Over The Edge

How Open Source is Expanding the Horizon for IoT and Edge with VMware’s Malini Bhandaru

Episode Summary

Today’s episode features an interview between Matt Trifiro and Malini Bhandaru, who leads open source IoT and Edge efforts at VMware. In this interview, Malini discusses her work in The Open Source Technology Center at VMware, the valuable role that open source plays in IoT and Edge, and interesting use cases for how Edge is solving macro problems.

Episode Notes

Today’s episode features an interview between Matt Trifiro and Malini Bhandaru, who leads open source IoT and Edge efforts at VMware.

Malini has a Ph.D. in Machine Learning from the University of Massachusetts, and prior to joining VMware, worked on big data for autonomous driving and OpenStack for cloud infrastructure management at Intel. 

In this interview, Malini discusses her work in The Open Source Technology Center at VMware, the valuable role that open source plays in IoT and Edge, and interesting use cases for how Edge is solving macro problems.

Key Quotes

"That's the beautiful thing [about Edge] -- it's everything. It just combines everything, whether it's security, cryptography, blockchain, machine learning, or collecting sensor data that comes across to you from different protocols-- it's just everything."

“By making these open source projects and a lot of the infrastructure available, you're opening the flood gates to many more adopters and many more applications and solutions coming to market sooner. “

“The IoT edge market's going to grow to a trillion dollars. VMware is keeping its eye on the future when there's going to be more edge adoption and larger edges.”

“The telco edge is going to be pretty beefy and it will enable applications like virtual reality and augmented reality and connected cars. Some edges are going to be very, very beefy edges with a lot of compute, a lot of knowledge, a lot of storage at that edge.”


Over the Edge is brought to you by the generous sponsorship of Catchpoint, NetFoundry, Ori Industries, Packet, Seagate, Vapor IO, and Zenlayer.

The featured sponsor of this episode of Over the Edge is Vapor IO, the leader in edge computing. We want to be your solution partner for the New Internet. Learn more at Vapor.io


Connect with Matt on LinkedIn

Malini's Blog at VMware

EdgeX Foundry

Episode Transcription


[00:00:00] Matt: Hello everybody. My name is Matt Trifiro  and I'm the chief marketing officer of Vapor IO. And I'm also the co chair of the state of the edge project at the Linux foundation. And I'm here today with  Malini Bhandaru of VMware, where she leads  the open source IOT and edge efforts at VM Ware.

[00:00:14]So, Malini, tell me a little about how you even got into technology.


[00:00:18] At home, my dad's a scientist, he's a nuclear physicist and an astronomer. So it was kinda natural. So I don't take much credit for that. It was like, you know, science everywhere at home.

[00:00:30] Oh, that's great. so what, type of technology first captured your interest?

[00:00:37] I wanted to become a doctor. And now I'm a computer scientist. I have a PhD. That's the difference?

[00:00:45] Matt: Your doctor, just not just

[00:00:47] Malini: Kind of let it go, doc.

[00:00:49] Matt: Yeah. That's great. That's great. what led you to IOT and edge? it's certainly a narrow space or at least it's growing now, but

[00:00:56] certainly when you started.

[00:00:58] Malini: You know what, it's not [00:01:00] natural. That's the beautiful thing. It's everything. I mean, if you were a kernel Linux developer, you, you have a Colonel on that edge device that has to do the compute. Okay. But it's on a, maybe a smaller constraint device. You need cloud because at the end of the day, if you have lots of these edge nodes, all the data has to come to one place.

[00:01:20] You want to analyze it. You want to have a dashboard view of it. You want to maybe monitor and manage it from a very remote location. You don't exactly want to go to the oil drills or the winter binds that are all over the place. So the awesome thing is it just combines everything, you know, whether it's security and cryptography and blockchain and machine learning and, and just collecting sensor data that comes across to you from different protocols.

[00:01:48] It's just everything.

[00:01:49] Matt: Yeah. So when you say IOT and edge, you don't just mean the devices or the compute that's near those devices. You mean the everything in the value chain that's needed to deliver an end to end solution. That's that's pretty cool. [00:02:00] Do you remember when, when the idea of edge as a slice through the entire world, you know, big became something that you saw as, as a focus.

[00:02:16] Malini: Why colleagues, when I was at UMass, during my PhD, he created a little internet device. It was fitting on a pinhead and that's, you know, my God, the hardware technology has become so small, so easily accessible and affordable that you can stick it on anything. You know, it's like a little dark, like now that talking to the internet type of thing.

[00:02:37] So for me, it's like, you've pushed, compute right in. To your remote environment. And it's just everywhere. If you want to like maybe monitor earth vibrations, you know, some seismic region, region only part that's going to be hard and troublesome is the rocket dies boxer because you can go to these places easily.

[00:02:59] So for me, [00:03:00] we brought a really, I know it sounds clugey. To the edge deep into your environment. And that's the power of what's happened in the last 30, 40 years because technology has advanced so much.

[00:03:14] Matt: Do you remember when it was that you saw that device?

[00:03:16]Malini:  well maybe 30 well maybe 30 years ago that was long, long ago.

[00:03:19] Matt: Oh really? Oh, that's really interesting. You know, cause I, cause I did some as part of the state of Utah project, I did some research and I tried to find the first reference to edge computing in the academic literature were any literature actually. And I could only trace it as far back as the, the paper that the guys at MIT wrote before they started at the mine.

[00:03:37] So Actimize which was at 90, 1990, something

[00:03:41] Malini: yeah, that's around the time during my PhD days and we didn't use the word IOT, we didn't use the word edge. Then it was like all the enabling factors. Here's this little pinhead internet connected device, you know? And my colleagues named Dennis still three. So [00:04:00] I've got this last name where the tree. So he did this little thing.

[00:04:03] He's had a couple of startups and

[00:04:05] yeah.

[00:04:06] Matt: So you got your PhD in computer science. Did you immediately go into the industry or did you do something

[00:04:13] Malini: Yes. Back in those days, I got a PhD in machine learning and AI, it wasn't such a hot topic. Then there weren't that many industry jobs. Then there weren't that many Academy positions then. So I said, okay, let me

[00:04:28] Matt: Yeah, it was hot in 1987 and then it wasn't hot until about eight or nine years ago.

[00:04:36]Malini: So, so in those days I took a job at GTA labs, which has since become Verizon. And basically we did some kind of pattern matching. They had these telephones that were, you know, in the little boxes and kiosks and people might cut cables and you might lose and you know, your connectivity, doesn't an outage.

[00:04:55] So we were trying to like, just use patterns like these alarms, where they [00:05:00] spatially temporarily to kind of say,

[00:05:05] Matt: So you are doing predictive maintenance with, with edge devices before anybody even given them a name.

[00:05:12] Malini: Yeah. It does edge devices. Didn't communicate. So that was their communication.

[00:05:17] Matt: That's really great. That's really great. And then, and then did you go straight to Intel? Was that

[00:05:22]Malini: no. So it's a little journey to get to Intel. I worked at, GT labs. Then I worked at nuanced does speech recognition and their product, you know, dragon software got acquired. I mean, I'm used by multiple other companies. There was the early speech recognition stuff and they had like 94% accuracy. When you like read a few paragraphs, maybe 10 paragraphs.

[00:05:47] Matt: it's not as good as the tools are today, but it was certainly was

[00:05:50] amazing. It was

[00:05:50] Malini: Yeah, but then in a limited vocabulary set, like say you're a radiologist, they give you like 10 little paragraphs, you read it 94% accuracy, which [00:06:00] reduces all the time that the doctors have to spend transcribing, texting and recognize my voice. And that was one of the criteria and like, okay, I can join this company.

[00:06:10]Matt:  That's great.  That's really interesting. I wonder, I wonder if they trained it with, with accents, do you know.

[00:06:15] Malini: they did. And then they had a version that did something called English, like Hindi and English for the Indian environment. So, yeah, they were pretty ahead. And before the nuance product, they had tried some Google voice recognition. It was so bad. We said, forget it. We will never get home using this two directors.

[00:06:33] it's

[00:06:33] come a

[00:06:33] long way.

[00:06:34] Matt: Gotcha.

[00:06:39] Malini: Yeah. After nuance, it was in Dell. And when they invited me to join, it was like, wow. You know, I have done my PhD. I used to, you know, DEC had this alpha computer and before that it was like TIS Explorer. And they were like honking big machines. The TA Explorer was the size [00:07:00] of a large microwave or a dorm fridge.

[00:07:03] It would heat up by 10:00 AM in the morning. And then I'll say, okay, the fan. Points to you by 4:00 PM would heat up if I didn't put AC. So this was like the fussy kid in the house, you know? And so when Intel gave me the chance to work with them and then. Exactly like I'm heating up, like, you know, I'm going to collapse steps.

[00:07:24] So when Intel called me, I was like, wow, I have an opportunity to work on processes that are faster than the multi-core all the things that were like holding me back when I was doing my PhD to do machine learning AI. So it was, it was a dream come true. I knew really nothing is how I felt. And then you realize like a lot of hardware is like a lot like software, but it runs much, much faster.

[00:07:47] And you do cashing on your software. You do cashing on your hardware, you do minimal instruction sets your larger ideas and functions, you know, in your software. So it was an awesome journey.

[00:08:06] [00:08:00] Yes then that's the origin of that cloud and virtualization. And that's what brought me to open source.

[00:08:15] Matt: That's interesting. So, so your, your first exposure to open source in any meaningful way was when you were at Intel.

[00:08:21] Malini: Yeah. I mean, they'd never done it before. I might have used it before I had used, like when I was working more than 15, 16 years back on an IOT edge project, it was for remote monitoring and management. I use like open Swan and free Swan for setting up VPNs. They use Linux, but I'd never really contributed anything.

[00:08:42] And so my open source journey started at Intel and I started working on open stack.

[00:08:47] Matt: got it. I got it. And so, and now you're a VMware, and as we, we talked about the start of the show, your primary responsibility is open source with IOT and edge. do, did you join VMware to, to [00:09:00] basically be a liaison to the open source communities?

[00:09:06] Malini: Nearly four years back, I think started their own open source technology center. So here's this, you know, established software company that does virtualization. They even use a lot of open source. They have an open source project called open vSwitch, but they hadn't quite embraced open source and. Part of them hiring the condom.

[00:09:28] And I joined as a VP was to launch this open source technology center to make us more, open source, friendly compliant in its use, a good citizen and contributing back. And then duck was building this whole open source group in this company. And that's how I joined.

[00:09:44] Matt: Oh, that's great. That's great. So you were there right at the beginning,

[00:09:50] Malini: I took a little time to join, I think a year and a half. I was like, yeah, I'll think about it. Yeah. Okay.

[00:09:58] Matt: Yeah. Well, I mean, [00:10:00] so I mean, you're a scientist with pretty serious, you know, technology chops. what led you? It seems like a little bit of a, of a, of a career pivot to, I don't imagine you write much code anymore, or do you, do you write a lot of code?

[00:10:13] Malini: No at VM-ware. I have that opportunity again.  and it's nice. It's

[00:10:19] Matt: I miss it.

[00:10:20] Malini: and then you like. Yeah. When I was at Intel, early on, I did write a lot of code when it first things I worked with Ana performance, you know, solution for Intel chips. So you have multiple cores, but they are connected by a ring or a mesh or something.

[00:10:38] So they can communicate, send data between each other and things like that. Not what frequency should you operate this? I mean the faster you run it, you consume more power, you heat up more, et cetera. So you can actually modulate this frequency. Cause now there's technology to control it through software, but at what frequency do you do it?

[00:10:57] And if you have like eight calls or 18 calls or [00:11:00] 64 cores, you can just say, I'm going to change them all at the same time. You have to watch what's running on each of them. And then maybe do them in little chunks at three, three, three, so that you have the latency interrupt, handling latency not messed up.

[00:11:13] So I did a lot of software and then when I moved into. You know, open source and cloud. I got a chance to do some more software, but with time started, reducing many are like 78 people then doing code reviews, VM, where I can do little bits of software, I can prototype things and then somebody else might even throw it out and say, I can do it.

[00:11:34] And that's normal too. So. I've gotten back into writing some Python code I've gotten into writing code. I can still look at Java and understand it. So, yeah, so I get to do everything now because I don't have to lead a big deal.

[00:11:50] Matt: Yeah, that's nice. And, so can you describe to me your job today? Like what you, how you allocate your time? So it sounds like you have a lot of freedom. Like what, what, what are, what [00:12:00] do you see as your objectives and what are you trying to accomplish on a day to day basis?

[00:12:04] Malini: So right now, because edge IOT focused, it's not just going to be like. Like, you know, I've worked with a project called ethics Foundry, so it's writing a little better code or maybe make X Foundry easier to use or better documented or whatever. But at the end of the day, your edge is not going to be one edge.

[00:12:26] It's about lots of these, so that it's going to scale. And then how do you monitor them? How do you manage them? And might you use something like Kubernetes to run at the edge? So you have, you know, high availability of resiliency. Like you can. Hope to get away with just one single server, one single raspberry PI at the edge, because how do you update it?

[00:12:47] What if it feels, I mean, you have to plan for that sort of thing. And that's why it makes sense to think of your edge as a cluster of machines. It, depending on how valuable it is, or, you know, how [00:13:00] much is writing of your economics on that? And if you're going to make it a little plastic at the edge, then maybe you want to use Kubernetes because the most popular kid on the block, it has the large community working towards it.

[00:13:13] A lot of people know those APIs. So then you want to think about your edge application running in such a cluster. And that's some of the things that we. We are, we're now focusing on like, what if I'm running one of those things left edge projects. Elephant is the Linux foundation that you are very well aware of.

[00:13:32] How do I run that maybe little Coubernetties cluster? How do I monitor them? How do I manage them and, and which open source projects in this palette of solutions can I use? And is there any gap over there? So yeah, one of the things I firmly believe is is you can write open source. But you're dreaming up things that you need features you need.

[00:13:53] And the best way instead is to have a problem that you're trying to solve and say, look, I can't do that. I need to do it. [00:14:00] And, and that brings out the gaps and the problems you need to address first. So it's not like a solution and finding a problem that it fixed, but having a problem and then a solution

[00:14:13] Matt: Is there any particular problem you're working on now

[00:14:15] Malini: So right now,

[00:14:16]one of the things we got interested in, and this is because Pat Jill singer VMware's CEO, this is what are we doing in sustainability. And as part of that, we have a microgrid project at VMware and we have two buildings with photovoltaic cells,

[00:14:35] Matt: What is a microgrid for listeners that don't know.

[00:14:38] Malini: So the grid is another word just for saying your whole electric grid.

[00:14:43] Okay. That's what brings in the power from a power generation plant. Be it a nuclear power plant or coal firing plant. And it comes down transmission lines right up into your house. And. Typically one way flow. It comes from the power plants through those [00:15:00] transplant, into your home, into your office, whatever.

[00:15:03] So now a micro grid is it's like a little Island that's homes, offices, whatever, these consuming end points, but there might possibly be some. Generators over there, maybe a winter by maybe what a world take cells are some batteries, and it can disconnect from that main grid and its transmission lines and still operate as a company plead unit as a little Island.

[00:15:28] That's separate. Okay. And the future of the grid is not only do you take power from upstream utility, maybe you can sell back or return. And why is that important? Because think of this whole electric grid as like, you know, power coming in and power going out, like, you know, a water pipe and a tap. And then if the tap is not open and what is just coming in, it can just bunch up then explored.

[00:15:55] So there was always this issue of flow. And then. And the government [00:16:00] also requires it to be resilient, secure, and then meeting the needs of the customers. And that's not easy. And even with all these renewables, your sun comes up in the day and it's not that night, but people are using power at night. So there is a difference in supply and demand.

[00:16:18] And that brings the need for, you know, these carbon rich fuels like coal. To meet that, that difference of consumption and supply. So what if you could, in the smart grid, save some energy and battery. When that sun was there, the wind was there and supplied back to the grid. Doesn't have to bring in those coal firing plants.

[00:16:40] That's the whole notion behind de-carbonization that's a perfect edge use case.

[00:16:46] Matt: So why, why so connect those dots for us? So, how does IOT and edge and the technologies, the hardware and software technologies that encompass that well, and to your point, all the way up to the cloud, [00:17:00] how does that make a microgrid possible? What's different.

[00:17:04] Malini: So let's just back up a little. Why do you need, are your DNA? I mean, why the edge in IOT? let's say you have electric vehicle charges. You have to know what's happening constantly. Who's judging. Who's not charging how much power is being used. So that's the real time aspect of this data. Then maybe you have your HVAC system.

[00:17:26] Maybe you have a data center. So you want to be tracking in real time. What's happening now. You know, if there is Slack, if there's extra power available that you could maybe give some other consumers. So you can chant power from here to there or from a battery or from a generation plan from location to location B without bringing maybe a coal fired

[00:17:47] Matt: Got it. So, so if I understand correctly, what IOT and edge enables is the ability for a software application, to. Be aware of the conditions [00:18:00] out in the field what's being used, where power is, what the battery levels are, and then to make decisions about where to route route power, where there might be a surplus of power that could be redirected somewhere else.

[00:18:11] I see.

[00:18:12] Malini: And they can be even another little commercial angle to it. Like, Hey, I can turn off my washer, dryer now or something, and I can give you public and maybe you're ready to pay more for it or something. And one of the things, another thing, and this might be because where we are located. So Pat wanted to build this micro grid to see how we can.

[00:18:32] You know, help towards the smarter grid. Yeah. And also we're right next to Stanford. What are they doing in this space? And they have Stanford bits and Watts, you know, and the whole initiative around better batteries. And what if electric vehicle charging? Becomes like a 50% adoption kind of thing. Like every other car is one of these, I think about the 75 kilowatt hour kind of little batteries, mobile [00:19:00] entities, they might be driving.

[00:19:01] And if they're not, they can give the power that they've saved up. To the grid and that can help small things, but what are the other issues? Cause there's security. There's all this distributed nature. There's unpredictable illness. I mean, we can't trusted. Matt is going to give his battery today to charge back something.

[00:19:17] I mean, you might be driving somewhere, maybe not now because it's covered, but, but you see, so there is this uncertainty there's. Question about availability, the security, there's also this sort of notion like a policy you might say, Hey, you know, during nine to five, I'm ready to share with you my power because I'm charging it, the cheapness of the sun or something, and I'm not driving anyway.

[00:19:39] So there are all kinds of things that it opens up. So, this is like a perfect diary edge you scares. And I say, why don't we use X Foundry here to collect that data from these devices? It's also very interesting space for us because one of those consuming points is a data center. So today, if you were to ask Google, how much [00:20:00] power do you consume?

[00:20:01] They won't share that with you, but they'll share with you like two years, two years back, they will spending as much power as the city of San Francisco. Yeah, that's enormous.

[00:20:14] Matt: an enormous amount of power

[00:20:14] Malini: then it, and

[00:20:15] then it was like, yeah. And it was, maybe in 2015, it was half of that. So you see in just like three years, they've doubled, who knows what they are today and there are other data centers and what can we do?

[00:20:29] To reduce that footprint, maybe workloads that are not necessary to respond immediately can move with the sun or move with the energy to another data center and you can reduce your footprint and things like that. So

[00:20:44]Matt: can you tell me what, role EdgeX Foundry, like what, what, what its purpose is and how exactly it fits into this microgrid project?

[00:20:52] Malini: For one thing, it's a project that my team and I are working on one of the merits of it. And the one that I [00:21:00] think is very significant, is it support for multiple Soutbound device protocols, like, you know, co-op and Bluetooth and just internet, IP connectivity types. So

[00:21:14] Matt: Well,

[00:21:15] and weird stuff like canvas and mod bus,

[00:21:17] Malini: Exactly my canvas is in your car. so there's so many of these. I mean, do you want to be like each person or each company or each provider? They have to build all these protocols. That's a lot of engineering effort. There's not much value in that once you have it, there's nothing really proprietary.

[00:21:35] So why don't we, you know, like, It's like stone soup come together and build it together and all of us leverage it. And that's the value of open source projects. So all boats can float. So that's one of the things that edge X does provide the southbound

[00:21:51] Matt: Yeah. You know, I recently interviewed Jason Shepherd, from the data and he, he was describing it as a, sort of like a universal API [00:22:00] to all these, these different, weird protocols that are trying to express the same thing. Like, turn this on, turn that off. My temperature is 85 degrees. but there's not a universal language for accessing them.

[00:22:13] Malini: Yeah. And that's part of like open source when you

[00:22:15] have a solution. Yeah. And then you can have some standardization, some software nowadays standards comes after software. So software is the King that makes it happen. So that's kind of where it's trying to go. And whether you represent it as a Jason object or whatever, But you still need that little conversion land.

[00:22:34] And that's the beauty. Once it's coming through the Rosetta stone, you're all talking English or something and whatever,

[00:22:39] Matt: Right. So, so I mean, obviously VMware's business, as far as I know it, isn't in delivering micro grids. so how, how does VMware see edge and IOT and how the larger business is going to interact with that?

[00:22:54] Malini: their stats, their stops, it was an IOT BU and right now, you know, it got [00:23:00] dismantled and are IOT edge work has moved into our end user computing group, which has had an, a big uptake, ever since COVID, you know, I mean, there are businesses shift, but I've seen a lot of uptick there and that has.

[00:23:14] Many many customers and it's like 65,000 and whatever, whatever. So we're trying to now build and get into edge IOT with maybe using that

[00:23:27] workspace one solution.

[00:23:29] So they have a dashboard.

[00:23:31] So you were asking. Whereas we am where with IOT and edge. So it's had starts and stops IOT and edited VMware. We had a special, you know, business unit for it. We had a solution called palace IOT center, but it's like having a big company and it's been in the cloud space and it has to pivot and it takes time for a bigger company versus a small agile company to figure out.

[00:23:58] The economics of it, [00:24:00] the scale of it, especially when you're talking in the big cloud space and those sort of prices to smaller edges, but COVID, I think has done something it's increased focus on one of our solutions called workspace. One. It has about 21 million end points that they monitor and manage 65,000 customers.

[00:24:22] And those were always like, you know, your laptops, your cell phones. They had a human behind those edges behind those 21 million end points. But IOT, I just don't always have a human over there. I mean, some wind turbines or maybe lots of

[00:24:36] humans, but now one sitting there all the time. Exactly. So, the path to IOT edge now is coming through in a workspace one they're thinking of connecting to devices.

[00:24:49] They already have hardware vendors, but it's more becoming like a proxy pass through to a single dashboard. That said the IOT edge, market's going to grow to a [00:25:00] trillion dollars. And maybe this is the entry point with the established enterprise customers, WMware is keeping his eye on the future.

[00:25:07] When, when you know, there's going to be more agile adoption and at least larger edges, like the Telogis type of thing. So

[00:25:16] that's how we're  thinking.

[00:25:17] Matt: And you mentioned earlier that, that your philosophy of edge, it's not just about literally that the last mile network, because it impacts everything. And so it does impact VMware's core business today. and there are new opportunities, you know, people are talking about running virtual machines on edge devices and, you know, as those edge devices look more like servers, they're going to be running VMware.

[00:25:35] And

[00:25:37] Malini: And then there's going to be the telco edge. The telco edge is going to be pretty beefy and it will enable applications that were chill reality and augmented reality connected cars, because you'll have to come closer to some place where the car is running, as opposed to some cloud firefighter away because of the latency.

[00:25:57] And some edges are going to be very, very beefy edges with [00:26:00] a lot of compute, a lot of knowledge, a lot of storage at that edge. Like the cottage, because one time we went camping and there was no cell phone connectivity. So you can believe in trust that you have connectivity. Some ways it take a left and take a right turn or whatever.

[00:26:15] Matt: Yeah. So, when you think about edge and open source and IOT and open source, how do you see those relating? What's the, what's the role that open source is playing and why is it so important to you personally and to VMware?

[00:26:38] Malini: Making it accessible to more people. So the early. Early adopters. The early solutions were those who did it like edge to cloud. Okay. I mean, that was like your Google nest solution. Like your Amazon Alexa solution. I say something like what's the weather. Like it goes all the way up to the cloud. It comes back and says, Hey, it's sunny in California today.

[00:26:59] So there are these [00:27:00] edge to cloud solutions and that's fine, especially if you're at home and you have your wife. But, you know, as you try to reach a larger landscape of applications and solutions that have different demands, like my nesting, that beeped a little while back. Yeah. You know, I can get out of my front door and maybe walk to the right end of the street, which is, you know, like at least.

[00:27:21] A hundred yards before it says, beep somebody came to the domain that started and I have to catch a robber or anything. Okay. So you want lower latency. So it's not going to be like from the device to the cloud, it's going to be something much closer, especially for things like temperature, sensors, and out of some safe range or a pressure valve or a nuclear power plant.

[00:27:45] I mean, you're just not going to send it to a father of eight point.

[00:27:50] So that's the thing. And then, the early adopters were companies that could do that whole edge to cloud solution. They were already in that space, but now by making these open source [00:28:00] projects and a lot of the infrastructure available, you're opening the flood Gates, many more adopters and many more applications and solutions is coming to market sooner.

[00:28:11] I mean, otherwise, who were the ones who made like. It was the at and T's and, and Berkeley and all when they was the early days of Lynette. And now Lennox is everywhere. Look at the adoption, look at the possibilities that have opened up.

[00:28:27] Matt: Yeah. And that's really interesting. It seems like a lot of technologies that we. I mean almost take for granted or at least we've allowed to become our standard light, the cloud, for instance, you know, even vaguely defined like we, you know, developers don't think twice about provisioning, an ECE to incidents or storing data and S3.

[00:28:46] And we as individuals don't think twice about using SAS applications. And so there's a moment and, and we're not there an edge. And I'm, I want to ask you about your sense of, of, of timing, but it seems like there's a moment where. [00:29:00] most of us, will, well, let's talk about the people that build these things.

[00:29:04] So the developers in the, in the engineers and the ops people that build these things, there's going to be a moment where, as you said, the edge is just part of the fabric of the internet. And we don't think of it as a separate thing. In fact, maybe we even stop using the word edge for all I know, right.

[00:29:17] It's just part of the internet. and it becomes a platform. And I think open source is what helps make it a platform? Do you have these, the standard structural components that, you know, everybody has to do this sort of like non valuable, heavy lifting, right. That everybody yeah. Asked to do. And then, you know, so like you said, an early days, it helps to have a problem you're trying to solve.

[00:29:38] Cause thing, you, you shape the domain and you say, well, you have to do these things to not these things. But at some point you have a platform that has so many general purpose capabilities, you know, Uber didn't come about because somebody said, be great. If I could call a taxi from my phone, and then they built a smartphone and GPS and ubiquitous LT and all that.

[00:29:54] No, all those things existed because the Apple and Verizon and at, and T then somebody said, Hey, I've got a [00:30:00] screening. Yeah. So where, where do you look into your crystal ball? Where do you see edge computing starting to

[00:30:10] Malini: I think at least another three, four years, it's like, I mean, like Coobernetti's is at least like, you know, 10 plus years old or even longer. And it had words and they had bark. It had something else. Linux is 30 years old. Open stack is like 10 years

[00:30:27] old. Yeah. It

[00:30:29] takes time to reach.

[00:30:30] Matt: is just more like five.

[00:30:31] Malini: Well had borrowed before, so big brother,


[00:30:34] but it takes time to reset that boring as hell kind of thing.

[00:30:39] It's so easy. You can just take it for granted type. And I think we're not yet there, but even simple things like. A few years back when you're watching a movie on Saturday, Saturday, or Friday night, especially it would a stall, it would glitch because everybody was doing something like downloading streaming on like Friday night.

[00:30:58] So today that [00:31:00] experience is much better and who knows with edge, it might even be. A virtual game, you know, like hundred players or whatever, and it might have downloaded it into a little edge storage cloud somewhere nearby. And then you don't see any glitters. You have great performance who

[00:31:15] knows

[00:31:15] you might do

[00:31:16] surgery

[00:31:16] from remote. Yeah. And also knows that you're going to do some game playing on Friday night and say, or let's see if there's a latest, greatest some movie or game that you like to play. Who knows? And that's like things like analytics, like today, when you turn on your TV and says, Hey Molly, you might like these movies.

[00:31:38] It knows it learns from how you behave. And I can put these things ahead of time. Like prefect them for you, paper stuff.

[00:31:45] Matt: Yeah. Yeah. And, and in fact, when you look at like these predictive models, I mean, I mean, you know, this very well machine learning has demonstrated that with the process we have the day and the ability to process huge amounts of data, and frankly, to generate and collect huge amounts of data, [00:32:00] really powers machine learning to do miraculous things.

[00:32:03] I mean, to learn on its own, to play video games and things. And so, you know, a lot of these predictive algorithms, you know, like the nest thermostat was a example you used, you know, and it's, I don't know how it is today, but I had one, when it first came out and it was okay, predicting, you know what, temperature, wouldn't just turn the heater on in the morning before I get up.

[00:32:20] But it was also off all the time. It was just kind of right. But you can see how they will get increasingly accurate, the more data they process. Yeah. That's really interesting. What's should ask that, that that's going to be, that threshold's going to be enabled by. Lots of low cost sensors on every device in connected to the internet in some way, you know, whether it's over five G or ZigBee to wifi or, or anything.

[00:32:47] Yeah, yeah, yeah. Or any of those, any of those protocols, just look at the list of edge X, Foundry supports and that, and that gives you the choice. Yeah. That's, that's, that's really interesting. I wanna switch gears a little bit because. I [00:33:00] understand that for a period in your career, you worked on autonomous vehicles, is that right?

[00:33:07] very

[00:33:07] very funny. Okay. But I assume you have some opinions on them. Well, I'll ask all that sort of thing. So, so I think, you know, Oh, toss field. Cause they were interesting because, you know, they certainly captured everybody's imagination and now I think it's like we're in the trough of disillusionment because you know,

[00:33:25] the.

[00:33:25] Malini: no, no. You know what it. It's amazing.   so when I was at Intel, it was one company. I won't name names. It was a small company, started by a professor and he said, you know what? I can, I can do this. I can analyze this image data. I can control the car. And he had like $10,000 worth of hardware in the back in the trunk already.

[00:33:48] Cause a pretty expensive, but $10,000, it was like humming, buzzing. And then we were going to go for a test drive. I'm not the best driver. And I was a little nervous. I mean, like I'm [00:34:00] not great. And how great is this going to be? So I sat down a little, but in like a few minutes I felt comfortable. It just took entering into roads that were

[00:34:09] Matt: assisted driving assisted driving is very different.

[00:34:12] Malini: fully autonomous

[00:34:14] Matt: on a, on like a downtown road, unlike downtown,

[00:34:16] Malini: yeah. And I live in. Yeah. I live here in, on the border of Cupertino and you know, San Jose, West San Jose, hands-off just ready to grab and stop. It drove fully. The only thing is it was a little slow to enter traffic. It was cautious, which is good when it parked.

[00:34:36] And this was like already like, Four years ago, when it parked at the curb, it was a little distant from the Cubs, something that, I mean that's four years ago, it must have moved so much further and that's going to revolutionize. Revolutionize just about how we behave. I mean like, like when Zipcar came and do you, and I need a car, always get me share the car.

[00:34:58] Can somebody just bring the [00:35:00] car when I needed? Like, I started using Uber so much and not having to deal with airport parking

[00:35:05] break-ins whatever. And I think it's just kind of amazing over this 30 year span that I've been in this industry so much has changed all that AI machine learning would not have been.

[00:35:17] Possible without all this cheap cloud computing, without all this cheap storage without, without the data that you need millions and millions of units of to do anything, you know, smart and nice. Like you mentioned about the nest thermostat. I mean, I think in like another 10, 15 years, it will be so wonderful and very different.

[00:35:37] And, and it might even liberate all the people to go anywhere and everywhere, not just get the freedom and they can't drive, so it might make it all cheaper. It might also make it cleaner

[00:35:56] Matt: Yeah. Well, I mean, I, I, I believe in a future, [00:36:00] I think it's more like 10 years off. This is what you said. Yeah. It's took 10 years off where we're going to look back and actually driving a car by hand is going to feel like a rotary telephone, but we're just going to be like,

[00:36:10] I can't believe we actually spent time doing this instead of other things that we could do on our car.

[00:36:14] I mean, we can have a picnic on our car and take a drive. Right? I mean, I think it's gonna completely change how we relate to vehicles. And you're right into ownership. Like we don't all have to own cars necessarily. We can just like lease them when we need it. Like cloud it's a cloud. Yeah. It's like cloud service that Amazon, right?

[00:36:30] Yeah. That's really interesting. So, so has autonomous vehicles relate to edge computing? So. there's a lot of topics here. but one of them is, what workloads can run on the car and what growth can run on the infrastructure. And obviously deploying an air brag, emergency breaking that that tight control has to be in the car.

[00:36:48] You just can't have anything, but a very, very discreet, reliable with lots of backups systems doing it. but as you said, you know, even even $10,000 for the equipment, let alone the quarter million dollars of equipment in a [00:37:00] Waymo car. It's cost prohibitive for mass markets. that shirt will be cost produced and, you know, maybe, but, even in those cases, right, the best LIDAR in the world can't see around a corner.

[00:37:11] And so to the degree that. Decision support information can come from the infrastructure or to the degree to reduce costs. I can offload maybe latency sensitive, but not life critical and not latency critical workloads to cloud computers that are at the edge, essentially. How do you see, how do you think about it?

[00:37:30] I mean, that's kind of how I think about it is that there's going to be this set of trade offs something's happening in the car stuff. We're on the cloud there's would be cost, equations, benefit equations.  

[00:37:37] Malini: I really feel everything will have to be in the car. I mean, that experience when I was going on a camping trip in Canyon territory and not having connectivity, my kids were bored. Of course, at the stars,

[00:37:59] Matt: Well, you don't have to [00:38:00] rely on it, but again, look at the turning around turning corners example,

[00:38:03] looking

[00:38:04] around

[00:38:04] Malini: no.

[00:38:05] So, so, so I think. Maybe you don't like you'll have to have a little something peeping out around the corner or go slower because it's a corner and you don't have visibility. Like this is, think about the Holy grail is getting to a point as a good human driver. Okay. And it's going to definitely be better than a good human driver.

[00:38:24] There's no fatigue. There's no drunken driving all that. So just like you wouldn't overtake a car on a curvy road. You know, with all those dash lines and all this car is also not going to do it. It might even have to slow down.

[00:38:38] So I think it's going to become like that.

[00:38:40] Matt: That's interesting. Yeah, that's, that's interesting heuristic. I, think I, I'll stop my head. I agree with that, that, two, to reach a quality with a human driver.   we probably want that function on, the car.

[00:38:52] although it'll, it'll constantly, you know, the model also potentially constantly be upgraded and things like that.

[00:38:57] but I do think like, [00:39:00] you know, you look at some investment mental things, like w one of the things we're just improving track that flow and why it's sure people have been. I envisioned these networks of cars that collectively beehive themselves into, into flow, but it also makes sense to say, look, there could be some, some local.

[00:39:19] Alright, I'm going to say centralized. It's not really centralized, but Metro centralized processes that can coordinate all the cars, all the lights, all the freeway patterns knows when people get off work, knows these things. And again, could see around a corner. Cause the benefit of seeing around the corner is, you know, when you're, when you're driving in a car, And you've got a human driver.

[00:39:38] We put our seatbelts on and we feel safer with the seatbelts. We're looking out the window. We want to see what's going on. You know, we've got people that are telling you how to drive because there's lots of things going on that potentially you don't see. And that's the worst. Cause something kind of steps out in front of you and you have to do like a drastic braking action.

[00:39:54] Well, imagine. We're having a picnic in the back and we're, none of us are looking outside. None of us are, [00:40:00] you know, and then there's a sudden break and all the food goes flying. Right. And so you can imagine that, well, what you really want to do is you want to have some knowledge about that car that you can't see, or that pedestrian you can't see and gradually break up until the intersection.

[00:40:14] So I can it's

[00:40:15] it's. Yeah, but it's but I can also see how that, that isn't the primary solution that you need to solve.

[00:40:22] Malini: But, but that is like the best situation that you can be totally agnostic of. What's happening. You're safe. Maybe the say PPP, you know, brace, there's something happening. I in Massachusetts, a deal might just jump out,

[00:40:40] you know, see fast. That's they call these Lake levels. That's L four. That's pretty far off.

[00:40:51] And, and also there's ethical considerations. Who's at fault, there's a software

[00:40:55] issue, blah, blah,

[00:40:56] blah,

[00:40:56] whole bunch of

[00:40:57] Matt: Yeah. A lot of it's going to be just legislation. [00:41:00] Yeah. Insurance and legislation and all that. A lot of it should things to, to work out. to to work out.

[00:41:04] Malini: It's just a future. And then I also wonder how companies will feel. I mean, this software can run on any, any car at that point, and they'll just be some differences on how fast the data comes from the canvas, how quickly the brakes respond, how well they respond. And maybe in Carre, you have to give it like one second and another car, maybe half a second for to do it.

[00:41:27] So there'll be some tuning per car. And that can also be learned just like the nest thermostat learning very leaky, not well-insulated. I have to start heating it sooner. It won't hold the heat that long. So they're going to be some analogies like that.

[00:41:52] Matt: with that seems to be getting people to not trust it completely autonomously. [00:42:00] I'm not, I'm not that either. Yeah, it'd be, if you looked away, so, I wanna ask you a couple last questions before we wrap up. one of them is, you know, you're deep into the edgy ecosystem right now, and I'm interested in knowing, what is changing the fastest from your perspective?

[00:42:19] Like what's, what's,

[00:42:27] Malini: Just the number of applications. People are coming up with it. It's like that crowd sourcing the ideas. It's so fun. It's just like that thing you mentioned, you know, the cell phone came and people started saying, Hey, what all can I make possible? Like yesterday my son was making peach crumble. This is, recipes for peach cobbler or crumble or something.

[00:42:49] And you know what, in the old days you had like a recipe book, somewhere, your bookmark, something your grandmother had done, it's so different. And if you didn't have the book and they didn't find it and blah, blah, [00:43:00] blah, now it's in everyone's hands, you know? It just as a future that was not imaginable before and it's happening.

[00:43:07] And for me, that's what this edge thing is. And last year we had the good fortune to work with the university of Santa Francisco, computer science students. We went there, a bunch of other companies came to an, a pitch. You know, this is what I work on. Would you like to do a project on this? And it was overwhelming.

[00:43:24] We had 10 teams wanting to work with us and we had just two of us there to be like, you know, guides can't do it. We'll do one undergraduate and one, you know, graduated team. And then we said, okay, let's do three. It's kind of hard to say no to so many people. And what else excited. And the students came up with such cute things.

[00:43:43] Like we said, here's edited. So some of the ideas, and we kind of pushed him some directions to, you know, seeing where their minds were. One young woman and her team basically said, You know, I work with shelters. Could I have a camera and add that camera, you know, look at your [00:44:00] face and say, are you one of the inmates that are allowed to come in?

[00:44:03] And this is a rolling population at the shelter. Maybe this month it's ABCD and next month, it's somebody else. And. So you have to able to recognize a face and you have to also say who's in who's out. And, and also the record who came and went just for that safety. And she wanted to do so at Jax was free and had smart brains, his little team of four, they thought of it.

[00:44:26] They also got their hands dirty with a little bit of machine learning that saved the model at the edge. They trained it somewhere in the cloud and pushed it. The cloud is in this case, a laptop. But they realized that the camera's like $35 a raspberry pie, another $40. And they had this personalized recognition system, another team, we said, think about the smart grid, because I was already thinking about the smart grid and you guys have looked on 'em.

[00:44:55] Blockchain. I want to sell. I want to buy and you know, [00:45:00] all of them have to go to some kind of marketplace or is it decentralized? How much do I have who's bidding, who's ready to pay what? And you have that commerce going. And they did that project. That was cool. And they had so much good knowledge from the blockchain teacher, but this was the practical use of itsy.

[00:45:23] tape recorder now or whatever. Yeah, exactly. Does this, like, you know, it was starting with Uber and then another team never. So cool. So you have maybe a thermostat from one company and it talks Bluetooth and something else from another company and your television. And right now, even my washer dryer are IOT enabled. Is it, but what if Matt comes home and he wants to kick his feet up, have a drink, feed the dog,

[00:45:54] cook dinner. So what if there's a. [00:46:00] The smart flow. You can language some JS on some graph and it knows, okay, it's evening, it's six o'clock Matt's going to come home and it automatically does all this. And these kids did it. These young people, And the first month or so they did no code were first. Like how do we abstracted, how to be able to give these flows and how do we say it's a rainy day floor evening flown, which should I look at the light thing?

[00:46:28] It was so awesome. And. And that's for me, what's awesome. The interest. And right now they're like conferences with tracks. For IOT, there was a conference call, you know, open networking something, and they made it once for open networking and edge. So there's a full track these days for IOT and edge.

[00:46:46] There's the in T track and that's going to enable much more. So. From a conference perspective from number of people come for, there's so much updates. That's, what's changing for me that it's becoming so affordable and people can experiment with [00:47:00] $35. Asprey buys, little center, little cameras,

[00:47:03] anything is possible with all the smart brains and ability to do things now.

[00:47:08] Matt: That's really neat. And I can tell you're passionate about you just light lit up when you talked about that. That's really neat. And that's actually a nice segue into my next question. And I know this is shown as computing, but, The role of women in technology is, near and dear to my heart. I mean, any minority in technology, is there to have a heart.

[00:47:24] And so  you've had a long career and you, or a scientist and, you know, you've said so. So what,

[00:47:30] what was your experience like and what advice do you have? Two young ones that are maybe considering a technology careers.

[00:47:38] Malini: You know, we have got to catch them at the school level.   lose so many people and it's surprising we don't lose them in India. At that level, But in the U S there's a larger loss and I don't know why. And it's surprising and it's sad We have lots more [00:48:00] women going into engineering and science and

[00:48:02] technology nearly 50.

[00:48:05] Yeah.

[00:48:05] It's phenomenal.

[00:48:07] Matt: a system we should model. I wonder what that

[00:48:09] Malini: So many women engineers come out of India and. And of course in my generation, like some of my colleagues decided not to get mad because you're going to be perceived as less committed, et cetera.

[00:48:21] I just happen to be lucky. And I did go the marriage and the children, and I have two children, but another good friend of mine who is a, you know, quantum mechanics. Scientists. I mean, she works at BRC. She decided to not go the family part and, and be taken seriously as a scientist. So it's kind of sacrifices and those are less now, but they're still very significant.

[00:48:46] Last week we had, you know, A panel talk like how do you juggle families, especially in these COVID times. And it's amazing hearing some of the stories. Some women had to hide their pregnancy still. It was pretty, you know, you can't hide it anymore. The baby's [00:49:00] there type of thing. And even then it was. It was hard in those days, even in terms of maternity leave, you might get six weeks and nowadays we even have paternity leave.

[00:49:09] So the industry has move so much to make it possible for women to have a career and raise families and not just give up that part of their life. Another thing is. Don't let anybody tell you, you can't do it. You know that you're not smart in science or math, it's all about perseverance working at it. And then, you know, those solutions do come out.

[00:49:32] It is an aha moment when it clicks in your brain and we're all wired smart. Maybe we have some differences in our cognitive skills. Somebody has special skills. Somebody has language skills, somebody has emotional skills, but we need all those skills. And I think. You know, as an industry, we beginning to value women more,

[00:49:54] leave is a big, big step in the right direction.


[00:49:57]When I was going into engineering school for computer science, it was like a very competitive area. And several thousands did it in India. This was for a top school, the top 20 people, the Dean of engineering wanted to meet. So there were two of us, two women, we went there and he was like, you're a woman because you can't tell from the last name I said,

[00:50:26] yeah. And he was. Belligerently and you know, you're a waste of resources is a very good school. It's a government school. It's it's, you know, you just sit at home,

[00:50:36] Matt: is cause you,

[00:50:36] Malini: I'm a woman. Yeah. You know, in the top 20, in the 10,000 or whatever. No, no, no, not the first day. It's still, his point was you're a woman you'll get married, you'll stay home.

[00:50:52] And you know, some man would have finished the course would go take a job with feed. A family would be useful to society.

[00:51:00] [00:51:00] And both of us women finish the course, neither of us gave it up. So it didn't matter if he'd been in the face.  but, we both studied. We both did well with. Both. I worked in industry.

[00:51:11] We have kids who are in the industry now. So that was how we were taken very seriously as resource waste. And when we were interviewing, if we got a job, the guys would say, Oh, well, you know, you're a woman, so you got it. Or your dad works here and you got it. Some rubbish and the same men today. Now we're back in touch with them with WhatsApp.

[00:51:33] They're way different. Now they've made sure they have spouses and they take women more seriously. They have daughters, they're proud of their daughters. So things have changed. And these people now are VPs and senior engineers and entrepreneurs. So it's changing, but it's slow.

[00:51:48] Matt: Yeah. And it's, it's a multigenerational problem. I mean, Justin, the story you told that the reason you went into technology is because your father was sort of in that space. And of course it sounds like potentially one of the reasons [00:52:00] your children went into technology is because in that space. And so, yeah, as they, as, as these attitudes change and as we, as more people are educated and they have children.

[00:52:08]  we're going to be seeing a transformation. Well, that's a really optimistic point of view. It's nice. It's nice to hear. So Malini, you so much for joining us on this podcast.

[00:52:17] This was a great conversation and,  I

[00:52:19] look forward to seeing what you and VMware and edge Foundry do next.

[00:52:23] Malini: It was fun chatting with you, Matt.

[00:52:26] I had a blast. Thank you