The Rugged Edge Survival Guide: Transformative Industry Trends & Technologies for 2021

Welcome to the rugged edge survival guide, a rugged edge computing podcast by premio, where it's all about the hardware iyo join us on our constant search into how embedded computing solutions are transforming the enterprise business landscape. Hello, everyone, welcome to another episode of the rugged edge survival guide, a premio podcast I'm your host, Daniel Litwin, the voice of BTB. And folks, Thanks so much for joining us on another episode of the show. We really appreciate you listening along for more rugged edge computing thought leadership. As we get into today's topic, make sure that you're heading to our site, premio inc.com again, premio I NC for more information on solutions and services, and of course, other pieces of premio content. And also make sure you're subscribing to the podcast on Apple Podcasts and Spotify. Hit that subscribe button. And you'll have a full catalog of previous episodes as well as notifications when we drop new ones. So on today's episode of the podcast, we're exploring the transformative industry trends and technologies that are going to most shape twenty, 2021 and beyond intersecting how said trends are being shaped by a variety of different technologies, including compute, storage, and connectivity technologies. We'll be breaking down updates to automation, tech, automated devices, edge computing, the cloud and much more, speaking to how different market forces are leading these various transformations. So for insights on the technology and its importance, I'm pleased to welcome a frequent guest, Dustin setsu, product marketing director at premio. Dustin, great to have you back on. How are you doing? I'm doing OK. I mean, we're making our way through this pandemic and covid-19 situation. So I kind of wish to the audience and everyone that everyone's staying healthy and hopefully getting the vaccines as they come out. So it's. Yeah, you know, I guess the same. I remember us having a few conversations since the pandemic has hit and each one kind of starts the same with fingers crossed, hoping that we get past the other side quickly. And I don't know, all my optimism hasn't really paid off yet. Things are still bad. So fingers crossed that the vaccine distribution happens quickly. And we can get back to some semblance of normal. So we can all feel a little more empowered. I'm with you there. Yeah, 100% I think in that year period, I think technology as a whole has all shaped our lives and brought some benefits here and there. So in this thought leadership and discussion, maybe we can talk a little bit about some of those technologies. Yeah, absolutely. And we'll definitely intersect some COVID conversation here to start as well. So we better understand the market forces that are influencing the transformative tech we're breaking down today. So let's go and jump right into it. Dustin, let's start at a general level. We've been seeing, like you said, a massive intersection of hardware and software around compute technology as of late, which is creating both disruption and opportunities around strategic enterprise level technologies. Can you lay out some of the key reasons why such computing technology is so valuable for today's economy? Yeah, so I think that's a very great question. And you kind of highlighted what the economy is, right? So the economic factor of what COVID has done is definitely put a strain on the entire economy. So I think in the next coming years, I think business as a whole needs to take a step back and really look at where their investments are going into in terms of their ownership. But really, where do they want to put their investments? And a lot of those investments, I think what they did was really kind of put a strain into investing into a lot of these newer technologies to kind of shape better workflows, to support collaboration and just better business efficiencies. So, I mean, if you just kind of look at technology as a whole, I think what technology really does is that it drives major benefits for convenience. And what it really does is it streamlines many traditional legacy processes that probably took longer than normal. So how that's being valuable in today's digital economy, right. Is now being able to take a lot of this data to be able to process, store and connect the world, the global world altogether in order for better efficiency. So really, I mean, right now, what we're seeing and it's continuously trends over time is the digitization of data. And this data comes from all these devices that are coming online. And as these devices come online, they have information that needs to be aggregated. It needs to be processed. And really what's driving is helping shape these new models for intelligence, machine learning and ultimately artificial intelligence. So I think what COVID and the Potomac really kind of put forth. Is it really showcased and highlighted these extremely beneficial value add-ons of what technology is doing and really help the human and the global population to have more conveniences and more efficiency, to drive better operational goals and business goals. So, I mean, really, right now, I think where the focus is. And a lot of these business enterprises is taking this transformation of data and really moving it and looking at the workload of how it's going to provide insights, how it's going to give results, and really how it's going to really help them make better business decisions to prevent a lot of these risks. We mentioned COVID to start. I want to get your thoughts on those market forces as well. Have you seen any COVID driven wheels of the market, such as remote workflows or corporate consolidation or, you know, expanded and restructured networks, digital networks and telecom networks? Have any of these forces or any others that I didn't mention, have they had any impact on the need for more innovative compute technologies? And if so, what kind of level of impact are we talking about here? Yes, I think it's different on every different type of enterprise goal and what they're looking for. But in general as a whole, I would agree with your statements 100 percent, that a lot of what COVID did kind of shaped a lot of the work, remote workflows that shaped how people are interacting, how businesses are interacting with their customers, how customers are accessing information for certain business channels. But I think what COVID that pandemic really did is it really tested the elasticity, elasticity, the range and truly the value of what technology is and what technology was doing is that it provided. An accelerator for a lot of these newer type of models, so, you know, especially when we're talking about computer tech, right, the ultimate goal is taking these frameworks, aggregating that data and having the ability to process, store and connect all this data, which is very, very valuable in the 21st century, specifically in quarantine. So I have a few examples that we can kind of walk through that we experience. I think most people experience there and hope that, for example, being shifting from going into the office, daily workflow into now moving all your activities, all your interactions into a more remote workflow and corporate consolidation, like you mentioned. So the challenge there is kind of migrating all of these tasks, all these assets into an area to where it still delivers efficiency for the workforce. So I think where the cloud and all the cloud assets and being able to interact with the cloud dashboards and having access to all the resources for remote work that has been monumental during COVID specifically allowing different departments, organizations to still maintain and hit their goals. A second example that I think was interesting to really see, and I think we're still learning and still growing from this and I've been reading a lot about it, is the ability for a lot of educational systems and online learning for schools. Right I think this was a major challenge during COVID because I think learning is a very unique elements where it's a very unique process with between teachers and students. But school has really had to figure out how to leverage the technology out there to streamline the learning platforms. So whether that be technology and providing connectivity for students when they're sitting at homes, especially even with simple hardware and laptops, for them to even access this information. And then third is moving all this information lessons curriculum into a digital platform definitely was not easy. So I think that was something where technology can technology itself is continuously going to help this type of digital learning platform continue to grow as a result from COVID. And the third, I think, is a major shift in, like I mentioned, these digital channels for retail and commerce. So not only are these retail models looking to get closer to their consumers, but I think consumers are also taking an active step to really move into the digital channels because they're really stuck at home and not able to really go out into the retail stores. So I think a perfect example of this is that I will share my personal example that I've learned is just simple grocery shopping and buying your everyday essentials that you really do when you go to your local grocery store. And a key example here was, I think what Amazon did and their whole Whole Foods platform is they moved that entire shopping experience into a digital platform. And it was very unique for me is that I was able to sit on my couch as if I was scrolling through my smartphone and had a digital aisle like I was in a supermarket. And I was able to go through and actually pick out everything that I needed for the week. But if you go down one layer, even deep, you even have the interaction with the person that's picking the actual item from the shelf at the store. So, for example, if there's an item that is selected and it's not really available, that person can have direct communication with you at that moment in time through connectivity and have a decision on what to pivot to in terms of another object. So even after they collected all that, all the groceries, I mean, even having the actual ability to see the driver in this car make that delivery and drop it off in front of your home. That was incredibly huge in terms of leveraging technology during good times. And I think what we're going to see is a lot of those COVID driven market forces aren't just going to go away when we return to whatever normal looks like. In a lot of ways, they reflect changes that were already happening in these various spaces and really just accelerated the adoption of various technologies. And I think we're going to see, as you know, the business world looks to how can we make do with less or how can we continue to get more efficient as know, the margins for error become tighter and tighter. I think a lot of the trends are about to break down, will fit in nicely to, you know, to that broader context and some of the ways that I think we're going to see technology employed in the future. Yeah, I think you hit one thing that I want to touch on is that they're not going to go away. And I think. And when we're in crisis mode, specifically in pandemics, it forces businesses. Really come up with new solutions and think and implement things for change, and as you know, change is not easy across the board. Change is not easy to implement, especially when you're in a large, large enterprise organization. So I read something that was really interesting. I want to share. So McKinsey with the leading research consultant. They basically did a survey and they surveyed all these C-suite executives, all these senior managers, and they basically ask them a question. And in response to covid-19, what changes that they implement that they thought that were impossible, that they were able to succeed during this pandemic. So there's three items that they actually were extremely successful that they thought were almost impossible to implement over a long period of time. So that, number one, being increasing the use of advanced technologies in their everyday business operations. So in that survey, the expected timeframe days wise is that they thought that it would take them to implement these advanced technologies. 672 days, if you actually look at the time it took for them in a crisis mode, they were able to implement advanced technologies and almost around 26 actual days. Right, so the acceleration factor to that is that they did that 25 times faster in crisis mode. Another thing that I mentioned that they were able to do is really increasing the migration of these assets right into a digital cloud environments where their workforce can really access that. The expected days for that was almost 547 days. They were able to do this. 23 days. Right, so you can see that when there's some type of crisis, enterprise businesses are really starting to figure out how to be efficient, how to be more productive. And then last but not least, is the spending on data security. And it just is a natural progression as you move more assets into the cloud and it's going back and forth between different layers, different departments and their individual workplaces, whether that be at home, that expected day range with around 449 days. So they were actually able to use more data security around 24 actual days. So that survey was very interesting to me to kind of showcase that. What you mentioned, Daniel, that a lot of the changes that are made short term are definitely going to be there long term for the beneficial of the overall business or enterprise. Well, without further ado, let's go ahead and break down some of those transformative industry trends and technologies that are going to shape 20, 2021 and beyond. We've got three main ones that we're going to break down. But I think it's important to note that each of these technologies intersects with the other. And also intersects with various other important technologies that we may not totally hit on. So this is just a selected view of some top trends. If we don't get to your favorite trend. Don't send us any hate messages, but let's go ahead and jump in. We'll start with the first one, which is hyper automation. So this is a tech trend that is taking the more traditional automation that I think the average consumer or business professional is used to and empowering it with more refined artificial intelligence and machine learning. And this is seen across a variety of different tools, not just to support the automation itself, but also provide more holistic analysis, measuring, monitoring and discovery. So really just a broader and more efficient kind of automation. So when we think about machine learning, practically, how does it work with various enterprise tools that are driving artificial intelligence today? I think this will help us get a better picture for why hyper automation is a thing in the first place. I mean, hyper automation is just a very cool way of just streamlining automation. And if you just look at automation, what it's doing is just providing automated tasks. That doesn't require a lot of human effort that has been streamlined and it can provide an actual results based on some type of program will function. So taking one step back, right. What machine learning itself is not something that's new? I think humans are always looking to automate tasks for better efficiency. But what machine learning is, it's that it requires a large, large amount of data sets and a large amount of storage in order to take a lot of this information to develop these machine learning algorithms for success. So how this was done in the past and how traditionally what it's been doing in terms of data aggregation is that in order to actually drive these machine learning models, all this data through big data analytics would need to be pumped into these large data centers where they have the resources for the post processing. They have the resources for the deep learning to really enhance the machine's ability for what the ultimate goal is for. Automation is delivering some type of intelligence. Right So how does this benefit the. Private businesses, so the enterprise businesses are not just interested in the ability for faster, better processing, but really the key element for these businesses to have better control, once you have better control, you have the ability to make better decisions. When you have better decisions, you can make. Cut costs, you can prevent risks. And as a business and you look at a business as a whole, those are very key defining elements of making a business successful. So really, I mean, when you look at Hyper automation, the key is really to offer the ability for enterprises to very quickly Institute action, but also change in the response of real time events based on what the data is showing in real time. I think what's going to be helpful for the audience and the people listening is to really identify three levels of hyper automation and how technology really kind of facilitates that in a closed loop of hyper autonomy. So the first stage. That's the most important of hyper automation is what I call what a lot of people are defining as a cognition stage. So the definition of cognition as a whole is once you have cognition, you have the ability to acquire intelligence, you have the ability or you have the ability to acquire knowledge and understanding through thought, experience and senses. So if you kind of look at what these IoT sensors are into things. And what these things are doing is all these sensors that are deployed, whether they all have a way to sense. And harm and turn that into a data. So if you kind of look and compare an analogy of humans, humans do the same exact thing. But we do it innately. We do it naturally. We basically use our five senses. Right we use our senses of sight, hearing, smell, taste, touch. And what that does is that helps us humans, to really generate a contextual, instinctual awareness of the situation to make a better decision. So in the sense of hyper automation, the machine needs to have that cognition where it can leverage the data and basically build their own context and awareness. So this is a perfect segue in terms of how hyper automation, where there's this thing called the digital twin and what a digital twin is doing is essentially very similar to a human twin. It's a carbon copy of the actual asset that's being deployed. And why that's beneficial when you have a digital twin is that you can now do a lot of the data manipulation of modeling that can be used to automate these type of business decisions. So it can be used as, say, for example, a proxy to run some type of model in that twin before it actually gets deployed. This is the first stage of the automation that's creating this cognition, a second stage of hyper automation that it's very important that ties into the cognition that comes after is the intelligence. So once you have that secure a situational awareness and context, you can now build the intelligence. And once the machines can have the sense of this awareness, it can start now, really deliver their machine learning algorithms and really streamline the decision making. Right, so once the machine is able to have that early detection, once it has the models for predictive analytics, this is where the automation and the full business goal of control is coming. And that is a perfect segue way into the third step of hyper automation, which is having the full control or full closed autonomy to really provide the intelligence for real-time business results. And I mean, this is all made possible currently. And a lot of the automation is through a lot of AI enabled robotics that are using sensing through cognition intelligence to really kind of culminate into a full circle of automation. So, yeah, that's kind of the three steps of hyper automation. Lovatt, Thanks for that in-depth breakdown. So with that in mind, have advancements in hardware had any impact on making hyper automation more powerful or more feasible for a wider variety of business use cases? Yes or no? And if so, how? Yeah, so coming from the hardware side and being in hardware, premio as a whole has been in hardware computing electronics for 30 plus years. I would say hyper automation really relies on dedicated hardware to really facilitate the steps of the hypertonic right, whether that be cognition, intelligence all the way to full autonomy. I really, truly believe that hardware and software is truly binded through a true partnership because a lot of the software algorithms are written to run on perform performance based hardware. The software algorithm will dictate if it needs to balance between x amount of hardware x amount, of course, whatever it's doing in order to kind of deliver that type of level intelligence. But really, I think a good example of where this hardware advancements is making place is the. Elements of where deep learning is happening, right, so a lot of the data centers are using a lot of robust purpose built hardware or purpose built technologies to really deliver these deep, deep, deep, intelligent neural networks. So when we're talking about, like high performance compute, when you're talking about these massive scale of training machines for in-depth intelligence, if you really look at the hardware, it's just really robust, extremely performance based. You know, CPUs use high performance memory. There's non-volatile memory storage or faster read and write for hot-air storage. And there's even computational storage in this. Computational storage is something that's do what computational storage is doing and saying it doesn't need to kind of go back to the host architecture for the CPQ to do the processing. It's moving some of that processing sitting closer to where the data is, which is on that storage. Right, so what that does is that it creates an extremely amount of processing power when you need it in real time. I think a second example to showcase kind of where a lot of this purpose-built hardware is kind of driving this automation and this hyper intelligence is shifting into what's the inference model. And I spoke about this in past podcasts as well. But what inference modeling is this after a machine has been trained in the cloud or after the machine has been actually intelligence, you can move that model into an inference where you can use a specific purpose built hardware to make an inference. Easy example to understand is if you were training a machine to recognize images of an animal, say, for a dog, once it knows has a 99% accuracy, it can recognize a dog. You could move that into an infant's model to where the sensor will pull some of that imagery into a computer hardware. And then it can basically influence that very quickly that this is a dog. So I think another thing is it's really evidence to see where they're going. So there's a lot of major tier one companies out there that are trying to get as close to either better, faster processing or even connectivity. So just to share some industry news, right. Invidia, which is a leader of cpu's, I mean, there's a reason why they've purchased and acquired melanocytes. Mullinax is all connectivity and then they've also even purchased arm. Right arm is another processor for semiconductor processing. So another example of that is even AMD, right? AMD traditionally is a leader in processing and semiconductors. They're also recently purchased Xilinx for their FPGA. Right, so they can dedicate a lot of their function for these new machine learning models. Last but not least, when we're talking about all these things, where does premio fit in terms of hyper automation? So where we're really driving this market penetration is really looking at all these different technologies, but looking at it from a system level and from a compute level, storage level or connectivity level, we're able to really consolidate all our engineering, mechanical engineering and kind of deliver what we're calling a rugged edge computer for a lot of these newer edx type of environments. And where we really focus on that is really kind of focused in our organization of our industrial computers for a lot of these harsh environments. So we look at it from a system level approach. We really make sure the product is can endure these harsh, harsh, harsh environments, whether it be from freezing cold environment to a scorching hot environment. But that range can go as low as negative 40 to 85 C and D applications, balance of power and efficiency, performance and energy savings. It's extremely important. And being able to really deliver performance, but also having that ability to control your power is huge for these applications and then taking the ability of a famous design where it kind of ensures the longevity and the reliability of the product. Another thing is a shocking vibration by shock. And vibration is something that is very detrimental to a lot of standard computers. So once you're able to kind of deliver a solution where it's resistant to a shocking vibration, you definitely have more endurance and reliability for the product itself. My last main thing I want to touch on with automation is better understanding the applications that are making best use of it. So could you break down a few of the applications that are benefiting the most from more powerful hyper automation? And connect the dots for us with how this is resulting in more efficient processes, as well as output for the companies that are using automation? Yeah, so kind of going back to what I mentioned. And I think if you kind of instill those three key elements of what automation is from your cognition, intelligence and control, I think a good. Example of that. And in terms of an application, is industrial automation specifically focused on robotics automation? So, I mean, if you look at robotics as a whole on a factory floor, robotic on its own, without any type of programming, without any type of cognition intelligence, it wouldn't be able to do pretty much of anything. Right, so when people or when many facilities are using robotics, they're trying to eliminate and create hyper autonomy. They're trying to create hyper automation. So something in their processes. So if it's a robotic arm that is able to decipher things, moving down a line in order to pick it off the line and move it to a packaging bin, say, for example, a soda can, for example, at this robotics has to be trained. It has to have the vision, technology has to have the computer vision technology recognize and sense that this what the contextual awareness that they're given, they have the intelligence to over time through more data to say, OK, I know for a fact that this robotic arm can now be able to choose. And then once you have that final kind of control, you humans really can take a step back and let the sort of hyper automation kind of run its course and really take an evaluative, measurable step back to say how they can improve inefficiencies based on the automation that's put into place. And actually, I've got one more follow up question there for you. How do you see automation continuing to evolve moving forward if hyper automation really is just a further refinement of existing automation and machine learning techniques to make for more robust data analysis and capture and really create tools that are powerful for the end users? Where do we go next with automation? You know what? What comes after? How do you continue to refine the technologies? It really becomes a way for extreme amounts of machine intelligence, for artificial intelligence. So right now, the machines are very kind of reactive into the sense of that. They're moving into an hyper automation state. I think in the future, these machines really are going to move into kind of like a terminator world where I can really kind of start to make decisions before we even actually can make a decision. So we'll see. mean, the technology will continue to grow, but I think it really stems into the ability to make a decision Well before things are even happening to control risk. All right, Dustin, let's jump into our next big trend for 2020, and that would be the empowered edge. So this is another trend that kind of plays on the same theme as hyper automation, which is taking existing technology trends that have already been making an impact across various industries and just further refining them or taking them to their next logical step. So whereas edge computing, just regular edge computing, already focuses on bringing information processing collection and delivery closer to the source, empowered edge looks to take that one step further to support the steadily increasing use of IoT devices while still pushing for data collection and processing as close to the source as possible. Now, obviously, if we're talking about IoT devices which when deployed at an enterprise level can be numerous, that becomes more challenging. So I'm curious your thoughts on that. If edge computing is already pushing for this kind of compute proximity, if that's already the goal of computing, what is the value of a push for unempowered edge? What's really the difference here? Yeah, so I think empowered edge, another way to look at empowered edges to really just drive it with the key goal of it is really it's called intelligent edge. Right and like you mentioned, it's similar to the automation trend. It's really moving these applications, specific workloads are intelligent models that have been trained closer to where the data is. And then they can move into a quicker inference. Right and this really is driving the performance at the edge specifically. So applications and their workloads are unique. Now because these applications get to dictate the type of compute power required to really run these different models. So key here is that every business is different, every edge computer model is different. All these business investments that are looking at the return on return on what they can do with that computing really need to understand their application and what the application is doing and what type of compute power is that they need. So they can kind of map out that data flow from where the endpoint all the way back to the cloud. So what's something that where premio is really focused on a theme amongst our engineers and our embedded team is that we are really moving into intelligent edge. But really we define that empowered edge as something that we're kind of carving our way into, which is a rugged edge computing. So rugged edge computing is embraces all the key elements of benefits of moving data closer to where data is being generated from these IoT sensors. But it really kind of decentralizes a lot of this local processing in a low latency environment, in a product or a computer that has been hard and before the reliability and the organization of these computers. So, you know, what we've seen is that there's these new demands, right, to empower the edge and there's new applications and workflows that require these very specific purpose built computing architectures. And what that's doing is that it's really challenging a lot of these engineers who have traditionally haven't really designed these type of environments. And they have to deploy and come up with these solutions outside their comfort zones where they're traditionally just always having a controlled environment. So quite literally, where premium fits into the dirty sandbox. So I like to say is that the products that we come up with design. And we facilitator's ruggedized computers really are made to be down and dirty. So we really recognize the need to balance the software advances with harvest strategies, really by providing a rugged, high performance computer that is able to survive in the most physical set physical settings. Right in the environment. So we look at everything from the external enclosure, everything from the internal components, everything from the mechanical and the thermal engineering that we put into our DevTest Labs to ensure that in the environment where the rugged edge computer will be, that it's going to still be able to interact with this mission critical data and not have any downtime, which is very, very critical for a lot of these empowered as we continue to kind of move into the empowered edge. I think our leadership in our rugged edge design and our manufacturing scale really allows us to open up a new world of IoT integration, automation capabilities for a lot of our clients and our integrators into these edge computing deployments. An essential part of what is pushing edge computing forward are obviously, the specific transformative technologies that are demanding closer to destination computing. And I think that intersection of hardware and it is empowering a broader push for edge computing is important to highlight. So the big three technological improvements to computing architecture that, at least in my opinion, are doing a lot of the heavy lifting. Here are improvements to CPUs and GPS, expansion of 5G network capabilities and cloud computing in general. So can you give us some insights on these three and how they're intersecting with rugged edge computing and how are they empowering an empowered edge? Sure this is a pretty interesting topic. And these are very big, transformative technologies that are shaping across a lot of different market verticals, but in simple form. Right these transformative technologies and where they're in their scale of their design is essentially a convergence of the latest and greatest performance consolidated into what we put into a ruggedized computer design for these hard environments. So from my perspective, I think it's a very interesting time in compute because many of these applications and the workloads that the required to run are demanding hardware, more hardware requirements in terms of better processing, better memory IoT and even more connectivity. So to go back to kind of that first transformative technology, whether that be GPU or CPU and gpu, they're basically fundamental computing and engines that have been dedicated for applications at the edge. So both of them are traditionally share a very close relationship when it comes to machine learning and processing power. But they really have their distinct advantages when you're kind of looking at the machine learning model. So, for example, CPUs always have been considered kind of the brain of any type of computer architecture. And what a CPU does is basically kind of it's all these transistors that are embedded onto a semiconductor silicon. And what's CPU basically instructed to do is and execute commands and process them based on programs that are designed in a very multitasks serial and sequential order that are able to deliver low latency. But one of the, I would say, disadvantages of a CPU is that once it. It's been designed to do more serial sequential computing. So in order to move on to a next task, it basically uses multiple cores, complete that task. And then go to the next task. So how you get more performance in the CPU is very simple, right? The more CPU, the more CPU cause that you put into a silicon, then you have more cores to actually do a lot of the task and the multitasking for these machine machine learning algorithms. But on the other hand, you have these high performance dupas. And the difference of the GPS and how they're different is the use a lot smaller course. And there's a lot more of these course. But they're beneficial in the sense of parallel computing. And what parallel computing is, is that it's no longer using a serial or sequential type of compute. It's very dedicated through high throughput tasks. Right, so when we're talking about high throughput tasks, CPUs are phenomenal for machine learning when it's dealing with image processing or three rendering of images. If you look at those two types of architectures and those two type of processing architectures, what's really defining that growth and what's defining even more performance as the trends continue to grow are all based on standard architectures and standard protocols. And one thing that's a very common industry that everyone's continually to talk about is that PCI 4.0 serial bus and interconnect. So PC. Ieee 4.0 is just the next generation to computational hardware, and opko ieee 4.0 does essentially doubles the performance of a huge and three. So a good example of this is for the techies out there. Right? so if you have a pc, ieee 4.0 by 16 type of card, if it's on a 4.0 generation, it doubles that performance. So that's by 16 can do around 32 gigabytes per second in terms of performance. How does all that really tie into kind of the future of machine learning and how that's going to be localized at the edge? These accelerators, these are all basically performing accelerators that are being designed in these ruggedized platforms. Right and being deployed in these ruggedized locations to really interact with the data. So that's going to be one of the key, the key elements and key transformative, transformative technologies in edx compute. The second thing that you mentioned, 5G networks. And I think 5G everyone is familiar with 5G by now, with all the telecommunication carriers not promoting it. But what really the benefit of 5G is doing and why it's so transformative is that it's delivering kind of this ultra low latency communication protocol for all these new devices. Right and when you're talking about Ultra low latency, you're able to kind of pass data for these applications in some millisecond level, but also at a high frequency of millimeter wave that delivers 99.9999% reliability. Another thing is when you move into 5g, you have enhanced mobile broadband and enhanced mobile broadband is that now you have these 5G networks. And there's no towers that are able to interact with devices and provide extremely fast downlink speed of 20 megabits per second. And even up to speed of 10 gigabits per second. And the last really benefit of 5G is that machine to machine communication. Right, so what that machine, machine to machine communication is providing is like a one square kilometer footage, these devices connected to over a million device subscribers. And then essentially, it's a 90% reduction in the energy usage. So kind of the last transformative technology. I think you mentioned that really benefits a lot of the edge. Computing architecture is really balancing the elements of cloud and why the cloud is so important. Over the years, the cloud has proven its benefit. And I don't think it's going to go anywhere. It's really about how these models are learning to kind of power and use the elasticity, elasticity of these large centralized data centers out in that network edge. But what you're going to really start to see with edge computing is that you're going to start to see the scalability through. What what I called are these checkpoints and what those checkpoints are basically checks and balances of the disaggregation of where the workloads needs to be placed in terms of data from endpoint to cloud. So a simple representation would be you're going to have mobile endpoints that communicate to some type of edge network. And then the edge network will go to kind of a national data center and data center will go to kind of a micro data center, and then that Micro Center will kind of go to the final data center. So if you kind of look at the path of pathways of the data, I mean, the question that every enterprise is going to define is at what stage in those checkpoints is what type of learning. You want to do, what type of machine learning, you're trying to do with the data that's passing through. I want to get your thoughts on a stat here, dustan, because I think it highlights exactly why unempowered edge necessary. So here we go by twenty, 23. There could be more than 20 times as many smart devices at the edge of the network as in conventional IT roles. And I don't think that's unrealistic. I mean, in most enterprise environments, we're already seeing at least some IoT integrations, whether they're simple, like a thermometer built into a broader facilities network or something much more robust, like IoT enabled manufacturing equipment. Right, so as this becomes more commonplace, as 5G supports the proliferation of iot, what sort of hardware is going to be needed to support that scale of then subsequent computing? Great, great question. I think I kind of touched on it in my last response about the cloud. Cloud is going to be kind of structured. I really do believe there's going to be like multiple layers of frameworks on how the data needs to be processed, stored and analyzed. So, I mean, as we continue, we have all these smart devices at. Edge and come online. I think the key is it's the truly map out the data flow from endpoint to sensor and understanding what type of hardware or compute power is needed to really interact with data inference that data store the data. You push the data and really creates this data continuum really for two key purposes. And I mentioned it all throughout all the trends, is that the ability to allow these platforms to have machine learning capabilities, but ultimately drive some type of intelligent results for a better business decision or a better business insight. And finally, here are the empowered edge. What are some of your thoughts on where this kind of IoT deployment is going to be seen? Where exactly do you imagine we're going to see really at scale IoT deployments that require such an empowered edge? Do you think it's going to be in manufacturing settings, smart city settings, more residential settings? What are your thoughts? It's currently already happening in a lot of the smart manufacturing and the industrial automation of monitoring assets and control. You know, you can imagine in a manufacturing line. There's a ton of different automated tasks that are happening, whether that be through automation lines, whether that be through programmable logic controller sensors. So within the last five years, I think that's where a lot of the empowered edge is really taking true benefits from to really streamline a lot of these trends, starting from hyper automation and really driving the robotics of this. Right, so we work with a few robotics companies. We work with a few metrology companies and vision companies. And what they're essentially all they're doing is interacting with the data, building models to really drive more inferencing and decision making to help streamline their productivity and really drive more efficiencies for their production. Perfect Thanks for that. All right, Dustin, let's get into our last main trend that we were wanting to break down on the podcast today, and that would be a high security. So I think as a trend, this is kind of the bow to tie all of our thought leadership together. So far, because while hyper automation empowered edge capabilities, a wider network of IoT devices and autonomous enabled devices, while all of this means more efficiency, and more smart enterprise functionality, it can also come with a lot of new security vulnerabilities. And these vulnerabilities are often ones that IT teams are just not prepared for or trained for, especially when dealing with tech like AI. So let's dig in there. What makes AI deployments, in your view, particularly vulnerable to new security threats? Yeah, so I think if you kind of walk through some of the discussion points that I made earlier, how you get to AI and how you to get to some type of business decision is all based on data and what makes data, particularly vulnerable. And specifically talking about AI is that once you move, once you're able to take that data and move it into an AI model for something to react silly or make a change or make some type of business. That's fully automated through hyper automation, the empowered edge that opens up an opportunity for a lot of this vulnerability for cyber attacks. So, for example, in a simple example, I think we can talk about would be autonomous vehicles. You know, autonomous vehicles are right now in kind of proof of concept for a lot of different car manufacturers. What they're doing is they're basically doing these tests drive from point A to point B to really prove their autonomy, software algorithms while working with all of the sensory data. But let's say, for example, they get to a stage of full level five autonomy where we're at a stage where we have autonomous cars driving around all throughout society. Right that potentially can open up a lot of different security risks based on the AI. Right, so if a hacker or someone went into it and was able to kind of take ownership of an autonomous vehicle to create either acceleration or some type of thing, that that would be extremely detrimental for the AI. Right, so I think one thing that's key, when you're discussing all these cool trends and all these cool things that are helping us with convenience and streamlining a lot of these things, I think another element of that that ties it all together, like you mentioned, is really understanding the elements of hardware security, understanding how to protect very, very critical data for the application itself. Do you have any more specific examples of these? Kinds of vulnerabilities being exploited by bad actors, so just some grounded examples of this at work, and then can you also explain what the consequences of being of using AI as a vulnerable security area? Yeah, I think a good, easy example. I think a lot of the people listening to probably understand at all deals with identity theft and like privacy and information. So I think it was a few years back. I remember, if you remember Marriott, the hotel chain Marriott had like a huge, huge data breach on all their personal information and why that's so dangerous. Right and I think from the level of privacy is that all this personal information and identity theft. Now can be sold to on the black market. And what that does is that creates a security vulnerability for identity theft. So it's very harmful if, you know, enterprises or companies who are dealing with debt are dealing with all this data to not have any protections in place to protect user data or privacy itself. Well, and obviously, that's a high profile example of AI being targeted. Is the same happening to smaller to mid-sized companies? Is the temptation to go after a vulnerability is the same for deployments that are smaller or are bad actors mostly trying to exploit vulnerabilities in the most high profile scale deployments? Question? I think it depends. I think hackers or people who are looking to exploit are going to look for the easy elements of where security protections are built into place. So I don't think it really matters honestly, if it's high profile or small, medium sized, I think if there aren't protections in place for AI security itself, that just opens up a can of worms and a can of issues for exploitation specifically and security breaches as a whole. So especially with you, when you're dealing with technologies that deal with data technologies that dealing with image capturing specifically. Right because image capturing, deal with actual information of physical identity, people faces and all of that is extremely huge. So I think it really I don't know if it's high profile, low profile, as long as there is a loophole in or if there's a way in, I think it can be exploited. When we think about security, I think it's also important to hone in that there are three key factors or I guess aspect to AI security. So there's obviously protecting the AI systems. That's the most straightforward version. There's also leveraging the AI to enhance security. So using the AI as security. And then anticipating how attackers may use AI against you, which is the more proactive approach to preventative security. So how should IT and cybersecurity teams, in your view, weigh these various factors of AI security? What is important and where the first one protecting air power systems? If you just very simply is basically using the AI and training. It to really have the protections in place to where it's becoming a little more protective for these machine learning models itself. Because it's I mean, essentially what you're kind of doing is using AI itself to protect itself. The second point that you mentioned, I think was key is leveraging AI to enhance security defenses. And kind of what I really is, if you look at it, it's based on a machine that's been trained with data to really detect and make a decision. So kind of, you know, for the good and the bad right, you can actually use AI to enhance the security defense. So you can, you know, the same way you would train a machine to recognize images. You can also use the AI and machine learning to really understand specific patterns, to really uncover attacks and really automate parts of whatever that cybersecurity process is. Right because most of these cybersecurity attacks, there's know, if you investigate, you look into it. Most of the times there's elements of weaknesses. There's patterns that develop. So if you can really train the machine to kind of do that ahead of time, then it really creates a layer of enhanced security. And then last but not least, as you mentioned, by anticipating kind of that nefarious use of AI by attackers, you know, if the AI is smart enough and through time, through data and it's becoming. Smarter and smarter and smarter, if it can raise a red flag to identify a potential attack and actually even deploy a solution or say like a defense mechanism way ahead, way ahead of the attack and control that environment, then you really, really have a strong element of security. Now, where does compute technology fit into supporting AI security? Why? why is that an aspect of the broader AI security conversation that is essential to hone into? I'm not going to tie into every single computer technology because I think in terms of machines and learning, there's going to be so many different layers. That can be another podcast just to talk about security, cybersecurity as a whole. But I think where it's most important is to tie-in to kind of on the hardware level from a computer architecture. What are the kind of key pillars or the key protocols that are defining kind of hardware security? All right. So one of the things that are built across all our products is based on a platform. And a protocol that's defined by industry standard, and that is the Trusted Platform Module 2.0. And in the TPM 2.0 module, what is doing in the physical hardware chip that is sort of down onto the board. And that opens up all these different functions for hardware, encryption security. You can do authentication through certain type of fingerprint ID. You can even have bootloader codes to where it has to make a signature before you can even access a layer of software to even boot into the system. So I'm kind of, in a general sense, right from a hardware side. Once you have this TPM chip on most microcontrollers, it actually now allows that data to be encrypted on the physical local platform or computer before it even actually needs to pass some of that data out of its box into a different network. So once essentially, I mean, if you're able to encrypt once the data is encrypted and protected, even if there is potentials for attacks, and if there's access to that information, if it's already been encrypted, there's no real downfall, because the information has all been encrypted and it's unaccessible. All right, dustan, I think that about does it for our conversation. We've broken down three key trends. So far that are going to be shaping twenty, 21 and computation technology that includes hyper automation, the empowered edge and AI security. So let's go ahead and wrap up our podcast with a final question, or you might have a follow up depending on your answer. But the last main topic, just looking into the crystal ball a bit. So how should enterprise level professionals, in your opinion, specifically ones that are working with these technologies, how should they move forward into twenty, 21 with all of these key trends in mind? And where should they start to strategically and invest in compute power that can support all these various trends? Yeah, so I think to recap. Right everything from hyper automation, from edge compute and kind of that security layer, really the investment comes into understanding what the application needs. And then the application. And we'll define really where the compute architecture needs for to really drive the machine learning and the intelligence. So, I mean, the return on investment for many of these enterprises is really going to be coming from what's at the end result of the intelligence. Right, so once you can move into an intelligent layer, that intelligence is going to be able to shape a lot of the different changes in the operational organizational structure and their business objectives. All that if it can be done in an automated situation or automated environment, it's going to be extremely, extremely, extremely beneficial for these major enterprises. So I think it's important to really keep their eyes open, ears open to what's changing in the market, specifically in a lot of this hardware technology, because like I mentioned before, I don't think the hardware is going to go anywhere. It's just only going to get more specific. Each specific type of hardware design is dedicated to a specific purpose, and that purpose is to really figure out more ways for better or hyper automation, bringing more intelligence to the edge, aggregating that data and really delivering. Extreme amounts of artificial intelligence and hopefully we'll get to a point where artificial intelligence is not taking over our world, but really helping us to even become even more streamlined and have more convenience and not only in our personal lives, but also in our business lives. All right. Dustin setu, director of product marketing at premio, thank you so much for joining us and giving us some more insights today, breaking down the transformative industry trends and technologies that are going to shape 2021 and beyond, and more importantly, intersecting how these trends are being shaped by compute, storage, and connectivity technology. So, again, Dustin, I appreciate your time. And if folks want to find out more about premio and some of the work that y'all are doing in this space, how can they get in touch or learn more in order to kind of find more information about a lot of what we're designing product wise? Please visit our website. And that is premio and C. There's a section, too, where if you're really interested in learning about how certain computer architectures are helping a lot of these IoT and machine learning inference model applications at the edge, we have a lot of case studies that really kind of dive into those stories. So, yeah, looking forward to more technology and twenty, 21 to come. Fantastic, Dustin. Appreciate your time. We'll chat again soon. Thank you so much. Thank you, Daniel. Have a good one. Everyone stay safe. And thank you, everyone, for listening to another episode of the rugged edge survival guide, a premio podcast, if you like what you heard and want to listen to previous episodes, make sure that you're going to our website. Premio inc.com again, premio I NC or subscribing to the podcast on Apple Podcasts and Spotify I'm your host, Daniel Litwin, the voice of B2B till next time.