This episode of Over the Edge features an interview between Matt Trifiro and Stephen Goldberg, CEO at HarperDB. Stephen is an established thought leader in the IoT space, with previous professional experience as a CTO and CEO of startups, holding several roles at larger organizations like Red Hat, and leading digital transformation projects at a number of Fortune 500 companies across many verticals.
This episode of Over the Edge features an interview between Matt Trifiro and Stephen Goldberg, CEO at HarperDB. Stephen is an established thought leader in the IoT space, with previous professional experience as a CTO and CEO of startups, holding several roles at larger organizations like Red Hat, and leading digital transformation projects at a number of Fortune 500 companies across many verticals. He’s been published on sites like Tech Target and quoted in a number of articles and publications like Forbes and ZDNet, as well as being a speaker at IoT World, SAP Sapphire, and Salesforce.com’s Dreamforce. Stephen holds 2 patents and received his Bachelor of Arts from Trinity College-Hartford in 2006.
In this episode, Stephen talks about being a self taught programmer that initially wanted to work in anything but technology. He explains the process of co-founding HarperDB to deal with the rigidity and complexity of databases, and how the company makes it easier to globally distribute data faster. Stephen also discusses how bureaucracy, in many ways, is the biggest challenge to innovation and the adoption of new technologies.
---------
“For example, while we felt like the iPhone is an extremely complex device, technologically and internally, it exposes a very simple interface to the world that even a child can use. And, so our thought process was that a database should be the same for a developer. It should be a thing that a developer can just sit down and code, and ultimately, we have built far more than a database in the end. But, that was kind of the premise of what we wanted. We wanted something that would scale with you as your application grew, that made your life easy.
The bureaucracy, like the speed at which people move, their ability to think about how things work, that is now the one that I think is the biggest challenge is that this stuff is now and it's here. And, if you really wanted to do it, you could, but I think it's also just changing how people think about stuff. So, like you know, willingness to adopt new technologies, willingness to adopt new architectural paradigms, not trying to bring sort of the same cloud centralized sort of cloud model that you implemented to the edge, because it's not gonna work, because it doesn't scale.”
---------
(02:00) Getting started in tech
(05:00) Starting HarperDB
(08:35) What makes HarperDB unique
(16:00) Understanding HarperDB
(22:00) Use cases for HarperDB
(26:45) HarperDB’s cloud product
(35:15) New product release
(36:30) Internet of Things
(46:00) HarperDB in gaming
(48:15) The future of tech
--------
Over the Edge is brought to you by Dell Technologies to unlock the potential of your infrastructure with edge solutions. From hardware and software to data and operations, across your entire multi-cloud environment, we’re here to help you simplify your edge so you can generate more value. Learn more by visiting DellTechnologies.com/SimplifyYourEdge for more information or click on the link in the show notes.
--------
Connect with Stephen on LinkedIn
[00:00:00] Narrator 1: Hello and welcome to Over the Edge.This episode features an interview between Matt Trifiro and Stephen Goldberg, CEO at HarperDB. Stephen is an established thought leader in the IoT space, with previous professional experience as a CTO and CEO of startups, holding several roles at larger organizations like Red Hat, and leading digital transformation projects at a number of Fortune 500 companies across many verticals. He’s been published and quoted in many technology publications, as well as being a speaker at IoT World, SAP Sapphire, and Salesforce.com’s Dreamforce. Stephen holds 2 patents and received his Bachelor of Arts from Trinity College-Hartford in 2006.
In this episode, Stephen talks about being a self taught programmer that initially wanted to work in anything but technology. He explains the process of co-founding HarperDB to deal with the rigidity and complexity of databases, and how the company makes it easier to globally distribute data faster. Stephen also discusses how bureaucracy, in many ways, is the biggest challenge to innovation and the adoption of new technologies.
But before we get into it, here’s a brief word from our sponsors…
[00:01:21] Narrator 2: Over the Edge is brought to you by Dell Technologies to unlock the potential of your infrastructure with edge solutions. From hardware and software to data and operations, across your entire multi-cloud environment, we’re here to help you simplify your edge so you can generate more value. Learn more by visiting Dell.com for more information or click on the link in the show notes.
[00:01:42] Narrator 1: And now, please enjoy this interview between Matt Trifiro and Stephen Goldberg, CEO at HarperDB.
[00:01:50] Matt Trifiro: Hey, Steven, how
[00:01:51] Stephen Goldberg: are you doing today? I'm doing quite well. I'm excited to be here. Thank you for having me. Yeah, that's
[00:01:55] Matt Trifiro: awesome.
[00:01:56] Matt Trifiro: So one of the things I like to ask all my guests, cuz I, I get a different answer [00:02:00] from just about everybody is how did you actually get
[00:02:02] Stephen Goldberg: into technology? Mine was not a, a well thought out plan. My uncle started a software company in the early nineties. It was a ISP billing software company when the internet was sort of first emerging for consumers.
[00:02:17] Stephen Goldberg: And my father thought it'd be a good idea if I flew out and spent this some time over the summer with him and kind of went from there and just always. Enjoyed programming and playing with computers and various sort of side jobs. And I didn't want to do this. I wanted to do anything but technology, but, and I studied something else in college.
[00:02:38] Stephen Goldberg: Yeah. I didn't want to be a geek. What'd you study in college. I actually studied theology and I focused on like mysticism in Islam and Christianity and Judaism. I, myself am not religious, but I just found sort of the, uh, no, that is
[00:02:53] Matt Trifiro: so cool. I St I studied philosophy linguistics and cognitive science and college, so I totally, I totally get it.
[00:02:59] Matt Trifiro: I totally get it. [00:03:00] Yes. But you did mention programming, so did, are you self-taught.
[00:03:04] Stephen Goldberg: Totally. Yes. And so is Kyle, my CTO, you know, we have a few folks with CS degrees, but the vast majority of Harper DB in my company are self-taught programmers. So
[00:03:14] Matt Trifiro: help me understand was the work with your father that was before the theology
[00:03:18] Stephen Goldberg: study, right?
[00:03:19] Stephen Goldberg: Yeah. So when I was I'm 39, so yeah, I was like 13 years old. It was like during the very early nineties, like when, you know, you dial up and before even AOL. Programming on a Unix based operating system and learning kind of HTML. And while self-taught, I was very lucky because there were, you know, 20 or 30 people who helped, you know, me learn in that room and who were some of the people who helped create the internet.
[00:03:44] Stephen Goldberg: So I had a huge advantage, so not sort of self-taught in the driven way, but I didn't have any formal.
[00:03:51] Matt Trifiro: learning through osmosis. Yes. And deadlines, exactly deadlines. So you came out of school with a deep understanding of mysticism [00:04:00] and religious studies. You said specifically you wanna do anything but technology, what did you end
[00:04:04] Stephen Goldberg: up doing?
[00:04:06] Stephen Goldberg: I graduated college in 2006 and then I went to Europe and I bummed around for a while. Then when I came back, I tried to get a job, but the market was collapsing. I applied for hundreds and hundreds of jobs in finance and project management in real estate in a million different things. And I couldn't really get hired anywhere.
[00:04:29] Stephen Goldberg: And so a family friend worked at a software company called I imaginations, which did patient education software. They hired me and I went and I worked there as an intern and quickly that transitioned into me working at red hat and from there than kind of many other things, but it was really more, just a function of, I knew how to program.
[00:04:51] Stephen Goldberg: I was technical and it was really the only valuable skill that I had. That was good enough to get me through kind of the market conditions at that time. [00:05:00]
[00:05:00] Matt Trifiro: And how did all this lead up to Harper DB, which is your current company, right?
[00:05:04] Stephen Goldberg: After I left red hat, I moved here to Denver to work at a consulting company, and I was solving challenges around large scale integration, large scale enterprise architecture challenges.
[00:05:17] Stephen Goldberg: And then I started my own consulting company and I hired Kyle burn Hardy who's my co-founder at, at Harper DB. He was my first customer actually at my previous consulting company. And he and I. We wanted to build a product, but we tried to run a consulting company at the same time and anyone who's ever done that, it's a really bad idea unless you're extremely disciplined, which we were not.
[00:05:39] Stephen Goldberg: And so we just ended up being a consulting company, but we wanted to build this sort of middleware integration platform because we felt like it was kind of stupid. And we were getting paid a lot of money to go the same, you know, company after company and do the same thing over and over again. That's where Heroku came from.
[00:05:55] Stephen Goldberg: Exactly. We love Heroku. And we think we have taken a lot from [00:06:00] Heroku and we think of Heroku as a sort of a similar thing to what Harper DB is today. And we're big fans of it at the time. We just were frustrated by that, but we couldn't figure out we were stupid. We were young. We didn't know how to run a company.
[00:06:14] Stephen Goldberg: And we were really good at solving technical problems, but our customer, a software company in the bay area, they asked us to join them and they sort of aqui hired us. And we were very happy to not be consulting anymore and not running a consulting company cause we were bad at it. and so we went and joined.
[00:06:35] Stephen Goldberg: They were a large scale sports and analytics company focused on big data in social media for, you know, football and live television. And we just had this extremely complex. Backend technology to power. All of it. We ended up actually buying a Cray super computer to power it because our AWS bill was so high.
[00:06:56] Stephen Goldberg: We realized it'd be cheaper to buy a Cray super computer . [00:07:00] And we did that and we were just very frustrated by how complex that backend was and how much of our resources was being used to maintain it. And sort of some of the similar problems we'd run into as consultants around integration. And at the time we had thought that was a middleware problem, but ultimately we realized.
[00:07:17] Stephen Goldberg: The rigidity and complexity of databases. Made development really hard developers just want to code. They don't want to figure out what their database scheme should look like, what should be indexed, where things should be deployed. And so we ultimately left that company to build Harper DB because we felt like.
[00:07:36] Stephen Goldberg: While the iPhone, for example, is an extremely complex device. Technologically internally, it exposes a very simple interface to the world that even a child can use. And so our thought process was a database should be the same for a developer. It should be a thing that developer can just sit down and code and ultimately.
[00:07:54] Stephen Goldberg: We have built far more than a database in the end, but that was kind of the premise of what we [00:08:00] wanted. We wanted something that would scale with you as your application grew that made your life easy. And we came up with an idea over a year before starting the company, but we thought that someone else would solve the problem before us, um, that someone much smarter who went to Stanford and had a PhD in data and computer science.
[00:08:18] Stephen Goldberg: But that didn't happen. And so here we are. ,
[00:08:21] Matt Trifiro: let's talk a little bit about the problems that you solve. The first thing that came into my mind, and it's gonna sound snarky, but I'm smiling when I'm saying this, which is like the last thing the world needs is another database. Right. And what I mean by that is there are a lot of databases.
[00:08:35] Matt Trifiro: And so if you're going to sell and compete in that market, you have to be doing something different for a specific set of use cases. Can you tell us what Harper DB does? That's truly unique and which class of developer says, oh my gosh, I'm so glad you built this
[00:08:51] Stephen Goldberg: many, many people said those exact words that you said in the, when we first started the company and it was a little disheartening, to be honest.
[00:08:58] Stephen Goldberg: Now, I think it's funny. And [00:09:00] I kind of agree with you.
[00:09:01] Matt Trifiro: Well, there's plenty of database companies that have been successful, right? So it's not like, or it began and ended with Oracle, right? I mean, cockroach is the most recent example.
[00:09:09] Stephen Goldberg: Absolutely. Yeah. Or snowflake or whatever. One of these used cockroach snowflake, you know, Mongo, DB there's many men flood.
[00:09:16] Stephen Goldberg: Yeah. All. Yeah. So really we did start the product because while all those products exist, You know, Mongo went after this idea of, Hey, let's just make developers' lives easier, but it's not that easy. And it doesn't really support a lot of use cases like SQL and analytics. And you mentioned Heroku, you also can't really easily just write your whole application in, in Mongo DB.
[00:09:41] Stephen Goldberg: So that is why we started the company and developers do love Harper DB. For that reason, there's a hundred thousand of. Currently using it, you know, to build their applications, but that's not how we make money. To be honest, like the way we make money is a very specific problem that we do solve. And it was totally accidental.
[00:09:58] Stephen Goldberg: And we didn't like plan [00:10:00] to solve this problem. We built Harvard to be in no JS, which is a web application framework, which means that it's not, it doesn't function like a database. Would it functions more like a web app? And as a result, it is extremely horizontally scalable. You can run it anywhere in the world.
[00:10:17] Stephen Goldberg: And it's extremely fast and lightweight. And so what we found was the major problem that we are good at solving is when you are a large company that has a globally distributed user base, let's say that's a gaming console. Let's say that's a large scale app, like mobile app, you know, whatever that might be.
[00:10:35] Stephen Goldberg: Let's say it's streaming media and your users are all over the world. There's lots of ways to get your application and your API. Globally distributed, but ultimately those things all call back to a centralized database called a origin database. What that does is it creates a physics problem where you're fighting the speed of light to get from, let's say, BU air sort, Tokyo,
[00:10:58] Matt Trifiro: let alone the [00:11:00] network cops in congestion and, you know,
[00:11:02] Stephen Goldberg: Exactly.
[00:11:03] Stephen Goldberg: Once you add those in, then it really, you know, you get about 60% of the speed of light. And so that is a major problem. And so what Harper DB can do is you can put it anywhere on planet earth. It can globally distribute your data locally to where your end users are. And what that means is then I in Denver am interacting with a node of Herper DB that is in Denver, at a 5g pop by Verizon or on a AWS instance or in an Equinix state facility.
[00:11:30] Stephen Goldberg: And so that means that me, the end user, as I query an API, that API is actually talking to something nearby and it significantly reduces the speed for the layman. What that means is if I am logging into an app, that app is gonna load faster, that console's gonna interact with me faster. And for the company, what it means is that it's gonna save them, have a better customer experience and save them a lot of.
[00:11:54] Matt Trifiro: that's super interesting. I have so many questions. You mentioned that there are tools today that allow you [00:12:00] to pretty easily distribute your content of CDNs. I think we're referring to, or distribute globally your API. What do they call those? The service that distributes your API?
[00:12:10] Stephen Goldberg: So like you could use containers and things like Kubernetes, or there's some companies that have productized that like section IO or edge gap or Ori, or, you know, there's a lot of companies in that space that do a good job of making that.
[00:12:22] Stephen Goldberg: But you're saying
[00:12:23] Matt Trifiro: that database is sort of left behind, at least in front of the people you talk to. And when I think of a database. I think of a, a big piece of iron sitting somewhere, and I realize this databases have gotten a lot more sophisticated and they can get split up and they can be replicated and duplicated and all kinds of things.
[00:12:40] Matt Trifiro: But you're absolutely right. I mean, I, when I've talked to developers, database gets to a certain size or a certain global distribution and you've gotta start doing like really horrible things, really unnatural acts to make that work. So the idea that you've built a system. Can be, it's almost like putting a sequel front end in [00:13:00] front of a CDN in a way.
[00:13:02] Stephen Goldberg: Yeah. And the two way is important because we're, you know, Redis does a great job of cashing your data all over the world, but that's not helpful if you want to read and write that data because that cash becomes stale. Databases have been left behind because people were really focused on things like consistency, meaning like cockroach, which you mentioned in Oracle, do a really good job of making sure that the data.
[00:13:25] Stephen Goldberg: Stays consistent everywhere. But then what that means is that it's hard to horizontally scale that, and you do have to do unnatural things and sometimes you just end up with a limitation. What we have built is a system where I can write to any node in a cluster. It is asset compliant. Meaning you guarantee that right on that one place.
[00:13:44] Stephen Goldberg: Let's say that's Denver, let's say that's Tokyo, but. As it gets around the rest of the world, it's eventually consistent. The reason that's important is because you get the guarantee up front, but then that replication happens significantly faster. And because you're not [00:14:00] worried a about the consistency as much.
[00:14:02] Stephen Goldberg: The other thing that we've done is because we're really a web application framework and not a database. The way that that replication happens is at the moment of right. It's natively integrated. And it, we use things like web sockets and which are more of a web framework and less of something you would think to use in a database.
[00:14:20] Stephen Goldberg: Everything is done over rest APIs and we're a stateless platform. So there's a lot of more like web three paradigms that we used and less sort of focused on the core data problems. I think part of that is. We are really just developers. We're not genius data people, and we didn't get so hung up on solving certain problems and those tools do a great job of those things.
[00:14:44] Stephen Goldberg: We don't do a great job of those things, but we do solve the problem we solve really well and really fast.
[00:14:49] Matt Trifiro: This podcast is about edge computing. And to increasing degree, this concept called the open grid, which is, it sort of pays some homage to grid computing, but that's [00:15:00] more of scientific computing, but the idea is okay, so this is very relevant to both of those.
[00:15:04] Matt Trifiro: I think it's relevant to edge because edge is a physics problem, as much as anything. I mean, it's a, there's a physics part of it, but there's also data locality data sovereignty part. I can put Harper DB nodes. As far out in the field, as I want as close to my end user or device or sensor or whatever as I want.
[00:15:23] Matt Trifiro: And I can put as many of them around the world as I want. And so what you're saying is if my, you know, I don't know my traffic light in Las Vegas stores, I don't know the, the photo of the car that drove through the hood light and my identical streetlight in Barcelona does the same thing. Help me understand.
[00:15:43] Matt Trifiro: The path of that data. So like Paul Harper, DB in both cases, different sides of the world, same database, right? Same company, same database. I wanna like eventually. Do statistics across all the traffic lights, what happens to that data? So I write it in Las Vegas, what happens to it? And I write it in Barcelona.
[00:15:56] Matt Trifiro: What happens to it?
[00:15:57] Stephen Goldberg: Yeah. And so that is a big [00:16:00] differentiator for us as well, is that a lot of people use what's called like a store and Ford method or like a hub and spoke model where ultimately you would store in Las Vegas and stored in Barcelona. And one thing I should clarify. You wouldn't wanna store an image in Harper DB, but the metadata about that image, like, um, that you could gather from it, the scan of the
[00:16:19] Matt Trifiro: license plate and how fast the guy was going.
[00:16:22] Stephen Goldberg: yeah. Normally the paradigm is that gets captured in that edge location. And then it's like an, a Mongo DB that capture that with realm. And then they'd send that back to some large cluster in the cloud. That's probably running on Aw. And so that's fine for certain use cases. There's a lot of T use cases where that works well.
[00:16:41] Stephen Goldberg: There's certain use cases where it's not great. The way that Harper DB is different is it's a mesh network. So there is no centralized, you know, mothership. You can set that up with our paradigm if you want to, but natively is a fully mesh. Peer tope model. Every node is aware of every other node in the cluster.
[00:16:58] Stephen Goldberg: And as I send it to Las [00:17:00] Vegas, the Las Vegas node sends it to Barcelona directly over a persistent web socket connection. If they lose state. There is a catch up routine that a node, when it comes back along line says to all the other nodes, Hey, what did I miss fill me in? And it sends them to all those nodes in, in real time.
[00:17:17] Stephen Goldberg: The way that that benefits is that instead of waiting, you know, to go up and then back down, all of that happens. Real time to all of them and they have that awareness of each other. So that is why it's faster. You can configure where you want your data to live at schema and table level and decide. I want this set of nodes and, you know, to have these set of data and this set of nodes of this and this to be globally shared, but ultimately.
[00:17:42] Stephen Goldberg: Because we have that globally shared schema and everything is aware of what the schema itself looks like. Maybe not where the data resides. You have that flexibility to set up however you want. And that's where a lot of that performance comes from.
[00:17:53] Matt Trifiro: If I capture data in Barcelona, but I query the database in Las Vegas, [00:18:00] the Barcelona data hasn't gotten to Las Vegas yet.
[00:18:02] Matt Trifiro: How does that query work? So it queries into the mesh and then what happens?
[00:18:07] Stephen Goldberg: It depends on the query and it depends on what you're executing against. So. You query in Las Vegas for the data that's in Barcelona. If it has not yet got there and you are querying the Las Vegas node, it will return null and say, Hey, I don't have anything, but it takes less than about 50 milliseconds to replicate that anywhere in the world.
[00:18:28] Stephen Goldberg: And so that is where. If you need that guarantee that that data's going to be there for FinTech. For example, that's a really great use case where like, you need to make sure that all the nodes have the same data all the time. Cockroach is the best solution in the world for that. And you're not gonna find really anything better than that, but we are also 5,000 times faster than cockroach because we don't do it that
[00:18:50] Matt Trifiro: way.
[00:18:51] Matt Trifiro: Is that a real number? 5,000?
[00:18:52] Stephen Goldberg: Yeah. It, well, it's like 4,900 and change. Yeah. Uh, yeah. Just to clarify, it's not exactly 5,000. It's very close to [00:19:00] 5,000. I'm rounding up slightly cuz I'm the founder of a tech startup. But yeah. So that's kind of the paradigm there of how that works.
[00:19:07] Matt Trifiro: I was gonna clarify my question a little bit because, so I understand like query in Las Vegas, the data's not there yet.
[00:19:12] Matt Trifiro: But I heard you say something else earlier, which is your data's a mesh network. It's not stored in any one place. Does that mean that no one node typically has all the data that actually generally, if you've got a large database spread across the world, that your data is gonna be spread across many nodes and to get the whole database.
[00:19:31] Matt Trifiro: You'd have to go to multiple nodes, is that it
[00:19:33] Stephen Goldberg: depends. So you get to choose how you want that set up. So oftentimes people have like a, a cluster of three nodes in a location that will have all, all of the data. Normally, what we found is that because Harper DB is extremely performant on a node level, as well as at a cluster level.
[00:19:50] Stephen Goldberg: And this surprised me, I'll be honest is that people tend to do full replicas everywhere. Generally they replicate their tables everywhere, all over. [00:20:00] So that each node has the whole database. The reason that's possible is. You know, it's not possible for all types of queries. Like if you're gonna do a select star from a table that has a trillion rows of data in it, that's not gonna work.
[00:20:13] Stephen Goldberg: However, like if you were to do a range, hash lookup on that, it will work and it will be fast. So for the types of things that we're doing, it's possible because they're not these super complex snowflake, like analytical queries. They're more like, Hey, I need this data. I know what it looks like. I want to get it.
[00:20:32] Stephen Goldberg: And then I want to update it fast. That's more, the types of use cases we're focused.
[00:20:37] Matt Trifiro: you mentioned a hundred thousand users
[00:20:39] Stephen Goldberg: in the developer community who are using Harper DB. That's correct. Yeah. Is it
[00:20:43] Matt Trifiro: based on an open
[00:20:44] Stephen Goldberg: source project? We have a freemium premium model. So we have a free version of Harper DB that we've distribute free forever.
[00:20:50] Stephen Goldberg: You can spin that up on a Docker container. You can install it via NPM install, Harper DB. Those are people spinning it up all over the world. Aren't where those numbers come from. [00:21:00] How
[00:21:00] Matt Trifiro: long has Harper DB been?
[00:21:01] Stephen Goldberg: We've been around for five and a half years. And we were founded in March of 2017, but the team itself we've worked together through three companies.
[00:21:10] Stephen Goldberg: And for over 10
[00:21:11] Matt Trifiro: years, I'm guessing that to date, most of the use cases have been the back end to a web service. A hundred percent.
[00:21:19] Stephen Goldberg: Yeah. Do you see
[00:21:20] Matt Trifiro: uses of it where it's backing, you know, a sensor network or IOT, or, you know, supporting sort of autonomous things that aren't involving like a web browser.
[00:21:30] Stephen Goldberg: We've done a lot of actually IOT projects in telematics for vehicles. We've just actually launched a project with a company called Edison interactive, where we're helping to power golf carts and doing low latency applications with them in Verizon. And so, yes, there's quite a bit of like IOT related stuff.
[00:21:46] Stephen Goldberg: Honestly, that market has been a little bit slower to move than sort of the edge API market. And part of the reason that we have focused more on gaming and streaming media is because those markets are here and now, and like, I kind of break the edge into you [00:22:00] have IOT, and then you have the edge, how I would classify those.
[00:22:02] Stephen Goldberg: Like IOT is more where there is a thing involved that it itself has some form. Intelligence, but Harbor Tobs a great fit for that. And we've done a lot of work mining as well, like monitoring electrical relays and high frequency sensors. Part of the reason it's a good fit in that area is because a lot of databases are optimized for read or right.
[00:22:23] Stephen Goldberg: Whereas Hartford TV, you can definitely scale more on the read side, but you can still scale pretty massively on the right side. So you do about 20,000 rights, a second on a single node. And so in a lot of T use cases, They're very right intensive because the sensors put off huge amounts of data. And so we found we're a good fit there.
[00:22:40] Stephen Goldberg: We're also about 150 megabytes and can run on anything from a raspberry pie up to a massive bare metal machine. So there's a lot of flexibility in
[00:22:49] Matt Trifiro: that the feature set that you were mentioning, where you can pick individual schema or tables to be, you can pick the distribution pattern essentially. Do you see that technique or that [00:23:00] technology or some variation of it being used in the future to sort.
[00:23:03] Matt Trifiro: Tier your data. And what I mean by that is one of the challenges in edge, especially with sensors is, and you got some sensor that's generating, let's say a trillion data points every seven days, and you may wanna have all that data local for some period of time where you can act on it fast, where you don't have to pay to send it anywhere.
[00:23:20] Matt Trifiro: Right. But you might want some subset of that data stored, you know, globally or, or in other locations. How do you view those sorts of problems through the lens of Harper? .
[00:23:30] Stephen Goldberg: Yeah, we've actually been doing projects like that for years. And so it's super easy with Harper DB. It solves that problem out of the box with no customization.
[00:23:38] Stephen Goldberg: So it's pretty easy. Typically what we recommend to folks is you have a data table or a raw sensor table. You send everything from that sensor into that table, and you can write what's called custom functions in Harper DB. So these are your own no JS applications that can be APIs. That can be ML. That can be anything you want.
[00:23:57] Stephen Goldberg: And so you can put some logic in there that says. [00:24:00] I noticed in the data table, this thing happened. I'd like this, you know, from this time to this time, let's grab that set locally on this node and then write that to a different table called the event table. And that data table may reside. You can set that up.
[00:24:15] Stephen Goldberg: So it only resides locally. I in the field on the device and the event table can then be replicated up to wherever you want it to go. We also have a feature called time to live, which was built exactly for this use case you're describing. And so you can set that. So it'll wipe out a table after a certain amount of time, because a lot of now empty,
[00:24:34] Matt Trifiro: empty your waste basket every 30 days, like my Google driver.
[00:24:38] Stephen Goldberg: Yeah, you can set it to one, you know, every 10 milliseconds or every 10 days or whatever you want that to be, because you're right. Like you might get 7 million rows of data in a day. And the challenge in industrial IOT is that I don't need that until I need it, but when I need it, I need every single one of those data points [00:25:00] so that I can have the best fidelity of my data possible.
[00:25:03] Stephen Goldberg: And so we built a lot of those features natively in Harper DB, and it, it works really. Harper
[00:25:08] Matt Trifiro: DB. At least the free product is something that I have to install. And if I want to distribute my database globally, I have to have the machines either. I have to lease them from somebody like Amazon or some bare metal company, or I have to pay to have them installed and co-located myself.
[00:25:24] Matt Trifiro: But I also understand you have a cloud offering, is
[00:25:27] Stephen Goldberg: that available? Now we have a cloud offering and there is a free version of the cloud offering. So you can spin, you can spin up a free tier on that as well. This is embarrassing, but I don't know if that's limited to a certain number of nodes. I know that I know that it's limited to the amount of Ram.
[00:25:42] Matt Trifiro: There's some limitation that if people with serious production workloads would need to move to pay. So tell me about the cloud product. How widely distributed are you globally? Like what's, what's your thought and growth plan for the cloud product?
[00:25:53] Stephen Goldberg: Yeah, we have done a bad job of investing in that service.
[00:25:56] Stephen Goldberg: We're gonna spend the next 18 to 24 months really growing that. So to. [00:26:00] It really only runs on Amazon in the United States, as well as Verizon me. And so you can deploy it on Verizon me, or you can deploy it on Amazon, the product itself, which is the same thing running on the cloud or running on a device, or you run it wherever it's in a really good place.
[00:26:18] Stephen Goldberg: We have some stuff we'd like to add, but you know, we feel like it's in a good place. And so we're gonna spend. Two years investing extremely heavily in the service because we are partnered with companies like Equinix, Google, Amazon. And so we are want to make it so that you can really deploy that anywhere on planet earth, on your own.
[00:26:35] Stephen Goldberg: From the studio, it go to studio dot Harper, db.ao, click a button and deploy it in thousands of nodes. We do also have partners. Can help you do that today. I, if you want to, but we'd like to build that directly I into the studio and, and that's currently what we're working on. And we, the cool thing about Hartford DBA, unlike a lot of databases is that it is container native.
[00:26:59] Stephen Goldberg: So it [00:27:00] just very easily runs in a container. And so getting it deployed globally, I is a pretty E easy pro.
[00:27:06] Matt Trifiro: I was just thinking about it visually about, you know, putting all these notes somewhere. And I thought back to what we were talking about just a few minutes ago, which is you could have a version of the database and the subset of the data running on the device.
[00:27:16] Matt Trifiro: I actually don't know of any, certainly not, not, it's not common of a single database that spans the infrastructure in the device. That's really interesting model because I would think, but if I'm the developer who has to ride code on the. and I have to figure out how to get that back up to the guys that are doing the analysis.
[00:27:34] Matt Trifiro: If I have Harper DB, I just have to write to the data. If soon someone's configured it correctly, that data will eventually get to where it needs to go. That's really interesting. As the device developer that'd be
[00:27:44] Stephen Goldberg: great. Yeah. We thought it would be something that people would like. And, uh, to be honest, a lot of people have written stuff.
[00:27:49] Stephen Goldberg: Uh, we haven't made a lot of money there, but we do feel like we've built the features to make IOT developers lives a lot easier. So you. You can literally install it on a raspberry pie, hook it up to a sensor. You can actually write your [00:28:00] application code in Harper DB two to consume directly from that sensor.
[00:28:03] Stephen Goldberg: We have a node red module for folks who just want to connect it via node red to the sensor. And then you just call local host to Harper DB. And so that saves you a bunch of money. It saves a lot of problems. And then. It's pretty easy to configure that note of Harper DB to another node or a larger server, and it'll replicate and handle all of that for you.
[00:28:22] Stephen Goldberg: So really what we're trying to do is we want you to do the stuff that's fun for you and that you are good at and like let us solve those other problems. And so we've built as much of that into the platform as possible. And I will say back to your point, we are in an unusual position where if you look.
[00:28:38] Stephen Goldberg: Extreme DB, for example, is a really lightweight small database that you can stall on anything. And it's great, but it's not something you would wanna run as a, you know, a massive cloud service and it's lighter weight than Hartford DB. And then you have, you know, a snowflake, but imagine trying to solve snowflake on a red pie, that's not gonna happen.
[00:28:56] Stephen Goldberg: And so we've done a really good job of threading this needle in between those two [00:29:00] things. We're not gonna be everything all in between. Our, our goal is kind of the orchestration layer of that, the movement of that and the development side of it. So
I
[00:29:10] Matt Trifiro: tend to think of the. Not just as of abstraction, but as geospatially right, same we do.
[00:29:15] Matt Trifiro: Cuz there really is physics involved. Like, like the data is bit, but it, it exists somewhere physically. right. That's and people don't really, really see that. And so what's, what's interesting is it's almost like you've recognized that. And you've built an
[00:29:30] Stephen Goldberg: abstraction layer. We have GeoJSON and Harper DB.
[00:29:33] Stephen Goldberg: So if you ever wanna play with it, you can do geospatial querying in Harper, DB also. Uh, you might find that fun
[00:29:39] Matt Trifiro: yeah, that's cool. You know, when I really roll forward into the future and I think about, okay, well, how are all these things gonna work? I imagine a significant portion of how the things will interact.
[00:29:50] Matt Trifiro: The internet let's say, or the cloud, uh, will be through a digital twin. And so there needs to be some way to bring in massive amounts of data, represent it in some [00:30:00] database and then, you know, represent it to a model and an AI model that can then reason against the digital twin. Do you see Harper DB operating that
[00:30:07] Stephen Goldberg: world?
[00:30:08] Stephen Goldberg: We did a bunch of digital twin projects early on. We actually started in IOT. So when we launched the company, we were exclusively focused on IOT, almost everything we did was in that space. And then we sort of moved. To the edge market, just because, you know, we have to make money. We're a company that said, like, I think Harbor DB's a phenomenal platform for digital twin.
[00:30:30] Stephen Goldberg: We actually thought that would be our strongest use case. I think, you know, because of some of the stuff you and I were already talking about to do digital twin properly. If you look at an airplane, if you look at a car, if you look at a manufacturing facility, The fidelity of data you need is insane because otherwise you don't have an accurate digitals win.
[00:30:47] Stephen Goldberg: You know, you need temperature readings, you need velocity, you need all these things. And so the volume of that, and then we've tried to build in things like geospatial, querying. Um, we've built enough. Things that you can [00:31:00] do all that you can do TensorFlow for machine learning directly in Harper DB to make that possible.
[00:31:06] Stephen Goldberg: Some of the challenges I see that we are not good at solving and that I think are gonna be super industry specific or super, you know, things specific like an airplane is a really complex thing and trying to implement. A digital twin takes a lot of specialized knowledge and a lot of rails that you need to be on.
[00:31:23] Stephen Goldberg: And so we can be a, a good building block for that, but not a, like a full solution for that. If that makes sense.
[00:31:30] Matt Trifiro: Yeah. The other thing it's interesting about Harvard DB, you know, you talk about running these workloads is no JS workloads. Like, so to some extent, as you're building out your cloud product, you're.
[00:31:37] Matt Trifiro: You're building out a cloud product that does compute also. So you may be in the awkward position eventually. I mean, it's a high quality problem, right? Where you're competing with your, with your partners, cuz you're offering. But as I realize, it's a very specialized workload, but that is a, that is kind of a really, again, it's a little bit of a mind shift.
[00:31:54] Matt Trifiro: Whereas when people talk about edge computing, one of the things that you talk about is like, well, don't move it. You know, a petabyte of data to [00:32:00] the workload move the 500 gigabytes of data, you know, of the workload to the data. And you've sort of operationalize that in a way, when I say run this workload, how do I tell it where to go?
[00:32:12] Stephen Goldberg: So if you go to studio Harper, DB IO, you can see your entire cluster of all of your nodes. You write your node JS code. There's two ways to do it. You can go into our studio, click a button and say, I want this on this set of nodes. I want this on this set of nodes, or you can just use GitHub actions and the sort of GitHub workflow and deploy using that directly into the nodes themselves.
[00:32:35] Stephen Goldberg: It's essentially just a fify server that runs natively inside of Harper DB that has access. To Harper DB's core operations without any hops. One of the problems we're trying to solve was simplicity and extension of Harper DB, and also giving developers. We want to collapse the stack. So they have less components they have to worry about.
[00:32:58] Stephen Goldberg: But the other piece of it [00:33:00] is if you think about like Lambda and serverless functions, those are great. And they're awesome. But even when they're running in a hyperscale cloud, like Google or Amazon, the Lambda may be running on one VM and dynamo. DP is running on a different one. And the problem with that is you have hops and.
[00:33:15] Stephen Goldberg: Creates latency. And so by putting them inside of Harvard DB, our core operations respond in nanoseconds, which means that you, you can do things really fast.
[00:33:26] Matt Trifiro: That is really interesting. Yeah. So let's, let's talk about hops and networking to some extent your customers and you to the degree that you grow, your cloud product are running a global network of notes, right?
[00:33:36] Matt Trifiro: So you mentioned the connection to the nodes is, is an open web socket. Are you doing anything or have you thought. the intervening routes, right? Because how it gets from Barcelona to Las Vegas, there's a, at least a dozen choice. Well, it's an infinite upper choices, but there's least a dozen good choices.
[00:33:52] Matt Trifiro: How do you think about that in the context of
[00:33:54] Stephen Goldberg: Harper DB? Yeah. So what, what I will say is we are currently using web sockets. Um, we are rolling [00:34:00] out a new technology in our 4.0 release that will not be on web sockets. That is on, uh, something new that I'm, I can't yet speak about. And I apologize. More advanced and it is true mesh instead of whereas websites sort of create peer to peer.
[00:34:15] Stephen Goldberg: This is a true mesh network. So part of that challenge is solved there in that the nodes in between Barcelona and Las Vegas have intelligence and can sort of assist in that with the true mesh. I will also say we're not a network provider, right? So we are partnered with Google. We're partnering with Amazon, we're partnering with Verizon, we're partnering with more and more telco providers, bloomin as well.
[00:34:38] Stephen Goldberg: Like we know what we can and can't do. And my background is not networking and is in development and I am not gonna solve the internet. I can do one very small piece to help, but we work closely with those folks. And like, you know, optimize has been phenomenal in helping us optimize things around that.
[00:34:55] Stephen Goldberg: And that is. Stuff we feel is done better by other folks [00:35:00] and that we rely on their expertise because like, otherwise we'd be boiling the ocean and way out off, outside our
[00:35:06] Matt Trifiro: skis. Let's talk about IOT. Right? So you mentioned, you seem to have a very informed view of IOT. You said the company started there.
[00:35:13] Matt Trifiro: Just walk me through what you thought. The original use case was gonna be where you think it went, and if you see it reappearing, it's
[00:35:21] Stephen Goldberg: going to reappear. I thought that it would be digital twin. I thought that that would be a big part of it. I thought that real time eventing would be a big part of we would do so we did a lot of projects.
[00:35:31] Stephen Goldberg: Whether that was, you know, we spent a lot of time in rock Springs, Wyoming going to oil and gas and mining sites there, helping them to do real time, power metering. You have these electrical relays, which are putting off, you know, 20,000, 40,000 data points a second, when something goes wrong, it costs them millions of dollars.
[00:35:49] Stephen Goldberg: So predictive maintenance, preventative maintenance. Things like that smart manufacturing. We did projects in sort of in transportation where Harper DB was running on fleets of [00:36:00] vehicles and sort of doing things like predicting whether a driver was falling asleep or intoxicated or going off route or things of that nature in real time.
[00:36:08] Stephen Goldberg: And so those were the use cases that we were kind of focused on. Ironically were starting to see them come back now with companies like Verizon, making the investment in the 5g. Me. And making it more of a possibility because the IOT hardware really never materialized. And so that was kind of what we saw as the problem was that the solutions were all there.
[00:36:29] Stephen Goldberg: Lots of people built to make those thing things in IOT, but we could never find a partner where we could say, Hey, do you have a product that we could put on the truck that makes it easy to get our product onto that truck to update our product, to. To the internet that is also waterproof and the battery's not gonna run out and that, like, it's not gonna catch on fire and it's not gonna melt.
[00:36:51] Stephen Goldberg: And the last thing in the world I want to do is be a hardware company. Um, the only thing I wanna do less than that is probably like go work in healthcare. But so [00:37:00] that was where we saw it sort of fall apart. But now as OT and the edge are converging, which is just starting to happen, I feel like in a real way, That is where I think it's gonna take off.
[00:37:12] Stephen Goldberg: And all those things will be real. And the IOT's original thought that, Hey, we're gonna put this stuff on these vehicles or we're gonna put them, you know, vineyard, or we're gonna put them down a mine shaft. I feel like that was a little over ambitious for where hardware is today. And I think the use cases where you can take advantage of the private 5g networks and take advantage of the near edge.
[00:37:34] Stephen Goldberg: I think those will come to life in, in the very near future. And I'm starting to see them come to life today, which is cool.
[00:37:40] Matt Trifiro: When you say the, the near edge, what do you
[00:37:41] Stephen Goldberg: mean by that? I mean, pops. So like telco pops, like little micro data centers. I
[00:37:46] Matt Trifiro: mean, so on the infrastructure side of the network, not on the device side,
[00:37:51] Stephen Goldberg: we have seen some projects we're working on two actually, where they're taking this, these same basically AWS outposts.
[00:37:59] Stephen Goldberg: You [00:38:00] know, a data center in a box, if you will and putting them on customer sites and creating sort of a private extension of the cloud that becomes that edge. And those do become real, but we have not yet. And maybe this is because we've been focused on other things and I'm just not smart about it anymore.
[00:38:17] Stephen Goldberg: I haven't yet seen the real IOT project where. You put compute devices on a bunch of trucks and that creates the cloud, uh, like a, you know, its own cloud,
[00:38:27] Matt Trifiro: an ad hoc vehicle network or something like that. Yeah. Those are farther off. One ways that I've come to think about it is you've got the infrastructure moving out farther to the edge and you have the stuff that's on the edge wanting to meet.
[00:38:37] Matt Trifiro: They wanna meet. Right? You want the compute compute data, all trying to meet. And, and it seems like the, the logical meeting place is literally at the edge of the access network. So at the cell tower, you know, the fiber intersection point or, or something like that, if you think about it within a, a local market within a hundred miles, 50 miles, let's say within 50 miles, the differences between a server running in a me environment [00:39:00] or in a micro data center.
[00:39:02] Matt Trifiro: it's the same as it being on premises from a performance standpoint. Agreed.
[00:39:05] Stephen Goldberg: Yeah. And that, and that's exactly what I think is those use cases where that work from. So you start
[00:39:09] Matt Trifiro: getting scale, you start getting, yeah. You start getting some of the economies of cloud delivered, right. To the edge where you're able to provide the same services that I'd have to go buy stuff and put it on my premises, which is the last thing I want.
[00:39:22] Matt Trifiro: So we're, we're you see us entering that?
[00:39:25] Stephen Goldberg: We're there. So you can go to studio dot Harper, DB IO. Today you can log in, you can click a button and deploy to 19 different locations than us market where on Verizon's 5g network. And then you can write, you can build your database in Harper DB, write the code and connect your thing to it.
[00:39:40] Stephen Goldberg: We just took a, as I mentioned, Edison interactive. Their golf carts are talking to 19 different Verizon pops throughout the United States, over the Verizon network, hitting our API. Can you walk
[00:39:51] Matt Trifiro: us through that use case in detail? That's a super interesting use case.
[00:39:55] Stephen Goldberg: Yeah, absolutely. So Edison interactive is, um, they have a, a many different [00:40:00] products, but their primary one is something called the shark experience.
[00:40:03] Stephen Goldberg: And it's a tablet that runs in a golf cart and you can order food on it. You can do, you know, your yardage, you. You know, listen to radio, you can do all these things, but there's many different APIs behind that. And they run on over 30,000 golf carts. The problem is that those golf carts are then making, you know, 20, 30, 40 API calls every time, the screen loads all the way back to one of the hyperscalers clouds and the database.
[00:40:29] Stephen Goldberg: And that introduced up to five second latency, that
[00:40:31] Matt Trifiro: five seconds of latency. What were the big bottlenecks? What were the, the things in that chain that contributed the most to that
[00:40:37] Stephen Goldberg: budget? Some of it was code. Um, some of it was, you know, 3g, 4g, uh, like network. You also keep in mind, it's a golf cart. So it's not like it's
[00:40:47] Matt Trifiro: in your house, it's on an LTE or something.
[00:40:49] Matt Trifiro: And it could be a, yeah, it could be behind a tree and exactly,
[00:40:52] Stephen Goldberg: uh, it could be anywhere. And then a lot of that was going, you know, if you're near. Where those APIs [00:41:00] originated. Let's say that's in the west coast of the United States. It's a lot faster. I'm in a golf cart
[00:41:05] Matt Trifiro: in Seattle. Yeah.
[00:41:06] Stephen Goldberg: Then you're great. Uh, if you're in a golf cart in the panhandle of Florida, that's not quite as ideal, you know, a lot more about the network than I do, but that's a lot of different hops to get there and there's a lot of distance.
[00:41:17] Stephen Goldberg: Yeah. And
[00:41:17] Matt Trifiro: if Verizon has a, a me down there that you can drop a Harper DB node on and push the data and some of the, the backend code to it, you're golden.
[00:41:26] Stephen Goldberg: Right? And so we moved not some of we moved the entire backend of the application to Harper DB custom functions, the entire database. So it runs in a fully distributed paradigm now.
[00:41:37] Stephen Goldberg: And so that when you were in Florida, you were hitting a me in Florida. And when you were in Dallas, you were hitting a me in Dallas. And when you were in Reno, you're hitting a me in Las. And so that is how that application works now. And we
[00:41:48] Matt Trifiro: reduce now are all these, all these little native services talking to each other through the database.
[00:41:55] Matt Trifiro: Yes, that's really interesting because you know, you look at the world [00:42:00] that let's just blame it on Kubernetes. It's not entirely Kubernetes laws. If you look at the, the world that Kubernetes has created with these multitiered service based applications and the need to route messages. Through a service mesh or some equivalent to, you know, and, and it's a pretty complicated tree.
[00:42:16] Matt Trifiro: It's really interesting. It's like, you know, maybe it was on purpose. Maybe it was on accident, maybe a bit of both, but you've built a distributed cloud in every sense of the word. and it's simpler. So where's the limitation. So, so if I was to choose between doing what you did, which is refactoring an entire set of backend applications that were written in what, like Ruby or pH fear or something, a bunch of backend applications into no JS Harper, DV functions versus replatforming on Kubernetes, how do you make that
[00:42:44] Stephen Goldberg: decision?
[00:42:44] Stephen Goldberg: One, if latency's not a problem for you, then, you know, it may not be as interesting, but as soon as you start to be this globally, you want to deploy in lots of places, then it does become sort of problematic. Harper [00:43:00] DB is not snowflake. So to be clear about that, like I mentioned, you're not gonna do these massive analytical complex, hugely expensive.
[00:43:07] Stephen Goldberg: Queries, but for your operational data, things like that, power, your API, that power, you know, your application interaction, that power integration. It's a great fit for that. The other thing which I did mention previously is if you need very strong consistency, We're not a great fit for that. So for example, like let's take FinTech and a banking transaction.
[00:43:28] Stephen Goldberg: if you, if you, the
[00:43:30] Matt Trifiro: money really should leave the account
[00:43:32] Stephen Goldberg: everywhere. Yeah. yeah, exactly. Uh, that, cuz that could be a problem. If you update, you know, Las Vegas and say you'll eventually get paid, we
[00:43:40] Matt Trifiro: think,
[00:43:41] Stephen Goldberg: yeah. Yeah. Or he takes the money out twice, which is uh, not a great thing. Even worse. Yeah. Um, so. That those are not great use cases, but we are trying to solve as many of them as we can.
[00:43:54] Stephen Goldberg: And when you mention Heroku, that's not a small thing to us. Like we, we loved Heroku. [00:44:00] We kind of were. Like think of it as what Heroku kind of wanted to be, but imagine that on the edge and then with some other things in there as well is kind of, that's what we are trying to build. Ultimately, I, in a lot of ways,
[00:44:14] Matt Trifiro: I mean, in a way there is a lot of similarity between.
[00:44:17] Matt Trifiro: The way that Ruby Ruby on rails ran on Heroku and the, the tie in to the Postgres database as a service that they built, it's a different approach to it, but it's very similar functionality. That's really, that's really interesting. Earlier, you mentioned that Harper DB is used a lot in gaming. What's the use case there?
[00:44:34] Matt Trifiro: How
[00:44:34] Stephen Goldberg: does that work? If you think about gaming, gaming has a lot of social APIs. It has a lot of distributed users, people all over the world. There's a lot of data that moves around and latency is a major problem there because if I log into my game console and I just want to do something as basic as set my presence from offline to online, or I want to see like what my purchases are or things like that.
[00:44:57] Stephen Goldberg: That you're talking hundreds of [00:45:00] millions of people you're talking about all over the world and massive data sets. And so very similar to the Edison golf cart case. We've, replatformed a lot of those APIs on Hartford DB distributed the origin of those. A lot of times we don't replace the origin. We just explode off of it.
[00:45:17] Stephen Goldberg: Cuz one of the things we can do is integrate into a dynamo DB, integrate into a fire. And then make those APIs run faster. So a very simple one. And one of the first ones you ever did was just you log in, in BU area you're offline to online. And that sets your presence online, everywhere in the world in a hundred milliseconds.
[00:45:35] Stephen Goldberg: Whereas in some cases that took up to a second to do that globally. And if you're playing with a friend, that's not a great experience that it takes him that long to, to be aware of things like that. It's like using slack. Yeah. Well, and so the reason we used. Socket cluster and socket. Well, we started with socket.
[00:45:54] Stephen Goldberg: IO is because slack used socket IO. So the crazy thing is under the hood. Yes, we're a [00:46:00] database, but we actually copied a lot of what slack did,
[00:46:02] Matt Trifiro: slack and Twitter and other things like that are, are sort of like the quintessential use case in some ways. Right. You just wanna write. The event, the tweet into some database and have it figure out all the other places that has to be to make sense.
[00:46:15] Matt Trifiro: That's really interesting. We talked about this, you know, IOT's day is eventually going to happen and digital twins eventually gonna happen. We're starting to see some of it now. Like, I mean, two years ago, you couldn't have deployed Harper DV on 15 edge nodes that Verizon runs in and coordinated fashion like that that's a new thing.
[00:46:31] Matt Trifiro: And pretty soon it's gonna be 15. Right. I mean, absolutely. I mean, there's an obvious progression there. When you look at the landscape of the world and all the moving parts, if you could sort of, you know, there's a bunch of dominoes that need to topple to get to where the sensors are connected to the internet, the data's being collected, it's being analyzed on, like, that's just part of our routine.
[00:46:50] Matt Trifiro: If you could push any one of those dominoes, what's the thing that you think would accelerate us into this future world. The fastest that if you could just nudge it,
[00:46:59] Stephen Goldberg: never in a [00:47:00] podcast. Give an answer that I thought should be edited. And I almost just did there. I mean, people is the thing. Honestly, I, I, I will say the answer cause I'm, I'm a, I'm a blunt person do it.
[00:47:09] Stephen Goldberg: It it's the people that work at the companies, right? Like it's not. So now that we Verizon and others and Vodafone and, you know, Google and Amazon have invested in this me stuff and Google's rolling out an awesome stuff there. And Amazon has great stuff there. Like it's there now. And like you said, for the vast majority of use cases, if you're within a hundred miles, Of a pop, like you kind of have what you need, the bureaucracy, the speed at which people move their ability to think about how things work that is now the one that I think is the biggest challenge I is that this stuff is now and it's here.
[00:47:46] Stephen Goldberg: And if you really wanted to do it, you could. But I, I think it's also just changing how people think about stuff. So. Willingness to adopt new technologies, willingness to adopt new architectural paradigms, not [00:48:00] trying to bring sort of the same cloud centralized sort of cloud model that you implemented to the edge, cuz it's not gonna work cuz it doesn't scale.
[00:48:10] Stephen Goldberg: You know, it's scaled to four geos. It's not gonna scale to 40,000 places like you just said. Um, I, I think that's it. I, I would also say. The Kubernetes piece of it, the services around that are good, but there is work to do. Um, one of the challenges we run a new constantly is the ability to have persistent storage in lots of places with Kubernetes, those two things being married is a problem for us.
[00:48:36] Stephen Goldberg: Aren't you persistent storage. We are, but if you deploy us in a container that are ephemeral by nature, and we have to do a lot of work to Mount that container to persistent storage, meaning. I mean the actual physical storage. Yeah. Right. Because what, normally, when you spin down a Kubernetes, you lose data container.
[00:48:52] Stephen Goldberg: Yeah. You lose the data, which is not ideal for her, for DB. And so the, those are some of the things I, I'm sorry, I give you way more than you asked for, but yeah. [00:49:00] So I,
[00:49:00] Matt Trifiro: so I heard a lot of really good answers, but it sounded like the one, the one that you, you were a little hesitant to say, because that involves like human beings, which is I, I call it cultural inertia.
[00:49:10] Stephen Goldberg: Yes. But it's changing. I think people are starting to see. Change now, but it's been a lot of blood, sweat, and tears for people on the
[00:49:18] Matt Trifiro: front. It is. But when you look at like the history of adoption, I mean like the iPhone is the exception, right. Everybody just got an iPhone in a week, but like, like ATM cars, it took 20 years to get half the us population getting money from the ATM because we, all those old people had to go to see the teller had to, had to stop with throwing funds, but yeah.
[00:49:36] Matt Trifiro: So, so things can take quite a bit of time, but I like you, I'm starting to see. Some of this, uh, materialize pretty, pretty meaningful. Hey Steven, thanks for joining us on this podcast. Really appreciate your insight. Very interesting arc to your product history. I do hope that the IOT business picks back up for you, cuz I think that's a really interesting angle for you.
[00:49:55] Matt Trifiro: If people wanna wanna find you and your company online, where should they go?
[00:49:59] Stephen Goldberg: Our website's a [00:50:00] great place to go. We do have a community slack. Almost our whole team is in there and highly available. So that's a great place to find us. And we're pretty active on social media, Twitter, Instagram, et cetera, so anywhere, but I, I really encourage people to jump in our slack cuz like our, my co-founder Kyle is CTO's in there answering questions and most of our senior engineers and it's a good place to find us.
[00:50:22] Stephen Goldberg: That's awesome. Thanks Ian. Awesome. Thank you so much.
[00:50:26] Narrator 2: That does it for this episode of over the edge. If you're enjoying the show, please leave a rating and a review and tell a friend over the edge is made possible through the generous sponsorship of our partners at Dell technologies. Simplify your edge so you can generate more value.
[00:50:40] Narrator 2: Learn more by visiting dell.com.