How do IoT and edge computing work together? Julian Chesterfield, Founder and CTO of Sunlight.io, joins Ryan Chacon on the IoT For All Podcast to discuss everything you need to know about IoT and edge computing. They explore the difference between IoT and edge computing, the IoT challenges that edge computing solves, the cost benefits of edge computing, the role of AI and ML in edge computing, the challenges of edge computing, adding edge computing to legacy infrastructure, computer vision, the real-world use cases of IoT and edge computing, and the future of edge computing.
Julian Chesterfield has worked in the cloud and virtualization industry for over 15 years. After completing his PhD at Cambridge University, he became involved in writing a technology called Xen, which became the foundation of the public cloud back in 2007 when it was adopted by Amazon and most of the emerging public cloud infrastructure providers of the day.
Julian began Sunlight.io as an incubation project in 2013 to build a lightweight hypervisor for low-power ARM processors. On seeing its value for data-intensive applications at the Edge, Julian and the founding team built it into a full hyperconverged infrastructure stack – Sunlight. Sunlight secured $6M in Series A funding in December 2020 and has gone on to launch new products, including the Sunlight Infrastructure Manager (SIM) and Marketplace. Julian previously held positions at OnApp, as Chief Scientific Officer, and Citrix, as Storage Architect.
Interested in connecting with Julian? Reach out on LinkedIn!
Sunlight.io is the thinnest, fastest HyperConverged Infrastructure platform built for applications that run at the edge. They make running and managing applications and infrastructure at the edge as easy as “in the cloud” across 100s to 1000s of edge devices. Sunlight works with efficient, ruggedized edge hardware – so you can consolidate all your in-location edge applications with centralized management, high performance, and high availability.
Their flagship product is the Sunlight HyperConverged Edge which is a full-stack, bare-metal virtualization platform that combines the computing, storage, and networking of one to multiple servers into a single system or cluster. Each cluster, deployed in a remote location, can consolidate multiple instances of Windows, Linux, or containers on x86, AMD, Arm, and NVIDIA Jetson.
Sunlight NexCenter is the centralized console and API that provides a single pane of glass to manage and monitor edge resources, take backups, move workloads, and deploy new remote clusters. A core feature of NexCenter is the AppLibrary which allows customers to build and access playbooks (images & recipes) for deploying applications and the supporting infrastructure to 100s or 1000s of remote clusters with a single click.
Key Questions and Topics from this Episode:
(00:44) Introduction to Julian and Sunlight.io
(02:09) What is the difference between IoT and edge computing?
(03:55) What IoT challenges does edge computing solve?
(05:53) Cost benefits of edge computing
(07:07) Role of AI and ML in edge computing
(08:55) Challenges of edge computing
(11:11) Can edge computing be added to legacy infrastructure?
(12:42) What is computer vision?
(15:14) Real-world use cases of IoT and edge computing
(18:47) Future of edge computing
(21:17) Learn more and follow up
– [Ryan] Hello everyone, and welcome to another episode of the IoT For All Podcast. I’m Ryan Chacon. And on today’s episode, we are here to talk about IoT and edge computing, how they work together, how they’re different, challenges when building out IoT solutions and what edge computing can do to help solve those problems. With me today is Julian Chesterfield, the founder and CTO of Sunlight.io. They are an infrastructure platform company focused on building those applications that run at the edge. A lot of value here. I think you’ll enjoy it a lot. Subscribe to our channel, give this video a thumbs up and hit that bell icon, so you get the latest episodes as soon as they are out. But other than that, on to the episode.
– [Ryan] Welcome Julian to the IoT For All Podcast. Thanks for being here this week.
– [Julian] Thanks for having me. Great to be here.
– [Ryan] Absolutely. I’m looking forward to the conversation, but I wanted to kick this off real quick with having you give a quick introduction about yourself, and the company to our audience, if you wouldn’t mind.
– [Julian] Yeah, happy to. So, my name is Julian Chesterfield. I’m the founder and CTO of Sunlight.io, and we build an edge compute platform that makes it very simple and easy for enterprises and, you know, customers to deploy and manage applications and services at the edge. And the edge for us is anywhere outside of the traditional core data center, whether that’s, you know, in industrial environments, in retail stores, supermarkets, restaurants, things like that.
– [Ryan] Fantastic. Sounds like you all cover a variety of different industries. Do you have any that you kind of are more hyper focused on right now or is it pretty widespread?
– [Julian] Yeah, I mean, you know, I think we’re obviously quite widespread, but I think the areas that we’re seeing the kind of the earliest adoption of edge compute infrastructure, I think particularly in the retail space, you know, retail restaurants, quick service restaurants where there’s a lot of, you know, modernization and technology acquisition, that’s starting to happen. And also in the industrial space as well, I would say probably the main focus areas for us.
– [Ryan] Fantastic. So let’s go ahead and kick this off. I wanted to ask you kind of a high level question. We’ve talked about edge computing in the past, very different kind of angles here and there depending on the guests. But today I wanted to kick it off by asking you a question. I know kind of comes up a lot in conversations that I have outside of the podcast, which is when we’re talking about IoT, we’re talking about edge computing, they often get talked about together. But can you kind of just high level explain it to our audience as to what is the real difference when people are talking about edge computing versus IoT.
– [Julian] Yeah, I mean, the way that I see it, I think, you know, obviously the two terms often get used together and obviously they’re very complimentary technologies. I think the way that I see it is that, you know, it’s really about the scale of device. So IoT devices are typically, you know, smart devices that run in an environment to be able to, you know, capture data, things like sensors, things like video cameras, you know, temperature sensors, movement sensors, things like that. Which of course are becoming smarter and have the ability to be able to make smart decisions themselves. The edge piece really is the bit that ties all those devices together. So, often there’s some sort of logic that needs to run slightly further back from all of those, you know, smart IoT devices that are out there to be able to run, you know, larger scale compute logic to be able to tie together all that information. So really the two things are complimentary, but often the edge infrastructure will run on slightly more capable nodes, more infrastructure that has more compute capacity, and so on.
– [Ryan] Fantastic. Edge computing has been around for a little while, but it’s still growing in adoption in the IoT space, what are the main IoT challenges that edge computing is kind of is focused on solving?
– [Julian] Well, I think really, you know, what we’re seeing is this, you know, more significant amounts of data generations starting to happen and you know, there’s a lot more latency sensitive applications now that need to be able to, you know, either process data, and you know, infer, you know, certain information from the data very quickly. So to be able to run control systems and things like that where you have to be able to make critical decisions very quickly in order to perhaps make a decision about switching a machine on or off or, you know, changing the temperature control for something. But, you know, and that comes from a lot of different sources and we’re seeing particularly the acquisition of a lot more video data now that allows, in the industrial space for example, using video analytics to be able to identify, you know, safety attributes in an industrial environment or being able to track movements and you know, movements of goods and things like that. And those types of decisions need to be made quickly, and often the bandwidth associated with that is very significant. So you’ve got, you know, large volumes of data being generated that you can’t necessarily, or you wouldn’t necessarily want to transmit into a large cloud data center, for example. You want to be able to process that data on the site and really that’s I think one of the key drivers for more automation and more intelligence being pushed out towards the edges.
– [Ryan] Yeah. Are there any cost benefits when it comes to kind of the edge pushing a lot of that to the edge as opposed to pushing it back up to the cloud, and dealing with it there? What other benefits are there, I guess associated with it?
– [Julian] I think increasingly there is a lot more cost benefit. You know, we’re seeing the emergence of a lot more edge appropriate hardware that’s becoming more commoditized. We see, for example, in the video analytics space, you know, things like the NVIDIA Jetson platform, which is a purpose-built system on Chipboard that has, you know, embedded GPU processing capability, and this is significantly lower cost than what we see typically today in the data center in terms of the, you know, the cost of you know, GPU capability and so on. And so we are finding I think that it’s becoming more affordable both from a transmission point of view, you know, not having to have very large, network bandwidth capability in order to transmit lots of uncompressed data up into the cloud, but it’s more affordable now to put compute infrastructure close to where the data’s generated.
– [Ryan] Fantastic. I’ve had some guests on talk about being able to incorporate machine learning kind of at the edge, you know, other and also run AI models and things like that. Is that something that, or I guess how does that all kind of fit into to the edge as well?
– [Julian] Yeah, and I think that kind of ties in with what I was saying before, and I think that’s one of the things that the edge compute capability can really bring to the table, is the ability to be able to run a machine learning model against the live data that’s being generated in that environment. So, you can make decisions about things like facial recognition or, you know, identifying objects, if it’s something like a quality control system on a production line in the factory, being able to, very affordably identify defects in quality control, and things like. That sort of capability, being able to run that on the edge compute infrastructure is really key. And some of that is leveraging things like GPUs to be able to accelerate that processing and drive that logic. We also see a lot of machine learning models that run on just general purpose computes as well. It really depends, I think, on the granularity of data, the number of samples that have to be generated and how fast that data has to be, those decisions have to be made.
– [Ryan] Absolutely. Yeah, it does always sound like it’s a decision on kind of to what extent are you building out or implementing the decision making at the edge depending on your use case, what your needs are and things along those lines. But one of the things I wanted to ask you, ’cause we haven’t really dove into this too much in the past, is when it comes to deploying edge computing or just kind of working on bringing edge computing into a solution, what are some of the challenges that you’ve come across or customers of yours have come across in the past that are kind of worth noting or worth kind of mentioning to the audience for them to be able to lookout for or understand that when it comes to edge computing there are challenges you have to consider or think about and how to kind of, you know, navigate them?
– [Julian] Sure. So I mean, you know, some of the challenges that we see, generally across edge environments is that, you know, you often don’t have a lot of technical staff in these locations. You know, if you think about, you know, an industrial factory setting or a retail environment, you know, these are not places where you have technical teams on site often to be able to deploy, manage, and support that infrastructure. So, edge computing infrastructure has to be very robust, ruggedized and very reliable. You know, ruggedization is another thing. A lot of these environments are not, you know, it’s not running in a air conditioned, rack mounted, server data center environment. You know, often these are servers that might be up on a wall or under a desk and you know, so they have to be able to run in these sort of ruggedized environments with just normal, you know, air cooled often we see some embedded system on chip type devices. And then, the other challenges I think also sort of scaling as, you know, the edge often, we characterize as being hundreds of thousands of distributed locations, and each location might be just a very small compute unit that needs to be able to run in that location. So challenges around manageability and scale for all those devices, those are the sorts of things that we see that really differentiate the edge, I think from, you know, traditional data center environments, cloud environments.
– [Ryan] Right, right. If a company has a deployment, it’s already out in the field, can you kind of implement edge computing practices out into something that’s already deployed? Or how does that kind of work if there’s a system already being used and, you know, the benefits of obviously edge computing could greatly affect the success, the collection and the interpretation of the data. How’s that handled?
– [Julian] So yeah, I mean, most edge sites aren’t greenfield deployments. If you take the retail environment, there’s often some compute, there’s some applications that are already running in those types of environments. What we’re seeing is this drive towards bringing in new technology solutions and being able to manage that more efficiently as these stores become more, you know, technologically advanced to be able to run those sorts of environments. So, one of the things we focused on with the Sunlight solution is very much, we build a type one hypervisor based solution, which allows us to run any type of legacy application. So you may have some sort of legacy Windows servers running, for example, in these environments, those can be lifted and dropped onto the Sunlight platform alongside, more advanced sort of containerized solutions that might run in those environments as well.
– [Ryan] Fantastic. One question I wanted to ask you kind of as it relates to this conversation, and it’s been a pretty hot topic recently, is computer vision and kind of how that works and how that’s a very good example of kind of the power of edge computing and what can be done on that side of things. Can you talk a little bit more about kind of the requirements that go into making computer vision possible and kind of just, I mean, even a high level, just so our audience understands exactly what computer vision is?
– [Julian] Yeah, I mean, at its core computer vision of course is, taking video feed data from, you know, high quality cameras and being able to infer and make decisions about, you know, data and the visual data that’s being processed. I mean, we see things like, you know, classic example that I use is the things like the just walkout technology that we’re seeing in retail environments now where we’ve gone from having, very manual checkout processes where you have, somebody sitting at a till or you’ve got a machine where you’re scanning a barcode in order to check out goods, to using fully automated computer vision driven logic in these stores where you’ve got hundreds of cameras now in the ceilings or behind shelves tracking people walking through a store, figuring out what they’re picking off the shelf, identifying what those objects are, and then, you know, allowing you when you walk out of the store to then, you know, charge you for what you’ve walked out with. And the logic that sits behind that is you’ve got to have a lot of, you know, machine learning models that run on compute infrastructure that sits behind all of those high bandwidth cameras. And that of course, that’s very latency sensitive workloads, you’ve got to be able to make a decision quickly, and you have to be able to track that in real time. So, that’s a great example I think of how the logic has really moved from this sort of very centralized model running in the cloud or in the data center out towards the stores themselves. And that’s the kind of thing that drives this, you know, technology acquisition in those environments.
– [Ryan] Yeah, absolutely. No, that’s fantastic. Thank you so much for kind of breaking that down. Throughout this conversation, we’ve talked a number of times about kind of use cases, and different things companies are doing or examples of solutions where edge computing is being deployed, and kind of the benefits. But I’d love it to kind of break down a few other examples a little bit more in detail, if you wouldn’t mind, just of organizations that are using IoT and edge computing kind of out in the real world, like some good examples just to kind of bring this all full circle for our audience.
– [Julian] Yeah, I mean, we see a lot of, you know, a variety of different use cases for this kind of new technology. I think, you know, good example is in the industrial space where, you know, IoT and computer vision, you know, camera devices are being used for safety. So that’s always becoming, you know, more and more important for companies that have to operate, these large sort of industrial environments where, you know, you may have, quite dangerous environments in some cases, lots of, you know machinery, lots of hot materials and things like that. And making sure that the people who are working in these environments are wearing the appropriate safety clothing, hard hats, things like that. You know, we’re seeing solutions where in software using AI and machine learning now these sorts of decisions can be tracked and made instantly in these environments to alert people if they’re not wearing the appropriate equipment or they’re standing in the wrong place in a factory, for example. That’s a great example. I think another good one we’ve seen in the retail space is around using sensor devices to be able to control electricity consumption. So particularly with things like fridges, refrigerators and freezers in stores now. Particularly in Europe, you know, we’re seeing this huge increase in energy prices, and so it’s become really important for these organizations to control as efficiently as possible the amount of electricity that they consume. So using smart software to be able to measure temperature and to be able to adjust temperatures for these types of devices is really key.
– [Ryan] Fantastic. Yeah, it seems like as edge computing grows, the value, the benefits, the applicable use cases just continue to kind of expand. And I feel like everyone that I’ve talked to recently is, you know, very high on kind of what edge computing is doing, the benefits it’s providing and kind of the value overall that it’s bringing to IoT deployments, whether they’re, like you said, Greenfield or they’ve already been kind of launched, and it’s quite interesting. It’s quite interesting to kind of follow along and see what as technology evolves and all the other areas of IoT the connectivity side, the hardware side, and just seeing what edge computing is doing to enable the analysis of data, the collection of data, the interpretation of that data, just a lot of different things that are enabling different use cases that probably weren’t as applicable before or as kind of or likely to be able to be kind of worked out. So it’s been a very interesting space to kind of follow along. And I wanted to ask you, as we kind of move through the rest of this year into just the general future, what are your kind of thoughts and outlook on edge computing? Like, where is it going, what are the kind of the next evolutionary steps of edge computing as it relates to IoT and IoT deployments?
– [Julian] Yeah, good question. I mean, I think, yeah, there’s no doubt that, you know, the adoption of edge computing technology is really happening. You know, we’ve talked about it for a long time. I think what we’re seeing now is that enterprises are really starting to deploy technology solutions. In a lot of cases they’ve been perhaps testing things in pilot phases and things like that, but they’re really starting to roll it out on a much larger scale. I think, you know, there’s a number of things that will happen. I think ultimately the edge computing technology market I think is going to be larger even than a lot of the sort of established data center technology marketers today. I think we’ll definitely see that. But I think, you know, one of the things we’re very focused on, I think is about what we are seeing this sort of commoditization of edge computing in particular. I think, you know, we saw this happen with public cloud where, you know, enterprises move from, you know, running their own data centers, buying their own hardware, having their own IT teams sort of managing this infrastructure themselves to kind of almost outsourcing it into cloud providers as a subscription model. So, you know, you can benefit from the sort of scale, and economics of scale that the big hyper-scaler data center operators can provide. I think we’ll see a similar model start to happen at the edge. And I think that will start with, you know, similar sort of subscription type models to edge infrastructure, to being able to deploy edge infrastructure, and then ultimately to sort of running this almost as a service for a lot of environments. You know, I think that’ll be something that we’ll start to see the adoption of you know, new and interesting subscription models in that space.
– [Ryan] Fantastic. Yeah, like I said, it’s a very exciting space to kind of follow along, for sure. Sounds like you all have a lot of great things happening in the space and for our audience out there who’s listening to this and wants to follow up, learn more, engage kind of after the fact, what’s the best way that they can do that?
– [Julian] So, get in touch with us of course. We’d love to talk to anyone who’s exploring the space or you know, who already has some good solid ideas about the types of technology and solutions that they want to deploy. Come to our website, sunlight.io, and get in touch and we’d love to chat.
– [Ryan] Fantastic. Well, Julian, thank you so much for being here. Great conversation. I really appreciate the time and I think our audience is going to get ton of value out of this.
– [Julian] Thank you, Ryan. Appreciate it.