Conversation with Flavio Bonomi
Good day, and welcome to edition 139 of our Digital Industry Leadership Series, today it is truly a great pleasure to welcome Dr. Flavio Bonomi, the founder, and former CEO and CTO of Nebbiolo Technologies, a Silicon Valley startup delivering the first complete Fog/Edge Computing Software Platform for the industrial automation market. Dr. Bonomi is a technology expert and visionary with experience that spans from low-level devices in silicon, to the broad level of networking in computer systems. He spent 14 years at Cisco Systems leading advanced R&D as a Cisco fellow. While there he also literally coined the term ‘Fog Computing’.
Prior to Cisco, Flavio worked at several Bay Area startups and was at AT&T Bell Labs where he was a distinguished member of the technical staff. He received a Ph.D. in Electrical Engineering from Cornell University, as well as a Bachelor of Science in Electrical Engineering from the University of Pavia, Italy. A consummate learner he has published over 100 papers in technical journals and conference proceedings and is co-inventor on – and I’m not kidding here – 60+ US and International patents.
Flavio, welcome to our Digital Industry Leadership Podcast.
Thank you, Ken, it’s a pleasure to be with you on these topics that I am very passionate about.
Yes, yes passion clearly defines you doesn’t it Flavio? Especially when it comes to Fog Computing which will really be the clear topic today.
I always like to start with one’s digital industry journey. What would you consider to be the red thread of your digital leadership journey?
I think it’s the IT-OT convergence, a sentence that is used a lot, but in my professional journey, I learned a lot about the IT world with Cisco, with AT&T Communications, computing, cloud, Soft Identify Networking, all those things. But at the beginning of my journey was actually systems theory & control theory. So now, the opportunity to bring all the learning from IT to the OT space, the operational space, that touches the ground, that touches the machines, the automotive, the cars, the robots, has been very motivating for me. And so, my red thread is that kind of sharing of cultures and sharing of technologies between these two worlds, from which I believe that one of the great values of these digital transformations comes.
Clearly, and also why you are literally having the right skills, at the right place, at the right time, given that we still talk about OT-IT convergence, but we’re probably much closer to it than we ever have been before. Certainly, Fog Computing, or Edge computing as others will call it, is a paradigm that well-describes that.
Let’s go back to some of your early careers; you got a Ph.D. at Cornell, 10 years at Bell Labs, so a typical under-achiever start there, but you jumped into back-to-back networking startups, and this was of course in the absolute heyday of networking. Interestingly enough both were acquired, the latter by Cisco which does explain how you ended up there. But what were some of your early wins and lessons learned during this time?
Well that was a very rich time in Silicon Valley at the time of the big boom, the networking, and the internet explosion, and I was coming from Bell Labs Networking, ATM, the technology was ATM, and there was a chance to apply to enterprise worlds rather than Telco worlds, and so I felt that Silicon Valley was the place for this because they were becoming interested in the same technologies that AT&T had been pushing for a while. So, getting in there was really the dynamics of small companies, sometimes quick and dirty prototypes, small little things that would really compete very well with heavy-duty designs from large companies like AT&T. So, the agility was one of the things that I learned.
And then it was the explosion of the enterprise world and the beginning of data centers and clouds, in which I found myself in the middle. And in fact, it was the convergence of the IP networking and Telco networking that was happening at that time. We were at the forefront of that with SONET coming in, together with the IP networking that Cisco was pushing, there was the big explosion of networking and it was fun to be with new people also coming from very different backgrounds, sometimes computer design, large chip design, and software. It created great friendships, some of which were taken into the next journey, in fact, some of the people I met early, Chandra Joshi, and Kannan Devarajan, and others were with me in these early startup’s creating the network of people that trust each other, that walk through multiple startups in their professional life, and this is what was happening here.
The networking, the humanity, the convergence of so many cultures was great. And then also the money at the end of the journey because it was really high time for acquisitions, and you felt at this point that the world was yours and there was no downside of it; retirement was near, and then you realize that things are not always shiny. Then came the 2001 crash and so-forth, so ups and downs of life, history, and professional lives as well.
And Silicon Valley!
Silicon Valley, exactly.
You talk about a company that did acquisitions extremely well, but also at a high velocity, Cisco in the day, and of course, they acquired your latter startup. You spent 14 years there as a distinguished engineer and later fellow, and I think at the time I mentioned VP and Head of Advance Architecture & Research, what a great time to be there with Mr. Chambers and the Internet of Everything, and of course the black boxes that really defined very much the internet for many of us. What were some of the highlights of your time there?
The company was really full of energy and enthusiasm, winning in everything they were doing, but also was bringing together talents and organization innovation in many, many ways. It was great because it could offer you multiple areas of growth. The first was the big routers, internet routers, GSRs, CRS, the beginning of the big designs there. Then I moved to the data center and storage with Mario Mazzola and Luca Cafieroand the famous MPLS Group, they were doing enterprise networking and then at the beginning of Datacenter Networking and Data Centre Computing. There was firstly the WAN and then the LAN with very different dimension costs and ways of doing things, and then again trying to bring both together, because they’re from the big routers and the switches, routers, and switches became again a convergence point, and that is where I always live at the boundary between domains. So that was very good.
And under Mario and Luca there was a lot of openness to do a new thing, and I really created this applied research group under Tom Edsall at that point, and he was amazing because we were acquiring companies into that organization as well as starting to work with research, Stamford Santiago, UC Santiago and there was a time when I had these top-top networking professors coming in on Friday for these brainstorming sessions, Nick McKeown the father of Software-Defined Networking, Scott Shenker, George Varghese, Balaji Prabhakar, these are big names in network research, and we were trying to put them into silicon and into real products.
It was like applied research but going into concrete products directly. It was really great fun, and that group became the core of the Advanced Architecture & Research team that was basically created around my 15 people there, and it became 30 people or more with good funding, and that was the greatest fun because we could give money to universities, and then to the startup’s doing prototyping, proof of concepts and we touched all the new trends happening in technology at the forefront of cloud, at the forefront of Software-Defined Networking, and the way to do security, deep packet inspection, and then came to the early ideas of internet of things, industrial internet with connected vehicle efforts, which was fantastic; going to Formula One races with McLaren talking about the future of the automotive, which is in fact happening now. It was great fun because it was the convergence of again my passion for the car, there was a mechanical object at the time becoming an electronic object, and so electronics and automotive were coming together, again another boundary, another transition, and much fun.
That was what really brought the idea of Fog Computing to the fore because we realized this car needs a lot of computing inside and outside, and the networking inside the car is changing, and it goes towards the internet – the internet wins everywhere. And then there’s the cellular connectivity or the wireless car to infrastructure. So, those were fantastic early movements that prelude to all we saw in the industrial internet, that was 2010 or so, and we realized we need a new way of computing, which we are learning from the cloud, but at the Edge. And that is when we started the stories of Fog Computing.
And I believe you are officially credited with having actually coined that term, and I believe you told me the story once, and it had something to do with the Monterey Aquarium as I remember it. But the end result is I believe the license plate on your Maserati says Cisco Fog if I’m not incorrect there?
Yes, let me correct you a little bit here!
Because it was at the Monterey Research Centre in Moss Landing. I gave a talk, and the talk was about extracting data from the bottom of the Ocean, maybe processing those robots at the bottom of the Ocean, and then bringing it up to the data centers. There was a lady at Cisco before – the wife of one of my employees in that organization. When we finished outside, and I have the picture of that day, 2010; she said, ‘Flavio, stop talking about Cloud Computing, you should call it Fog Computing, because it’s close to the ground, it’s like clouds down to the ground’. She told me that and I complained, ‘Come on Jeanie’, I didn’t believe it. Then I got in my Maserati, which still didn’t have that number plate on, and I thought, ‘Ah, Jeanie is right, so let’s try to use this. Then I kind of became the user of the name, but the origin was her, Jeanie. Then when I left Cisco, my team gave me the vanity plate with Cisco Fog on, and until Fog started dissipating into Edge, but that was fun. That was fun because people enjoyed understanding technology through the silly naming and wines, also grapes, and other stuff, smiling about it. It was fun.
Yeah, an absolutely brilliant name, I still think it’s as applicable as Edge, but the industry certainly seems to like the Edge connotation there.
Speaking of Fog, which I think actually is still a very good term, although the industry seems to have adopted Edge computing a lot, in 2015 you founded Nebbiolo of course, a great name but also representative of Italian Fog per se; a Silicon Valley start-up delivering the first complete Fog Edge Computing software platform for the industrial automation market. This one warms my heart because I do remember some of our early discussions both in the basement of your home, testing some of the hypotheses of this, but of course also enjoying your great espressos and cappuccinos. What compelled you to start this company, and to focus on the industrial automation space?
Well, first I believe that this was a future trend of notice, of relevance for the movement into industrial IoT. I was very convinced that this was a big trend, and I didn’t see much happening, in fact even at Cisco after we proposed it, and then outside I waited a year or so trying other things before getting the courage to start the company around this topic.
It took a while to gather the courage to do it because it implied a complete system, software, hardware, platform, infrastructure, things that were not very fashionable in the VC space at that time. So, it took me a while to decide to do it, but then since nobody was doing it, I decided to do it. That is when we collected the people and started talking to investors, most investors in Europe, because again here the receptivity was not there. That was exciting because starting something from the ground up, with your old friends that you recall back to action with you, all the things that you need to do to start a company, it’s learning every day and it's scary at some point because there’s nothing that you have done before that is exactly helping you, so you have to really learn, be flexible, be agile and fearless in many ways. In fact, fearless is the word in that it’s good that you don’t know how hard it is to start a start-up, otherwise you would never start.
At the same time, if you knew how building a family is not just the first moment of love and infatuation, you would sometimes not build families. But then in the end the growth and the experiences are priceless. So, I started with that courage knowing that this was the right thing for the future and seeing no movement in the area, so I decided to let’s do it, and we got into this. But again, what I said is that the funding had to come from Europe because they were more open to an investment of this type, which included at that point hardware, infrastructure software, and even some elements of applications. It was a very broad initiative that would have been much more appropriate for larger companies than start-ups. And probably you haven’t told me that in your wisdom, classifying what I was trying to do as maybe too broad or too ambitious. But anyway, we did it, we got into this.
Yeah, clearly a full stack is the way I’d like to refer to it, and it’s interesting you really were a pioneer in an early space at that time, at a pretty nascent time. Fast-forward now and you would think these types of investments are much easier, because both strategic investors and institutional investors are more used to thinking about things, but you were really at a formative time my friend.
Yeah, good, and bad!
What were some of the key use cases and wins for you at Nebbiolo?
Firstly, as I said before, the use case that I had in mind was more related to transportation than industrial. I had the car in mind and other use cases like smart cities, related transportation, energy; industrial were one step removed, but I learned of the potential of Fog and Edge for industrial from Kinoshita, Kino from Japan, an investor from Japan and former Cisco person, because he told me this is very important for industrial. So, I knew there were applicabilities there, but it was not the first objective. Then when the money came from TTTech and KUKA, that was also a very exciting time because TTTech – Georg Kopetz brought in KUKA as the other partner for this adventure, and KUKA was at the top of robotics at the time. Till Reuter was a very charismatic CEO; we met in Kitzbuhel as the Downhill Races were taking place, so from the Audi rented building we were looking at the big Downhill Races going on and talking about this new Nebbiolo company. It was fun, and the idea was to apply to one vertical first, to be focused at least on the vertical, target vertical, which was industrial automation, at that time the beginning of industry for Industry 4.0 was the talk.
It was exciting and it was important for me to put the focus on what could Fog, and Edge Computing do in industrial automation. It was early 2015, again pretty early in the picture but the ideas came very clearly, and they are still valid and are still being pursued, and in fact, they are the future in terms of one of the perspectives of digital transformation for the industrial. That’s what brought us to industrial automation, it was the investors coming from that space, and don’t think it was wrong, and there’s talk about the difficulties of that domain, but I think it was a good direction.
Yeah, clearly, as we fast-forward from 2015 to just recently, you actually exited the company to TTTech which I think was a nice home, long-term for this. Now that you’re able to sit back and think about child-rearing of your own start-up in some sense, what were some of the lessons learned now that you’ve gone everywhere, from doing the start-up to full exit on this, i.e., what would you have done differently if you don’t mind me asking it that way?
First, I think it was a very good focus to take, and it was important to take the time to really learn the pain points of the industry with people in that space. It became very clear from the beginning that there was so much we could really do in this convergence of technologies because, despite the cars and flashy robots and so-on, the electronic infrastructure side of industrial automation was very much in need of innovation. And so, it was very important that the use cases became clear from the beginning, from the consolidation of these industrial PCs to the analytics at the edge, and to this evolution of the automatic control of industrial automation. The use cases became very clear and that was one of the important learnings that are still there, that is still to be fully manifested, the potential that is there.
On the other hand, the other learning is that this world was so distant from the IT world, the gap was so broad, in a sense attractive because there’s a vacuum there to feel, but on the other hand, there’s a cultural shift that is dramatic for people involved. It became very clear that it was not an immediate adoption, it was actually painful because this world is conservative and wants to test things. And so that was the learning of the purgatory of IoT, that is the proof of concept, pilot time, where you run around spending your equity money to travel, to teach, and to show a platform at work that maybe is not fully mature, so it cannot be really deployed at scale, so cannot really bring in the revenue. That process of learning from each other was very good but is also costly, because when you really have to show as a startup that you have revenue then you realize it’s a long pull, it’s a bigger challenge.
So, the challenge is, to make money out of this transition, and these new technologies, for a startup, is not easy, and when you come down to that bottom line, that was painful. But still, we made important breakthroughs with companies like Hyundai, and the Toshiba Group, and various other big names. With Audi, we had fantastic proof of concepts that are still now referenced. So, we did an amazing job in breaking through the market as a little company, but it was not easy and it’s not easy for startups in this domain and we see that. Then the pandemic came as well to make things harder.
The learning is, this is a very good topic that will continue to mature and explode I think, so it’s a good area; on the other hand, it took a long time to mature than we expected, although it is a fast transition. If you think about it, five years is not long for this vertical, and in fact, so much acceptance is there and it’s really a revolution. It’s too slow for a startup in some ways, but it's fast for the industry. So, we’re still there.
Well said Flavio, and I go back to your intersection of IT and OT, you’ve sat at enterprise, IT on the Cisco side and other companies, and certainly on the OT side – I’ll call it enterprise OT side, for the last several years. That is the part that I think most IT vendors, analysts, and that community tried to judge progress on the OT side by what they’ve seen on the IT side, i.e., virtualization of IT over the last decade or two, right; OT should virtualize just as quickly. And I think the old adage always applies because I’ve lived on both sides as you know, a bug in software on the IT side and people are inconvenienced, a bug in software on the OT side, and people are harmed or worse, right?
And so, it’s the understanding and appreciation of that, that any startup that’s working in this IT-OT kind of convergence space we’re always very much cautioning about and making sure as you say you have early revenue opportunities that can grow in scale into long-term ones. And so yes, you have lived that in spades my friend!
Yeah, I know, I have a lot of scars on my back as they say.
Given your early predictions of Fog computing, what do you see is the next decade for this Edge-Fog pattern that you predicted so early? Especially in the digital industry space.
I’m very confident and convinced that what we saw… and this is my problem, my wife also tells me, ‘You see the future 100 percent right, but you don’t know when the future will come,’ and so the timing is the question. But the direction is very clear and so I’m very comfortable to predict now, that what we talked about is really what is progressively happening, that is virtualization will come into the OT space in its definition, that is more mission-critical than it is probably typical of IT. That will open the door to a new software deployment model, managed, containerized, based on virtualization what they called cloud-native. So, we will have a model of distribution of software that now goes from the inner machine, embedded to the Edge, to the cloud; so, you can take the same function and move it wherever you want.
This is really distributed computing at its best, this continuum is happening, enabled by virtualization that has to be very lightweight and mission-critical as you go down. That will decouple and is decoupling the software and hardware. In the old world, you would buy a box to deploy an application, and then you have a fragmentation of boxes around the floor that is very difficult to manage, secure, upgrade, and so forth.
So now there’s an infrastructure like in the cloud, and there are applications that can go over the infrastructure, and now it becomes a world of apps, the app store of industrial. So, now we’re seeing the same things that happened with the phone, and that will unleash great innovation, very dynamic and agile because you don’t need to put down a box to test a new analytics algorithm, you can give a container to the operator and say, ‘Test it out' and the guy clicks and brings it down. This also leads to all these Edge AI, ML, and digital twins, this is one of the exciting things of the future – can also be distributed from the Edge -up, so you can have this hierarchical machine learning, digital twins, and analytics that will again make sense of the data, and next goes into action because extracting insight from the data is good, but you have to also close the loop.
This leads to the important area of the evolution of control, which has really been my passion, seeing control becoming Software-Defined, for example, the PLC becoming a software virtual PLC, that is a big disruptive and dramatic change. So, Software-Defined Control will become again deployed as applications; you need a new robotic controller, okay I’ll give you the container that does it, instead of the box and the whole thing, but obviously, the infrastructure has to be ready for that. So now you have this continuum of softer models and intelligence from the cloud all the way down, and that’s the dream of IT and OT convergence, and it's coming.
I don’t know if you have seen it, there’s a beautiful video by Henning Loser of Audi that I probably didn’t share with you, but I’ll share it with you – it’s not public, but he is presenting the future of the automotive factory exactly in these terms. And it’s nice to see because we talked to Audi earlier about this vision, and now it’s what they’re going for. So, this is happening, there’s a lot of work still to do but that’s why we’re here, and we are excited about more work ahead of us on this.
And that begs the obvious question; now with Nebbiolo safely tucked away with TTTech, what’s next for Flavio?
Well, the dream has not been completely manifested. The dream we started within 2015, the Fog dream is maturing but is not fully manifested, that’s why I’m active with the Lynx Software – which by the way hosted the Nebbiolo in 2015 for six months because we wanted to use their mission-critical defense typical embedded hypervisor unique that can be certified for safety, it’s very very secure, so that’s why we went to Lynx in the beginning. That’s why I’m still working as an advisor to the Board of Lynx Software because we need that layer to close this gap towards the Edge; virtualization to the deep Edge, to the first mile or the last mile, requires mission-critical qualities, lightness coming from an embedded type of use into Edge. So embedded going to Edge is that area of activity that I’m involved in now that we wanted to achieve with Nebbiolo, but then it was too much to try to attack that area, the real deep mission-critical virtualization.
So, I’m doing it with Lynx and other companies that are working towards this, and to really manifest as with manifest the virtualization of control, to see that it's feasible, and Intel is also moving in the same direction. So, I’m at the last mile still working in the same direction, helping companies in a consulting fashion, you teach that that’s a good area and you’re succeeding in that, and we are all looking at the great things you did at Momenta in that direction. So, I’m active on that front with a number of companies. I see big areas still open which is the programming of this cyber-physical system, of this IoT, industrial IoT systems. It’s still an open question.
We’re working on the infrastructure, the infrastructure in coming, but then how do we really enable and expose the deployment of applications? You really need a way to program these applications, build these applications, and I think that software has to evolve, software that really responds to the needs of mission-critical systems, cyber-physical system – the automotive autonomy, the robotics, the control, and all that space is still fragmented. So, beyond the infrastructure is the infrastructure for the software that can still make progress. I’m active in that area, you’ll probably hear from me in that space soon.
So, I’m open and excited because I see the industry going in the right direction with more enthusiasm now that we’ve come out of this difficult year, or two years. So, hope is coming back for everybody and for me too as a technologist. So, courage, investment, also there’s this government investment in infrastructure that should give a lot more to areas like transportation, manufacturing, advanced manufacturing, robotics, autonomous vehicles, and so on. I think there’s a lot to do ahead. I’m excited, open – we’ll see, we’ll keep in touch.
Fully agreed, and again you are clearly somebody with the right experience space at the right time, and perhaps in the right place. I love this idea of Software-Defined Control, I think we could do a separate podcast just on that, but unfortunately, we’re running really short on time. So, Flavio thank you for spending this time with us today.
It was fun! I am very fond of you and what you’re doing, and this time was inspiring. I hope that the audience also got something out of our chat here. Thank you for the good questions, and best wishes to everybody in this new resurrection of the world.
Well said there Flavio, and the feelings are mutual.
So, this has been Dr. Flavio Benami, founder, and former CEO and CTO of Nebbiolo Technologies, and if I can get away with it, aka The Fog Man. Thank you for listening and please join us next week for the next episode of our Digital Industry Leadership Series. Thank you and have a great day.