Utilizing Tech - Season 7: AI Data Infrastructure Presented by Solidigm - 05x01: Commoditization of Edge Infrastructure with ZEDEDA

Episode Date: May 1, 2023

Edge environments were historically very specialized, but virtualization and cloud technology is enabling companies to deploy commodity platforms at the edge. This episode of Utilizing Edge features Raghu Vatte of ZEDEDA discussing this commoditization with Alastair Cooke and Stephen Foskett. Although the transition is still getting started, standard compute platforms are rapidly being exploited at edge locations, from warehouses to retail to industrial. Even is some specialized hardware is still needed, a unified platform can increasingly absorb a majority of applications at the edge. Another factor contributing to commoditization is the standardization of application requirements, with most now virtualized or containerized with standards developing for I/O and shared hardware resources. Hosts: Stephen Foskett: https://www.twitter.com/SFoskett Alastair Cooke: https://www.twitter.com/DemitasseNZ ZEDEDA Representative: Raghu Vatte, VP of Product Management and Customer Success, ZEDEDA: https://www.linkedin.com/in/raghushankar-vatte-970a9214/ Follow Gestalt IT and Utilizing TechWebsite: https://www.UtilizingTech.com/Website: https://www.GestaltIT.com/Twitter: https://www.twitter.com/GestaltITLinkedIn: https://www.linkedin.com/company/gestalt-it/ Tags: #UtilizingEdge #Commoditization #EdgeComputing #EdgeInfrastructure @UtilizingTech @GestaltIT @ZededaEdge

Welcome to Utilizing Tech, the podcast about emerging technology from Gestalt IT. This season of Utilizing Tech focuses on edge computing, which demands a new approach to compute, storage, networking, and more. I'm your host, Stephen Foskett, organizer of Tech Field Day and publisher of Gestalt IT. Joining me today as my co-host is Alistair Cook. Thanks, Stephen. I'm a blogger and a trainer and an analyst, and I have a side interest in microcontrollers. I've been working with embedded systems since it was 8-bit microcontrollers and clock speeds in the kilohertz. So I'm interested to see just how much things have changed in the hardware that we're deploying out to solve these edge solutions. Yeah, that's a real good point, Alistair. You and I are both data center veterans

and have experienced edge retail environments. And it used to be that these things were a total mess. It was all a bunch of proprietary hardware, specialized systems. You had to buy different things from different vendors. Even the backhaul, the network, the management, everything was completely proprietary. Yeah, nobody trusted anybody else. There was no consistency. There was no underlying platform that was used across it. When you needed to provide

Wi-Fi for your customers, it was a whole other connection coming in. When you bought your inventory control system, it was a new appliance, really, that just got shoved into a rack, and you bought your inventory control system, and it was a new appliance, really, that just got shoved into a rack, and you paid ridiculous amounts of money for these things. And they didn't provide any kind of advanced services or anything either. I mean, you know, failover, things like that. They just were unheard of. A lot of the stuff had to go back to the central office. And, you know, I think that a lot of that is changing.

As we saw at Edge Field Day, and as we are going to see at the next Edge Field Day event. A lot of people are standardizing, commoditizing. In a way, it reminds me a little of cloud and a little of the data center, but with a real edge twist. And that's why we've invited Raghu Bhate, VP of Product Management and Customer Success for Zadida to join us today. Welcome to the show, Raghu. Hey, VP of Product Management and Customer Success for Zededa to join us today. Welcome to the show, Raghu.

Hey, welcome, Stephen. How are you doing? Great. It's great to have you here on episode one of this season of Utilizing Tech. We'll call it Utilizing Edge. We saw you present at Edge Field Day? And I think it's really interesting because the goal of Zededa seems to be to build commodity hardware and a standardized platform for running applications at the Edge. Is that right? Yes. So Zededa does not actually build hardware, but what we wanted to do was commoditize how people can consume edge, again, how people

consume edge, how people develop edge. And this is the same theme that we have seen earlier, both in the data centers and in the cloud, right? These are very specialized, high skilled kind of environments. But once you actually have abstracted out the high-skilled nature or the specialized nature of it, then the adaption of those environments became like exploded. The same thing with the data center, where people started actually consuming it very high scale

and also the standardized way of developing things on data center. And now take a few more years afterwards and the same exact trend we saw happening on the cloud, where now the hyperscalers were able to provide this very, very specialized kind of services. But for the people who are consuming it, it was a very standard, very easy way to consume. And more importantly, for people who are developing it, it was a very standard way of developing. And this created that explosion of use of both the data centers and the cloud, and also the services that could be offered on top of it. We currently see a very similar trend happening in the edge okay where it is a very special uh beast in itself right like it is definitely for all the reasons that you are

talking about everything right from the hardware to all the services and the applications and how people consume it it's a very different thing than cloud and the data center but once those specialized functions are abstracted out and it is provided as a platform where where this platform is now consumed by the customers or the end users in the same way that they would consume a data center or the cloud and develop applications for it then you can use the same workforce to do this, do consume edge. Now, by doing this, what we feel at Zadida is that you can bring a lot of services, a lot of applications to the edge much quicker, you can bring them in a way that is economical

and that makes it very, very easy for the end user to use. And that is what I feel about the commodity side of the picture, right from hardware all the way to the services. I think there's a really important part in here of seeing a sort of unified platform that instead of the disparate platforms that Stephen and I were talking about in the older edge solutions is that what you're bringing is a platform on which you can deploy a variety of things, but which delivers a whole collection of services towards making it easy to deploy those applications.

Is this a trend you're seeing with the customers who are actually buying the platforms? Are they tending to standardize on this one platform at their edge locations? Or are they still being left with a lot of legacy architecture that they can't necessarily bring across um we actually see a very high desire of like uh um of standardizing on one unified platform right uh but the reality is the reality so they uh the customers have actually the end users have actually already invested um in a lot of platforms and these things are going to stay uh they are not going to be yanked out in in a day just because there is a new shiny thing uh that people can get their hands on right so the trick and also the right thing to do is to be able to give a path out of the current platforms

or integrate into those platforms so that you are not going to incur a very high cost of moving from one to another. But providing that path is the most important thing because there are applications and there are services that are being offered by the current existing platforms that are the bread and butter of these companies, right? No one is going to take the risk of like saying, hey, I'm going to young this whole thing out, put a new thing in, because this is an unknown commodity at the end of the day.

And by providing that seamless path of like step-by-step, small steps, what you can do is like you can take the desire or you can fulfill the desire of saying that, okay, I'm currently operating in one way, but the other way of the unified platform that can now help me implement my edge is the right way. But how do I go from one to another? And doing those small steps, incremental steps is the most important thing. But the desire itself is clearly there

and it is almost like a stated goal in every single one that we talk to but i think it's it's important to recognize the dirty secret of the edge which is that it's not all docker containers and newly built applications there's an awful lot of legacy windows applications that are still running out there and you need the unified platform really does need to support those as well. Oh yeah, definitely, right? And this is what I was talking about, the bread and butter. There are so many applications

right from your SCADA applications to the homegrown things that are running. And again, I think like we are all dating ourselves by talking about Windows. I hear people coming to me and telling, hey, my legacy Linux VM, right? Like, okay, so like that has already become a legacy. Like, okay, and like my legacy Docker. I don't even want to go into that picture.

But the thing is like talking about Windows, Linux, and Docker, and newer runtimes, all of them together is a norm right now okay I have not been to a single call where people are talk coming to us and saying hey we have this huge fleet of Windows applications we want you to like help us manage it right that is not what we are seeing what we are seeing is hey we have this huge fleet of Windows applications or Linux applications that we are we depend on. But there is a digitization that we want to do and we want to go from A to B. And B is very clearly a containerized world. And it is not just about the form factor of the application, but what you do with it how you interact with it right so those that is that is the place they want to go and they don't want to go in one shot most importantly they want to be

they want to have that insurance of like i want to keep running this i want to add these sidecar applications that will provide this value on the existing one. And slowly I want to move from my older generation applications to a new generation applications. And that is what we see. And I think that is the right way of going, because these applications have been built to run for like tens of years. Right. And now you are not going to go and yank them out.

And it is not the right thing to yank them out. So gradual that is that is what we see. And if you're going to be reinventing anyway, you might as well reinvent on modern infrastructure and standardization and virtualization and containers and orchestration and so on, instead of deploying more, I guess, future technical debt. One of the things I want to bring up as well is, and we're going to be bringing this up on every episode this season, is edge and edge and edge. There's different edges and there's different places that we're talking. Raghu, I'd love to get your perspective on this question. So when you are looking at people deploying a standardized commodity, commoditized infrastructure for running applications, is that more prevalent in,

for example, retail or industrial IoT or all the other locations, mining, transportation, food service, which areas have different needs? So there is a very clear desire to commoditize the hardware, right? It's happening for multiple reasons. And if we go back in time, not far over, like five, six years, I go to applications for built for hardware. So people would choose a hardware and say like, okay, this is the hardware, these are the specs that I want, everything from environmental to ruggedized to like compute to IO and everything else, right? In security too, they would say, okay, this is what I have. And like,

I want to build applications for this. That has changed. That has changed because of multiple reasons, but it has changed basically because it's very easy to produce a level of compute nowadays that is cheap. But more importantly, it's also reliable to a certain extent. Again, there are still very large sectors which have to deal with the specialized pieces of equipment, right? That is not going away. But equally large areas of the edge are now taking a step back and say, do I really need to spend so much on this specialized equipment? Can I not just have a commodity hardware and put that resiliency and all the other intelligence into the software layer rather than into the hardware layer, right? So that is what is happening. We see that quite a bit.

So, but I agree with you, but is that more prevalent in, for example, retail or food service than in manufacturing? Are there different verticals that are more willing to go with commodity platforms? So I would say like the retail sectors, the manufacturing sectors, the energy sectors, all of these sectors are trying to, or verticals are trying to go

to the commodity hardware, right? They are trying to go to the commodity hardware, right? They are trying to go to the commodity hardware. I have seen people who used to run real-time operating systems on like a very, very specialized equipment in industrial space, telling me that like, I want to like take away that specialized equipment and I want to run this real-time operating system on a commodity hardware, okay? You name it, right? That is one thing that is happening. The second, a very important thing that happened, at least in the past few years, is just the supply chain issues. Because of the supply chain issues what happened is like uh all these industries really got a jolt right when uh you have a service to deploy you want to deploy that service and now

when your service is dependent on a power supply not being available in a supply chain somewhere and your whole fleet not being able to deploy now people are taking a very deep look at like their actual service chain. And then they're saying, if I am dependent on these kinds of things, then like I am going to like have a risk in my service deployment. And now they are forcing everyone up the stack or their vendors saying that I don't want to be tied to a single hardware platform. Neither am I tied to a single architecture. I want to have the ability to pick and choose multiple different types of hardware, but I want you to make sure that your applications would be able to run on all of this. Give me the minimum viable hardware that

you want so that your applications can run, and I will source it from five different places. And once you do that, you should be able to run these applications. So what that makes, it's a very big change in my mind, right? What it makes is like, it creates this barrier of entry, or it lowers that barrier of entry for this hardware hardware vendors and now a lot more people can join that group and because there is a lot more people that join the group there is more innovation there are more different varieties of hardware that are available you can get something that is available where the parts are available for 10 years and cost X amount

of dollars versus the same similar kind of a hardware where where the parts are available for 10 years and cost x amount of dollars versus the same similar kind of a hardware where now the parts are only guaranteed for three years and it costs y right now depending on your use case and depending on what you want to deliver you can choose x or y now that is a very big c change and also the fact that these both x and y are these hardwares are also very reliable to begin with. Now, we are not talking about a reliability of 70 percent versus 90 percent. We are talking about a reliability of 90 percent versus 95 percent. So you have already crossed that one back.

I think there's been quite a change in what constitutes commodity hardware as well. I've been seeing a trend that x86 CPUs, which is fundamentally what's driven a lot of our virtualization, are turning up in things that don't look like a PC. Even from mainstream vendors, we're seeing smaller and smaller things. And at Edge Field Day 1, we talked a lot about Intel NUCs as being a really small form factor. But there's a lot of more of the industrial type small form factors, the fanless, big heat sinks, lots of IO ports on them that still come under the category of being commodity. Those sorts of things would not have been interchangeable 10 years ago. And so I think that's where Raghu's point about that supply chain risk is being reduced, because you can simply switch out from the device that's unavailable to another near-equivalent device very much more easily. Neither of these devices are things that,

as a normal IT consumer, you'd expect to see on your desk. But they are still replaceable, interchangeable, and that whole idea of being commodity hardware. It's gotten a much wider range of what is commodity than what used to be. Yeah, I completely agree with you, Alistair. The other portion that makes this hardware commodity are the applications that are running on top of it, right? At the end of the day, the way you develop those applications make this a commodity if you are tying your application very very closely to that particular type of hardware or that particular model then it is very hard to make it

commoditize but in the trends and especially and we you have to like go back to the cloud and how cloud is being developed uh or the applications and how cloud is being developed or applications for the cloud are being developed right like you don't really know what hardware is underneath so everyone is developing for whether it's a virtual machine or a container or some other form factor right what you're developing is based on a framework that has nothing to do with the hardware underneath and on edge it's a little bit more trickier, because you're not going to say like, hey, I want to deploy something on a cloud, but like it needs to have the serial port connected to it, right? That's not a very common thing. But it's a very, very common thing on the edge.

So one of the things that is happening is just in terms of the virtualization of this IO, and virtualization and sharing of these resources among multiple different applications. And this is a very big step towards commoditizing the hardware. It's not just about the hardware and the availability, but how you use it. By virtualizing, by introducing that virtualization layer

on top of this hardware and then being able to like run your applications on top of that virtualization layer and being able to like pass this io or whether it's ethernet whether it is gpus or a serial or usb or a display port right all of these things when you do that now all of a sudden your your application is no longer tied very very at the hip to this IAM. And that is a very big change that happened, I would say, in the last five to six years that makes this whole process possible. And also the form factor of that virtualization there has become smaller and smaller. So now you can dream about putting those things onto very small. So, for example, the smallest device that is out in production in very large scale with

Serida is a form factor of a Raspberry Pi. OK, we are no longer talking about this large specialized equipment to run virtualization, right? We are talking about these very small devices that will run virtualization and you can develop applications on these devices just and you can develop applications on these devices just like you would develop applications on a very large device which has a full-fledged virtualization layer on top of it so this is something that makes the whole underneath layer a commodity and because these two things are coming together in the last one or two

years or three years that is where like you can now take advantage of the full stack. Yeah, I've definitely experienced that. I've pulled out my hair trying to figure out which port is this thing? How come it changed? You know, how come the USB port is specified with this completely different nomenclature on this other system, or even a different port, a different device is suddenly dev what? And similarly, I pulled out my hair trying to use a lot of oddball small form factor hardware.

I mean, no dig against ARM generally, but in my experience, a lot of the Intel small form factor hardware is just a lot easier to understand and to use than a lot of the more specialized, you know, you mentioned Raspberry Pi. I know you were just using that as an example for the form factor, but honestly, a lot of those alternative ARM platforms and so on, well, it's hard to get everything up and running, get all the devices working, figure out the device driver situation. Whereas, you know, I think one reason that the Intel NUC has taken off so well is because they freaking work. I mean, they're standard chips, standard interfaces, you know, you can run standard software on them. And it's just a much more friendly and supportable environment still.

And of course, we're seeing, as Al mentioned as well, this whole world of fanless solutions that use everything from Xeon D down to the Pentium and next generation, Atom. And we're looking forward to seeing where Intel's going with that, where AMD's going with that, to see even more fanless, ruggedized, sealed boxes. So all of these things, I think, are contributing. But one of the things you said in there, Raghu, that really, I think, resonates with me is kind of flipping that upside down and looking at the standardization of the application. That is so true. None of this can work if the application expects specialized hardware. But increasingly, applications that are being run out there are able to adapt to a virtual environment and have come to be expecting a virtual environment.

Are there any ways I think that you're seeing particularly strong in that or particularly weak in that in terms of the applications that are running on an environment like this? It's not about strong or weak, but I want to think about it as simple and complex. So there are applications that are running where they are developed, assuming that a set of resources are available. Now, they are not tied to the hip in terms of the exact same resources, but they would say, hey, I need an Ethernet port, I need a GPU, I need a display port and a USB to do whatever I want to do, right? So that's one kind of an application where the application definition tells you like every single thing. And now you just deploy it and you deploy it based on

the resources that are available if the you want to deploy it at a site where the resources are not available you just go back and figure out another place right or just if that is the location that the application needs to deploy then you say like okay this is not possible at this time that's the i i consider that uh falling into the simple uh simpler uh applications where it is very black and white uh to say but then there are the other types of applications that are becoming more and more prevalent is uh people are taking this large um stacks of application uh that they that they are used to running in data centers and like cloud and everywhere and they're saying like why can't i run the same thing here it's now the application is not about a single entity but it is the full breadth of all the

things that are done right um so you can connect you it will start with a connectivity solution especially when we are talking with in the edge right it can be connectivity or a wi-fi or a lte or or an ethernet but then sd1 or a cloud connectors these are things that come in like very very easily and then they always have like a firewall element idsS, IPS element, those things are already part of that. And then there are virtual machines, containers, runtimes, you name it. So as a whole, this application is providing an outcome, right? Now, all of a sudden, the complexity of that application increases quite a bit, because it is no longer about like, am I running a VM or a container? It is about, am I running all of this in a certain way that the outcome is satisfied, right? Sometimes like

even very sophisticated techniques that are used in the Kubernetes world or in the cloud, where horizontal scaling or resiliency, the auto-restart, all of these things are put into place in this application definition. Now, those are the things that make Edge very, very interesting on one hand, but they also make it very, very similar to like the applications or the solutions

that are deployed in the cloud or the data center. And that is where a lot of the commoditization that we talked about comes back into the picture because you have done this once, right? And now the demands that are put on the infrastructure there are very heavy because now you are taking two or three smaller, very small devices. Intel NUCs are a great example or even ARM, right? Like ARM is not as mature and as standardized but it is getting there and we see a lot of arm too okay so now you take these things

and say i want to do exactly the same thing that i have done in my data center but in a smaller scale but the number of moving components in that in terms of the applications, connectivity, security, everything, those things don't change, right? It's just the the the scale in terms of the performance of these individual components might change, but the number of components and the connectivity and the dependencies between them don't change. And now you have to deploy and manage the lifecycle of all of these things at the edge, which is a completely different ballgame. Yeah, we do like to say that the edge isn't like the cloud,

and yet it is like the cloud, because we want to be able to deploy applications out to our edge locations the same way we deploy them out to cloud locations. We want to be integrating with our CICD pipelines to get the application updates out. But we don't have, you know, the massive elasticity that you get with public cloud where you're essentially building this tiny micro cloud at each of our edge locations. Yet, we still want to be able to have that

pooling and sharing of resources, as well as there's a really important thing of having multiple stakeholders in there as well. A single edge deployment, you may well have three different business units who have applications down at that one edge location, and that they're all authorized to make changes just to be a little part of it. And so there's a whole lot of that security foundation

we expect from a cloud that needs to be built into the solution. There's a lot of cloud that's an edge, but there's a lot of cloud that can't be done at the edge. And there's a little interesting dichotomy there. It is. And the way I look at it is like it's an evolution, right? When you went from the individual computes or computers

in a place to a data center, you took all the good things that you learned from there and you applied it to the data center and then you brought in a whole bunch of other things. The same thing happened to the cloud. And what is happening is that because Edge is the next iteration of this technology curve, what is happening is all the good things that are built in the cloud, everything from the way applications are developed, everything from the CI CD pipelines, how they are deployed at the end locations, whether it is edge or cloud, people are thinking, okay how do i uh what are the what is the thing

that actually works in the edge what they're saying is like this is a perfect way how do i make this work on the edge how do i make it as seamless as possible so that when i have to deploy an application in the cloud or in the edge then i do it in the same way right that the all the goodness that we have developed in the past few years or in a decade in the past decade in the same way, right? All the goodness that we have developed in the past few years or in a decade, in the past decade in the cloud side is now being implemented on the edge. So I think like all those things will come to the edge, but like you rightfully said, right? It is, the scale is not about how many compute or how many endpoints do I have in the cloud like is it a hundred thousand is it a million now it is more going to be the scale of how many clouds do you have right these are these

small micro clouds and now you're talking about ten thousand of this you're talking about hundred thousand of this and how do you manage a hundred thousand clouds no one has done that before right and it is a completely new kind of a problem that people are trying to solve uh but the good news is that like there is a lot of learning that we have done in the past and we are trying to address that but i i don't want to forget what you said which is very very important it is about the stakeholders too uh it's not only about how do I deploy these applications when it comes to edge, because there is no real single entity that owns it. Everything from operational groups to security groups, to development groups, to like you said, there can be multiple operational groups

in the same place, right? Now, how do you make that happen? How do you make it so seamless that when people try to go and try to deploy an application, then they don't have to think about 100 different things to just make their single application happen? By abstracting away all of those things, that is what will make Edge possible. And that is actually what will make edge scalable in terms of adoption. Yeah, I'm really optimistic about it simply because of what you mentioned. I appreciate you bringing that up, Raghu, that we really are bringing the best of the things we've learned in data center computing,

the things we've learned from cloud computing, especially. And we're seeing those adapted at the edge in really exciting and interesting ways. Some of the things are being used, well, in ways that would surprise people, as we will, I'm sure, be talking about with Kubernetes. Some of the things are being used in ways that are quite predictable in terms of commodity hardware. But all of those things, I think, are really taking the best of the past and bringing it forward into this new environment. And I will say one thing as well, that especially as we learned at Edge Field Day, the companies that are approaching Edge from the perspective of this is something new, this is something different, and I need to keep in touch with, as you say, the stakeholders

in order to develop a product that's appropriate for this. I think those companies are doing much, much better than the companies that look at it and say, let's see if we can throw what we used to do into this and see how it works out, because that's just never going to work because it is such a different environment. As you say, nobody's ever had to manage thousands of endpoints, thousands of locations, thousands of clouds before, and it takes a completely new mindset. So I really appreciate this conversation. Thank you for helping us kick off season five of Utilizing Tech, where we're going to be focusing on edge computing. Raghu, before we go, where can people connect with you and continue this conversation?

I think like all the things that we talked about today are the base principles, are the founding principles of why we created Zadida. And this is something that you can experience today. You can go to Zadida. There is a self sign on. If you want to experience Zadida, just go there. Zadida.com, www.zadida.com and just sign up on the self sign on. And you can then start experiencing what it is to do this in a programmatic way.

Yeah. Deploy these 100,000 edges if you want to and be able to like do a fleet management of all these edges and do it in that is driven by intent and driven by policy rather than one at a time. Right. So again, www.Zadida.com just go in and like use the self-sign-on button and click away and start experiencing Zedita

and see what Edge is really about. Thanks for that. And I would also recommend checking out the presentations from Edge Field Day 1, which you can find at techfielday.com or just Google Zedita and Field Day and you'll find them. How about you, Al? Well, you can find me through my demitast.co.nz site

where I write my own content. But also it's really a big part of my life is being involved with the vBrownBag podcast and having community members helping to educate other community members. Take a look at vBrownBag.com or the YouTube channel for vBrownBag. We're everywhere.

And hopefully we'll see some of you in person at events like VMware Explore. And maybe even we'll see a return of the Build Day live events this year. I'm really looking forward to getting back to doing that travel. Yeah, that is a lot of fun. I definitely recommend checking that out. And it's also a great community, especially, as you say, on-site at those events.

Also, please do look out for another Edge Field Day event coming. Again, go to techfielday.com to learn about that. And also, I do recommend checking out the Gestalt IT rundown, our weekly news show on Wednesdays, as well as the on-premise IT podcast every Tuesday. Thank you for listening to Utilizing Edge, part of the Utilizing Tech podcast series. If you enjoyed this discussion, please do subscribe in your favorite application.

You can find us in pretty much every podcast application under Utilizing Tech. And please do give us a rating or review. I know everybody asks for that, but it really does help, especially now that we're changing topics here with Season 5. You can also find us on YouTube. Just go to YouTube slash GestaltIT video. This podcast is brought to you by GestaltIT.com, your home for IT coverage from across the enterprise. For show notes and more episodes, go to our special site, UtilizingTech.com or find us on Twitter and Mastodon at Utilizing Tech. Thanks for listening to Episode 1 of Season

5 of Utilizing Tech, Utilizing Edge, and we'll see you next week.