottom-up (enabling products) approach, we support our customers in their embrace of IoT technologies and also help our systems integration channel partners to respond to the huge IoT solutions opportunity in front of them.
4. Where are you seeing the most maturity and growth in end user adoption of IoT?
Overall, we are looking to address two specific questions: Firstly, how to identify customers and solutions with the biggest potential upside, i.e. which users and Use Cases will deliver the best ROIs? And secondly, how to identify applications that can really scale, in terms of system size per customer and/or the number of potential customers? The answers define the sweet-spots in the overall market for us.
In addressing these questions, we prioritize engagements with forward-looking, entrepreneurial and innovative customers rather than any specific vertical markets. It is the customer’s culture and attitude that is more important than their application domain.
That said, we are spending a lot of time with manufacturers (particularly in terms of enabling smart factories) and with a wide range of machinery makers, who now see their products as valuable data sources in an IoT context (in addition to their traditional functionality). But we range across many verticals, and engagement depends mostly on the forward-thinking and innovation culture of the customer. So, in summary, the customer’s willingness to innovate, experiment and explore is more important than the vertical market in which they operate.
5. Where do integrators fit in the end-to-end IoT solutions ecosystem?
For IoT solutions to work end-to-end you need the right team of players, including both users (domain experts) and specialist partners (complementary services providers). So, we’re working with a broad set of partners, both major – such as Intel’s market-ready solutions programme – and smaller – like specialist systems integrators and local services providers – to reduce the user’s barriers to deployment of IoT technologies.
We still see a channel bottleneck in terms of skills and experience in many vertical domains, so working with innovative partners we can learn together how best to create new business value from IoT solutions. ADLINK will continue to act as a multi-vendor, end-to-end solutions advisor and provider, working with preferred systems integrators to develop the IoT solutions eco-system, and thus overcome the “getting started” and then the large-scale deployment issues that end users and machinery makers face today.
My take: There’s substance behind the fog: watch out, cloud providers
I confess, I’m not a great fan of the term “fog computing” as it focuses more on the problem rather than the solution. However, offers a relatively accurate description of IoT’s current state of affairs: a lack of clarity pervades, alongside more general agreement on standards and norms. These are symptoms of where we are, rather than inherent problems, which will be treated over time.
The foggy nature of things is also a smoke screen for what could be one of the most exciting areas of technology in years to come. I don’t want to overstate this, as it starts to sound like hype, but let’s think about the pervading architectural model: cloud.
Right now, we have a layer of rhetoric which assumes that all processing and storage will centralise to a small number of providers: this is variously termed as “the journey to the cloud.” Perhaps it may take decades, goes the thinking, but it will happen. Hybrid architectures are a stop-gap, a Canute-like attempt to stave off the inevitable.
Fog computing, a.k.a. highly distributed and self-orchestrated processing systems, flies directly in the face of the hyper-centralised cloud model. In the foggy world, technology is moving rapidly from a set of standardised boxes and stacks, to a situation where anything goes. The mobile phone or the home wireless hub are just as able to integrate sensors and processing, as any custom-built device. And they will.
When we do achieve a level of standardisation (and move away from this wild west), we can expect to see an explosion in both innovation and uptake. Organizations that have built their businesses on the centralised models will no doubt adjust the rhetoric to suggest that the cloud has extended right out to the sensors. But they will have their work cut out keeping up with the new competitors that will emerge, out of the fog, to take market leading positions seemingly from nowhere.