What trends are driving edge computing? Jason Shepherd, CEO of Nubix, joins Ryan Chacon on the IoT For All Podcast to discuss edge computing and IoT. They cover the impact of trends on technology providers, the current IoT platform landscape, the intersection of AI and IoT, what TinyML is, tips on a holistic view of IoT and edge computing, and the challenges of IoT and edge computing.
About Jason
Jason Shepherd is a veteran in the Edge and IoT space and is CEO of the device edge orchestration company Nubix. Previously, Jason was VP of Ecosystem at ZEDEDA, and prior to that he was CTO for the Dell Technologies Edge and IoT Solutions Division.
His proven track record as a thought leader in the market is evidenced through his leadership building up ZEDEDA’s ecosystem, the award-winning Dell IoT Solutions Partner Program, and co-founding the EdgeX Foundry and Alvarium open source projects in the Linux Foundation. He has previously served as a board member and the governing board chair for LF Edge.
Jason speaks and writes regularly on technology topics such as edge computing, IoT, AI, 5G, Digital Twin and Ecosystem and has been recognized as one of the Top 100 Industrial IoT influencers. He holds over 45 granted and pending US patents.
Interested in connecting with Jason? Reach out on LinkedIn!
About Nubix
Nubix brings cloud native and containerization principles to highly constrained edge devices that can’t support technologies such as Linux and Docker. Their platform greatly simplifies the development and deployment of applications on microcontroller-based devices that make up 2/3 of the overall edge footprint.
Key Questions and Topics from this Episode:
(00:37) Introduction to Jason and Nubix
(02:04) The trends driving edge computing
(04:35) Impact of trends on technology providers
(07:14) The current IoT platform landscape
(11:02) Intersection of AI and IoT
(13:28) Current state of IoT
(15:31) What is TinyML?
(20:00) Tips on a holistic view of IoT and edge computing
(21:45) Challenges of IoT and edge computing
(23:42) Learn more and follow up
Transcript:
– [Ryan] Hello everyone and welcome to another episode of the IoT For All Podcast. I’m Ryan Chacon and on today’s episode we’re going to talk about macro trends that are driving edge computing with Jason Shepherd, the CEO of Nubix, an edge computing company. Very fascinating and exciting conversation. I think you’ll get a lot of value out of it.
Before we get into it, please subscribe to this channel if you have not done so already, give this video a thumbs up, and hit that bell icon so you get the latest episodes as soon as they are out. Other than that, let’s get on to the interview. Welcome, Jason, to the IoT For All Podcast. Thanks for being here this week.
– [Jason] Thanks for having me.
– [Ryan] Yeah, it’s great to have you. Good conversation. I think we have planned here, but before we get into things, I’d love it if you could give a quick introduction about yourself and the company for our audience.
– [Jason] Yeah, so I always joke, that my motto is if it’s fuzzy, I’m on it. I’m always finding myself with the front end of new tech trends long background in engineering and R&D and CTO roles also move more towards the business side of things. Worked a lot in the industry open source projects a lot of different things, over, over time, really big believer in the network effect.
And how do you leverage, ecosystem to drive change and new things? Currently I’m the CEO for a company called Nubix. We’re focused on how do you deliver a cloud-like experience to the billions of tiny devices out in the field? Devices that are too small to run Linux, but how do you bring containers to those devices?
Which is interesting. I’ve been at Nubix for a few months, but it’s been clicks of, I was at Dell Technologies where I was CTO for Edge and IoT. I worked a lot with VMware, and then I was last at ZEDEDA, which is a company focused on distributed edge. Out of the same data center I can still run Linux, and now I’m, at Nubix, which is that next click down, micro controllers.
So been around the block when it comes to IoT and Edge.
– [Ryan] Fantastic. Speaking of Edge, I wanted to ask you a question to kick things off here around when we talk about trends in edge computing, we talk a lot about macro trends a lot about lower latency, bandwidth consumption, increased security, privacy, things like that. But beyond those typically cited technical reasons for that drive edge computing forward, what are some of those other, or what are some other, I guess I should just say macro trends that really catch your attention that we need to be paying attention to?
– [Jason] Yeah, I mean, I will qualify. So one thing too, so yeah, very common. People say latency and bandwidth and whatnot, but even latency, there’s nuances. Real time, way overused, real time in, in airbag is like microseconds and it must be deterministic, real time in a building automation use case is like 15 minutes.
You know, that’s considered real time in terms of temperature changes and all that. So first off, there’s a lot of people that cite these technical reasons. If it’s latency sensitive versus critical a big difference. Airbag, latency critical, Netflix, latency sensitive, and so there’s nuances that people don’t get into as much even in that conversation.
So I’d qualify that. For me, I think for Edge, it’s a, another major driver that’s less talked about is this notion of getting control of your data back, back control of your data. We’ve spent the past, 10, 15 years, a lot of folks moving toward public cloud. Investing in resources and centralized cloud locations makes great sense.
In a perfect world, we’d centralize everything in the cloud. It’s convenient. It can be cost effective. When you talk about public cloud that easy button of the centralization and sign up and pay as you grow comes with a price tag and, easy button sounds great until you get the bill. So I think Edge too, the conversation now is as much about not just those technical reasons, but how do I build a multi-cloud strategy. How do I take control of my data more at the edge? Where do I appropriately run workloads? It’s not like the cloud’s going away. There’s click bait about that, but it’s how do I control my environment more?
And the edge is where a lot of my data’s generated, so I need to be thinking about it from that point of generation.
– [Ryan] And is that kind of, one of the, is that when it comes to these macro trends, I imagine that they really are going to affect the approach that technology providers take in the IoT space. And you mentioned I briefly mentioned them there, but how else are they affecting the approach that companies take when it comes to bringing an IoT solution kind of to life?
– [Jason] Yeah, I mean if you look at the trends, so you know, historically you’ve got the, different buckets. You’ve got the IT infrastructure OEMs know, the of both hardware and software. You’ve got the big cloud scalers, so that would be like, Dell, HPE, Lenovo, and then you’ve got VMwares and Red Hats, and then, so they’re, that’s one crew.
And then you’ve got Microsoft and AWS and of course then the telcos. There’s all these different buckets, but when you apply that, what I was saying before around the data gravity moving towards public cloud and now people are wanting to pull back, it’s going to change the dynamics of what people were doing.
Cloud scalers used to buy all their hardware from your Dells and HPEs and then when they get to a, got to a certain scale in the cloud, then they’re like let me just go straight to the source and buy it myself and I’ll spec it myself. I’ll build everything. So predominantly they’re actually driving their own infrastructure and we’re seeing that happen in other places.
That cuts into the business for your traditional, hardware suppliers and whatnot. Now with this shift to edge, you’re starting to see the cloud providers trying to offer edge services, like AWS Outposts and Azure Edge and Google Anthos. But that’s like bringing that cloud experience back out, you know, on-prem, which is now encroaching on the bread and butter of the traditional OEMs hardware OEMs to end customers.
It’s, I call Edge the last battleground for the, or the next battleground for the infrastructure giants. Mean, edge is a hot topic for the technology providers because it’s a battle for customer attention and acquisition and whatnot. So the clouds are trying to bring their easy button back out, but then you get locked into their ecosystem.
The hardware players are bringing easy button out, but then you’re locked into their hardware. The software players, are bringing, easy button out now you’re, I’m agnostic, like a VMware, I’m agnostic to cloud and hardware. Everyone’s got a different, the telcos are trying to figure out where do they run stuff and what do they do.
So it’s a big, gonna be a big driver for scrambling for mindshare and customer attention.
– [Ryan] Yeah. One of the things that’s done that early on is when I joined, got into the IoT space, I paid a lot of attention to was the platform landscape and you know how every platform was clamoring for the attention of the end user and the buyers. Saying that our platform can do everything.
It can be built for, if it was very horizontal, it could be worked and used for any type of solution, et cetera, et cetera. And that’s changed a little bit over the years, but I’m curious from your perspective, how, what are your thoughts on the current IoT platform landscape in IoT?
– [Jason] I think the last thing the world needs is another IoT platform. I think that we have 359 too many. Nah, it’s it’s, I’ve been talking about this kind of stuff for a long time. Any new market, you, it starts usually with the platform raise. You know, the people are like, well I don’t know what to do with it.
What my, what’s my use case? I don’t know. You just buy my platform and you figure it out. It’s very much about that. And then eventually it flips to, the value first. And so I think we’ve seen good progress, and obviously the clouds have there’s what we just saw, Google killed their IoT platform and there’s folks that are coming in to, to help on that front.
But I think that we’re seeing progress on consolidation. Ultimately need more open, trusted infrastructure. And then the real money is made by applying domain knowledge and necessarily unique hardware and software around it. Believer in, in the need for interoperability, which also comes into play against, the clouds versus, the, all the other things we talked about.
Because the cloud model is to lock you in. So it’s which is sticky but that comes with some drawbacks. Yeah, I think it’s consolidating and I think that people are realizing that they need to have a go-to dance move, so to speak, of like where they’re going to make money and not trying to do everything.
– [Ryan] Yeah, it’s been interesting because what we’ve noticed recently is a lot of the platform or solution providers that are trying to pitch as more turnkey option or more complete option I should say feel like an off the shelf buy are becoming a quite more vertical focused and show trying to showcase that that solution is built for this particular industry or this particular problem, and tailoring all the different pieces of that to that particular problem as opposed to the horizontal approach.
And I think that’s what the industry, especially the buyers need, because trying to go through and do your own research and figure out which platform provider should I be working with? Which hardware do I need? What connectivity do I need? It’s a nightmare for companies that really don’t care about all those pieces and they’re really just trying to find a solution that solves their problem. So I think the more companies can help buyers get to that point and help them make better decisions, it’s a win for the industry as a whole, because obviously the more successes we have, the more scale we have, and IoT continues to grow.
– [Jason] Absolutely. Yeah. And I’ve been, we got the IoT business started at Dell, a group of us back in 2014 and following the landscape for some time and even back then, it’s like this is going to go vertical before it goes horizontal. And over time, the hope is that yeah, you have less choices or more consistent choices for the horizontal elements below, like the plumbing and then, people that are actually going to get traction.
And I think still the people that are getting traction are smaller smaller companies are at least very focused on a given outcome versus the peanut butter platform approach. Yeah. We’ll see how it plays out, but it’s definitely in the consolidation path.
– [Ryan] What are your thoughts on AI in, in IoT kinda intersection there, as well as just the general state of what’s going on in the space?
– [Jason] Yeah. I mean, AI, so first off, AI is, so many people we’re starting to get there now and there’s some very interesting slash scary stuff happening with generative ChatGPT is the hot topic now and, where does where does that head when it starts looping on itself and getting, smarter and all that.
But IoT and AI, there’s also a lot of people that say, oh, I do AI, but it’s basically like a fancy rules engine. If this then that, it’s oh, that’s AI.
– [Ryan] Just to say I have AI in my solution.
– [Jason] Oh yeah. But when you look at, IoT is obviously, predominantly about the sensing layer and being able to gather data and, then you can make decisions and act on it. AI applied to it can automate, those elements. Computer vision, both for IoT and, the relationship of sensing computer vision is the killer app for Edge.
Know, computer vision obviously being related to AI. Cameras are some of the best sensors around. The only people that think that it’s a good idea to send raw video data over the internet are people that sell you internet connectivity. Otherwise, that’s process it locally, at the device or, close to it.
And then send events up, you know, trigger, action based on, that context. I think that’s a, when you look at cameras as sensors tied to AI there’s a very tight correlation there. You the generative stuff is a little different. But at the same time, we’ll increasingly see real time telemetry data feeding into the generative AI over time.
It’s definitely becoming a lot more, I think, concrete and we’re seeing a lot of acceleration happening versus, five years ago it was, you know, definitely part of buzzword bingo. It still is, but.
– [Ryan] Yeah. I feel like that happens years before the technology is ready in a lot of these cases for sure. Another thing I wanted to, oh, go ahead.
– [Jason] IoT, I think people are getting fatigued with IoT to some degree just, you know, because it wasn’t like the big bang, but the reality is it’s happening. It’s just putting, bringing sensors and these types of technologies, into solutions is just stuff that you do now. Not that it’s without, it’s not without some challenges, but it’s just stuff you do.
It’s yeah I’ve said for a while, like, how often do I say, Hey, I’m gonna do some e-commerce today. You know that buzzword is gone, but it’s just some part of your daily life.
– [Ryan] But yeah, it’s still, whatever we call it, it almost doesn’t matter. It just does, bringing or using sensors to collect data from the physical world and being able to utilize data that wasn’t accessible. Companies there’s clear value there. It’s just, what do we call it as, and maybe not as important as what it does.
And it’s not if the buzzword was something different then obviously it went from M2M to IoT, so it’s still growing and it’s just becoming, the technology’s becoming better, more sophisticated, I think they are becoming cheaper and they’re seeing more successes out in the world to justify and get over the hump of or any fear of adoption.
– [Jason] Early days when we were doing it getting started with IoT, you need a label, there’s no doubt. But we’d have people coming in, this is when I was at Dell, like basically I wanna buy some IoT. Really? Okay. What color would you like? Check out the shelf.
– [Ryan] That that’s to like my point earlier about the way IoT is sold is changing. And I think purely focusing, obviously, don’t get me wrong, there are companies that have, are sell components of technology, but they’re selling back into the industry in most cases. So that’s a different kind of selling than the scale adopters buyers that are, work in agriculture that have, know very little about technology, but know they need a soil moisture monitoring solution.
They gotta figure out what, how to buy that. And making that easier is what’s gonna allow a lot of these solutions to really take off in adoption to grow. So I wanted to ask you what is and this was something that was sent earlier, but I’m not as familiar with it when the term TinyML, what is that exactly?
And how is that kind of driving I assume more intelligence on, on IoT.
– [Jason] TinyML, speak of another buzz term, so to speak but it’s really the, it’s the notion of compressing machine learning models into very small kind of fixed function models for very constrained devices. An example of TinyML would be when you say, Hey, Alexa, for your Echo, that there’s a little machine learning model within the echo that recognizes that wake word, and that’s when it connects to the internet and everyone’s oh, in theory, oh,
It’s listening to me. No, it’s, it’s basically listening for that wake word. So there’s a TinyML model there. We’re gonna see more and more of these tiny, functions being deployed in these constrained devices. And that’s, so historically on for embedded, I can’t run Linux.
I’m an Arm M-Class microprocessor. Historically for these devices, whether it’s like a smart light bulb or some connected product or whatever, like it’s does one thing for its whole life you do embedded programming for it. You hard code for that device to optimize it because that device is so constrained. This is compared to the data center where you got these huge servers and you can virtualize everything and you got the resources for that overhead. But what’s happening is it as these devices get smarter and smarter and the tool sets get better.
You’re gonna want to dynamically adapt that device software define that device. Just like we’re seeing in cars and all these other products like the day you buy a product is not, the product change will change over time. The value is defined over its whole lifetime. And so TinyML is a representative function that’s gonna start to change how you have to develop these really small devices.
You can’t just bake in a model and never expect to change it. Light switch does the same thing you know all the time. But if you start looking for different voice activated words or computer vision stuff using these tiny little models, you have to be able to update them.
So what we’re doing at Nubix is instead of traditional embedded where you hardcode everything how do you create an abstraction layer above that hardware just like you see in the data center with a, a hypervisor or, Docker on top of Linux, or taking those same principles and applying them to microcontrollers.
You can’t just drop Docker on an MCU. It’s it’s way too heavy but you get the benefits of that compartmentalization, the containerization. I can drop a TinyML model on a little function from, for security, a little something for, connectivity to, my cloud or whatever, or to the sensor below.
All of those functions that we would do further upstream can be done on microcontrollers with our solution. And the benefit is especially key. There’s, there’s time to market. You don’t have to do all this embedded coding. But also there are a lot more people with skillsets around cloud and IoT and AI development and things like that kind of coming down from the cloud and the data center than there are with embedded skill.
So the by containerizing the hardware, you’re actually enabling people that don’t have embedded experience to develop new functions. And TinyML’s a, a great example of it. And it also matters when if I’m building a connected product and I’m pulling in code from different people within my company and like partners and whatnot, I can keep that intellectual property separated.
Today with embedded, if it’s all compiled together, everyone’s got to give me their source code and compile it into a common image. And your IP is exposed to everybody. So it’s all stuff that we take for granted in the data centers is how do we get it from the gigabytes and terabytes that you would have kind of memory utilization that you have available in the data center to kilobytes. It’s a big difference.
– [Ryan] Yeah, absolutely. That’s all great stuff. How do you think about holistically, like how do you, or how do you talk about with people or advise people to be thinking about IoT and edge landscape just as a whole?
– [Jason] We’ve talked about, some of this as we went I think it’s really important. A) Think holistically. Don’t reinvent the plumbing, adopt things that, that, you know as standards emerge, adopt those, the value’s gonna be around the wheel. Think about as much as you can, the clouds are offering great services, but as much as you can decouple your edge investments from the cloud as close as possible to the source.
Yeah. So use APIs that bridge, to, to the, to different resources that are as agnostic as possible. So just watch for lock in. Ultimately, don’t think of just Edge as like a regional data center or online factory floor. Increasingly it’s gonna be, these really tiny devices everywhere that are doing more processing.
So think holistically. And ultimately it’s just about if you do it right, there’s, you have these different paradigms from data center to outside the data center, but can still run Linux to these microcontroller based devices. Can’t be the same platforms or the exact same platforms as technology, but if you do it right, you can stitch stuff together and where you run applications across that continuum just is a balance of, performance and cost.
So it’s baby steps.
– [Ryan] Yeah. No, totally agree. Last thing I wanted to ask you before I let you go here is just from your all’s perspective and what are some of the biggest challenges you all come across or have seen, when it comes to working with customers or the industry as a whole. You can kind of take it from any way you like, but just some challenges that are worth noting and thinking through for our audience out there.
– [Jason] I’ll start by saying the biggest challenges with technology have nothing to do with technology. It’s people and business case, which kind of ties back to people. You with IoT it’s, don’t have the skills. That’s not my job, that, that freaks me out, like whatever. But when there’s value, things tend to progress.
So we’re, we’ve definitely found more and more examples there. Just the fragmentation, out there is, has been a mess. But that’s getting better. As we talked about the platform landscape the edge washing has been a problem. People like, talk about Edge as like one thing versus these different paradigms with inherently different trade offs.
I think all of these different things and, with Edge now it’s vendors trying to push their platform ahead of what’s the actual use case and all that. It just, it confuses end customers and it becomes, I think, challenging but what we always tend to, this always tends to happen early on in any kind of new tech trend, and then you work through it and the value rises to the top.
– [Ryan] Yeah, it’s it’s interesting when I talk to guests about the challenges in the space. Like fragmentation in the industry is a big one. Justification of kind of business cases and things like that always come up of, getting buy-in internally. Lots of things. It just seems like there’s so many, but at the same time, I feel like a lot of the companies that you have the option, opportunity to work with they’ve experienced this and they’ve thought through a lot of these things before, and that’s a really important part about, picking a company to kind of engage with when it comes to building and adopting a solution. Jason, this has been fantastic. Thank you so much for taking the time. For audience who wants to learn more about the company, follow up with this conversation, ask any questions, anything like that, what’s the best way they can do that?
– [Jason] Yeah, reach out to us. We’re on LinkedIn info@nubix.io. You can email us directly and yeah, I’d love if you’re, doing embedded or you’re interested in extending kind of your cloud prowess down to the tiniest little devices out there that’s our thing.
– [Ryan] Fantastic. Jason, thank you again for taking the time. I really appreciate it. I think audience is going to get a ton of value out of it and so thanks for being here.
– [Jason] Yeah. Thank you so much for having me.