Data Gravity Why Cloud Databases Will Prevail

Eric Kavanagh:

And we're going to talk about something called data gravity and cloud databases and what that all means. So we have several guests lined up for you today. We're going to hear from Venkat Venkataramani of Rockset. We'll hear from John Kreisa, old buddy of mine through several different companies. Now, the CMO over at Couchbase and Lior Gavish of a company called Monte Carlo.

Eric Kavanagh:

But first, let's talk about data gravity and the whole thing going there. It's really quite interesting. Data gravity of course, information assets don't have physical weight. There's a hard drive that they sit on that has some weight obviously, but data gravity really refers to the fact that it's hard to move data away from where it is. It's takes time. It takes effort. It takes money. We spent decades moving data all around with ETL processes, extract, transform, load, building data warehouses, pulling from source systems into the warehouse so you can analyze can understand what was going on. We spent a long time doing all of that.

Eric Kavanagh:

Well, now, with the cloud being a new center of gravity, what does that mean for data gravity? What does that mean for on-prem data centers? And we have lots of other really interesting trend lines to track here, one of which is federation, right? Distributed systems. We have distributed databases where you have lots of different nodes. They each have a copy of the data. This can help with latency. It can help with concurrency, lots of different things.

Eric Kavanagh:

So over the year in terms of using data, leveraging data, well, you've got a couple different major categories of use case. You have transactional stuff, getting things done, selling products, et cetera, et cetera. But then you also have analytical databases. And now, that's actually coming back together again, which is interesting that you would see a convergence or reconvergence of that.

Eric Kavanagh:

But the bottom line is that data is still the foundation of every application that you use, every business needs to use data. You're going to need a cloud database sooner or later, most likely depending upon your business model. On-prem is not going to go away anytime soon, but I don't think there's any doubt that the cloud is now the new center of gravity in the world of data. And with that, let me go ahead and bring up Venkat Venkataramani from Rockset. Venkat, tell us a bit about yourself and Rockset and how you're only in the cloud? Tell us about that.

Venkat Venkataramani:

Thanks, Eric. Thanks for having me. Thanks for the opportunity. My name is Venkat, I'm the founder and CEO of Rockset. Rockset is a real-time analytics platform. We help companies achieve fast analytics on real-time data with the simplicity of the cloud. We talk about data gravity a lot. I think the real trend that is happening right now is DBMS used to stand for database management systems. I think they're becoming database management services.

Venkat Venkataramani:

I think there is a tremendous advantage in eliminating so much of the operational overhead capacity planning on all these delays that used to get in the way of companies leveraging analytics to make decisions better and use data to grow their revenues, reduced churn, eliminate risk, what have you.

Venkat Venkataramani:

And now, data going to the cloud I think one big consequence of that is data is going real-time so Monday morning quarterback playing is not enough. In the middle of the third quarter, I want to know how to win this game before the game is over. And that is what real-time analytics is all about and that is what we do.

Eric Kavanagh:

Yeah, that's very interesting. I love your comment there, database management system systems. Now, it's database management services. And that's really one of the beautiful things the cloud offers, right? Is offer it as a service. We're seeing analytics as a service, all kinds of different things as a service.

Eric Kavanagh:

And it's great because to your point now, the business doesn't have to focus on a lot of the blocking and tackling that used to be necessary to do that. You have to hire DBAs, they have to know what they're doing, you have to monitor all that. And these days, a lot of that stuff goes out the window, which to me, gets the business ever closer to just doing what it needs to do to improve its operations, to improve its top line, bottom line, et cetera. Is that about right?

Venkat Venkataramani:

That's absolutely right. I think even before public cloud infrastructure came up and now databases and data management systems are all going to the cloud, you saw the advent of SaaS apps, from Salesforce to what have you. Another thing that has really happened as the data gravity goes to the cloud or the center of gravity of data is what I call it, goes to the cloud is all of these things systems become interoperable.

Venkat Venkataramani:

Because APIs can work with APIs and you don't have to have extremely complex data plumbing and what have you. The ETL stack has been upended by ELT infrastructure. But now, a lot of these things can also interoperate with each other because you can take the APIs for granted, they're more backwards compatible and what have you rather than software that changes once every two years.

Eric Kavanagh:

Right. That's right. I think a lot about this stuff because I'm in the industry, but I think about the old days when Microsoft was the operating system in so many use cases. And there were so many versions of the operating system and that was such a pain for developers because now every time a new OS comes out, "Okay. Let's just make sure whatever they changed, make sure it works." All the testing, et cetera.And then what happens? IBM throws a billion dollars at Linux and now it's the defacto standard for operating systems.

Eric Kavanagh:

And what I think is so cool about open source, but also about this whole industry is that we're climbing up the mountain where you're getting more and more of the foundation that is pre-built for you and is robust, and you know it works. You don't have to worry about it. Just as an aside, I remember the whole Hadoop movement at one point you had like what? Seven distros or eight distros of Hadoop. And then it was down at a six, five, four, three, two. It's like, "Okay, guys, did we need all that?"

Eric Kavanagh:

I think that there is this really interesting evolution and maturation of the entire data fabric industry, if you will. And the bottom line is businesses can benefit from that without all the hassle. Is that about right, Venkat?

Venkat Venkataramani:

Absolutely. I used to work at one of these large database companies back in the day. if you really think about even what really got the open source movement going amongst a lot of other reasons, one reason I would say is developers love speed. They don't want to be waiting around when they want to build an application. I can't wait for my CTO to approve what database am I supposed to be using to build this app. I'm going to download this open source and off I go, right?

Venkat Venkataramani:

Now, so then I think development time probably got cut down from maybe years to months. I can go on, get something working on my laptop and then eventually it's in a backend somewhere. But now with services, you can do it in a day. You can basically have databases that never fill up, that are fast, that can give you fast analytics and fresh data. They all interoperate with each other, and all of these things can come, you can take for granted scaling and growing up, you can take for granted and so you can move very, very fast. And so I think the same incentives that developers had to give up proprietary on-prem, install, configure software to going into open source. I think it's the same incentives that I think are also taking developers from open source software to cloud services.

Eric Kavanagh:

That's really good stuff because again, it just expedites the value to the business. And maybe I'll bring in John Kreisa from Couchbase to talk about this. Like me, John's been around for a while so you've seen some of these trends over the years. I mentioned before the show Dr. Michael Stonebraker. I had interviewed him back in 2005 when I didn't know who he was and Philip Russell who's now with Gartner was like, "Stonebraker, yeah. He's a big wig."

John Kreisa:

You have him.

Eric Kavanagh:

Right. He was talking about one size does not fit all. And from that date after that, we saw all these data warehouse appliances come out. And now, my goodness, there's a cornucopia of database technologies that can be used that are fit for purpose. So tell us a bit about you and what you're working on and what Couchbase does.

John Kreisa:

Sure. Great to be with you. Thanks for having me on the show. Great to be here with my other guests. So I'm John Kreisa, I'm the chief marketing officer for Couchbase. Couchbase is a publicly traded company, we IPO'd last year and we offer a leading modern database for enterprise applications.

John Kreisa:

Our customers include about 30% of the Fortune 100 who build applications that really run their business. They run mission critical applications. And underneath the hood is classified as a multimodal database. So we're talking about services earlier with Venkat, it has core database storage and query capabilities with SQL engine on top of adjacent database underneath.

John Kreisa:

But it also has search services and key value services all on a common platform. So it's very, very powerful for developers and architects to build modern applications. And we offer it as a fully hosted service. Couchbase is a 10 plus year old company, but we now have a database as a service called Couchbase Capella and so that is the direction and the thing that we see customers adopting and realize you said the purpose of this call or this podcast is all around that and momentum in that direction.

Eric Kavanagh:

Yeah, that's good stuff and I'm glad you brought in the other database types too. Key value stores, object or document databases, for example. And I do see this trend going forward, where you have one company providing all the different types of databases or at least multiple different options that you can use. Horses for courses is what my old cohost used to talk about on this very show.

Eric Kavanagh:

And maybe you could explain to some of our audience members out there, what are some of these databases good for like a key value store versus a relational database, versus a document database? Can you give just some quick assessment of which one to use and what circumstance?

John Kreisa:

Sure, absolutely. So I'll start with the first one you mentioned which is key value, that's an in memory architecture which is very good at very, very fast look up. So we used commonly and it's a common use case for Couchbase for user profile store. So if somebody needs to log into a service, it's used by companies that provide telco services so validating. Before a text message is sent, you have to validate the ID that you're sending it between. It has to be done very, very fast. Millisecond or sub millisecond response times. So that's a common use case for a key value database component of Couchbase.

John Kreisa:

If you look at SQL related capabilities, that's going to be highly structured for querying data, typically highly structured data. Although, at Couchbase, we do offer SQL query on top of a JSON database. And JSON database is document databases, as you said, have their own scalability and flexibility benefits to customers in terms of the way they store the data and the data models. So very, very flexible architecture for data that have a high level of variability like the information you're collecting on users that you need to store.

John Kreisa:

Is it a family of two, family of four? What are the profiles of the people in the family? If you're providing services out to them in a highly interactive application, you need that flexibility. So relational databases tend to be fairly structured, but give you a lot of high speed interactivity. JSON databases give you a lot of flexibility in data model, and Then you have other databases like key value, which give you high, high interactivity. And then if you can combine those all into one platform, then you really enable some really interesting use cases.

Eric Kavanagh:

Yeah, that's a really good point too. And it's always about performance, right? For whatever the use case is, you want it to be performance. And I think that's one of the main vantages of going with databases as a service up in the cloud. Like Venkat, I think was saying you can have databases that never fill up. You used have to worry about these things, right?

Eric Kavanagh:

Now, it scales out as it needs to. It can even scale back as it needs to with some of these newer technologies, and that's fantastic. Because not that long ago, you had to provision for your peaks and then suffer through the valleys where you're paying a lot of money that you don't even have to pay because you don't need the service. But now because it's scalable and that's really cool stuff. And again, it is the foundation of business. So if you get your foundation fluid, if it's moving very quickly, if it does what you need it to do, then you can focus on all these other things to make sure the customer experience is good, to make sure you're getting products on time, whatever the case may be. You don't want the database slowing you down, right, John?

John Kreisa:

Right. That's right. As a consumer of a database, as a user in an organization that's building applications, you really need to focus on your core competency, right? Provisioning database, worrying about the storage. As you said, in the scaling of the various services it's providing, that's not core competency, right? That's not where you're going to differentiate yourself.

John Kreisa:

Where you're going to differentiate yourself is the kind of applications, domain knowledge and expertise in the applications you're providing and in the customers you're serving and consumers, and that's where you really need to focus. And that is the beauty of what cloud databases have now provided and are enabling as companies are making that transition and we see it a lot, so-called digital transformation initiatives are rampant across the customer base. And they're enabled to do that more quickly because of cloud database capabilities.

Eric Kavanagh:

Yeah. That's such a good point. And digital transformation is what every organization, frankly, is doing one way or another. If they're surviving these days, they're doing something to change. But as I look at classic brick and mortar businesses and see where the opportunity is to change, you're right, it's in the cloud. And with a multimodal database, you can solve a lot of problems in one fell swoop, and then just let your DBAs focus on other things.

Eric Kavanagh:

Let people switch their roles in the organization. But the point is, if you have just two or three cars, you don't want to hire a full time mechanic. You just got two or three cars, right? Let them be handled by someone else. And that's what we have with modern cars, right? Most of it is under the hood technology. You don't have to know how to build out a radiator to be able to drive a car. So you want to focus on what you want to get done for your business and not worry about all that other minutia, quite frankly, which was very important. It still is, but now the machines do it in the cloud, right?

John Kreisa:

That's right. The fact of the matter is with most cars to stick with that analogy, it's so complex. You are taking it to somebody else to do that, but that's what you want, because you're focusing on getting to where you want to go not how do I make this thing go? So it's a good analogy

Eric Kavanagh:

Yeah. It's important stuff. We want to be able to advance and the other cool thing too, is that up in the cloud, so much functionality has been baked into the cloud. Who logged in when? Who touched what? All this governance and audit trail stuff is just baked into the cloud. So it's a lot easier to troubleshoot, it's easier to know what happened at what point in time. Whereas in the old days that would be a bit trickier to get to the bottom of. Real quick, what do you think, John?

John Kreisa:

Totally agree. The ability to have knowledge about the usage and what's going on and be able to track that and have all the collective expertise within the company who built it, to manage it for you provides a lot of that value back to those users for sure.

Eric Kavanagh:

Yeah. Well, it's the foundation of business folks. We're talking to John Kreisa of Couchbase, Venkat Venkataramani of Rockset and next up we'll be talking to Lior Gavish of a company called Monte Carlo. They do data observability, which is really part and parcel to being able to manage data environments. You have so many data sources these days that you can pull in, more people know about data warehousing, thanks to companies like Snowflake of course. Rockset calls themselves, I think he's got a special comment he'll make about that, about their comparison to Snowflake.

Eric Kavanagh:

But the point is that more people are figuring out how all this stuff works. And the bottom line is if you can get that data where it needs to be, and then let it be leveraged by all your different parts in your business, that's the key. You don't want silos. You want it ideally in one robust place, they can service all the different needs, all the different apps that you have. We'll pick that up after the break folks, don't touch that dial. You're listening to the longest running show in the world about data, it's called BM Radio.

Speaker 5:

We're clear.

Eric Kavanagh:

All right. One down, three to go. That was fun guys.

John Kreisa:

Yeah. So far so good.

Eric Kavanagh:

Lior, are you ready for your close up? Yeah, it was good.

Lior Gavish:

Good conversation.

Eric Kavanagh:

Yeah. It blows my mind, man thinking about how much things have changed in just 20 years. I knew it was going to be big 20 years ago when I got into this industry and it's blown my mind just how much bigger it is than I thought it would be. It's exciting, fun.

John Kreisa:

You referenced that I do, it says [crosstalk 00:18:51].

Speaker 5:

Oh, yeah so.

Eric Kavanagh:

The people who are in the Zoom, you get to hear our chit-chat in the middle. You'll get to get the commercials. It's one of the perks the behind stage baby.

Voiceover:

It's the longest running radio show in the world focused on data. DM Radio, be a guest some time? Send an email to info@dmradio.biz. That's info@dmradio.biz. NS1 has a singular focus to deliver the best application traffic intelligence and automation technologies in the market so companies can build the best future.

Voiceover:

The most innovative companies in the world rely on NS1's flexible, automated and scalable solutions to build an intelligent, automated and resilient foundation to manage delivery and optimize application traffic. NS1 puts innovation within reach by eliminating boundaries between applications, users, infrastructure and data. It's the only

Eric Kavanagh:

Yeah, we use this guy for most of the commercials that we do. And after we started using him for a two years, all of a sudden I started hearing him on cable. Doing the cable commercials and I'm like, "There he is. That's Chris Davies." I'm like, "Hey kids, that's Chris Davies. Provo USA." He's our guy. He's amazing

Eric Kavanagh:

When we first launched the show back in '08, we were trying to find someone to record commercials. And I reached out to a few ad agencies and they were like, "Ah, you want a 60 second, it'll cost you a three grand. Like, "Are you sure? $3,0000 for a 60 second ad. I think there're a better way." We found this guy and he's like, "60 bucks." There you go.

John Kreisa:

Nice.

Eric Kavanagh:

Just cranks him out, man. Like I've given him probably, I don't know, 300 commercials over the last 15 years. I'll email him almost any time of the day. He gets back to me within 30 minutes, sends me the invoice. Another 30 minutes later, I got myself a commercial. I'm like, "Yes."

Venkat Venkataramani:

That's what all of our customers say when they use Rockset.

Eric Kavanagh:

Is it? All right.

Venkat Venkataramani:

It's so fast to get started? I can get my job done in a day, which I thought would take six months.

Eric Kavanagh:

Okay. So go to Lior first when we come back here, about eight or nine minutes and then I'll just go right into the rounds and we'll probably throw it over you, Venkat for.

Lior Gavish:

Venkat and John, just curious people that use Rockset or Couchbase, what percentage of the workloads are more analytics oriented versus production application oriented?

John Kreisa:

Yep. We're an operational database, so it's principally operational applications with analytics on the data that's inside about the usage. But it's principally operational.

Eric Kavanagh:

All right, folks, back here on DM Radio, talking all about cloud databases. And we're going to talk about an adjacent technology now, something called observability. And I've been watching this trend for about 10 years or so. I think it was really in the heart of the whole head duke movement that observability became a real thing. And it's referring to the ability to see which data is coming in or which data isn't coming in to be more precise.

Eric Kavanagh:

So a lot of times, especially for the analytics world, you have all sorts of data sources coming in. Sometimes in real-time, you can stream data these days using technologies like Kafka, for example or Flink, any number of other technologies. They've been around a long time, but the open source ones have really taken over and are now the [inaudible 00:23:17] technologies to use.

Eric Kavanagh:

But the point is you have all these different pipelines of data coming in and it could be a very difficult thing to manage. And so observability has to do with being able to track and see where data is coming in? Is it the right kind of data? Has it gone down for some reason? And that gives some indicators to when your pipelines are broken or when things are changing. So it's a real fun, it's very interesting space. And Lior Gavish is with a company called Monte Carlo. So Lior, tell us a bit about yourself and Monte Carlo and what you're doing in the cloud database space.

Lior Gavish:

Hi Eric, thanks for having me on the show. So my name is Lior and I'm one of co-founders of a company called Monte Carlo and we help our customers with data observability. I'll try to explain what it means if people haven't heard of it before, but it's the idea that a lot of companies today, essentially every company today is building a lot on top of its data infrastructure. Whether it's analytics for internal consumption, whether it's production use cases where data is the centerpiece, whether it's machine learning models that make predictions in real-time for the business.

Lior Gavish:

Whatever it is, there's an increasing reliance on data and on the data infrastructure that is ingesting transforming and creating those products. And just like any complex system, it will break for a variety of different reasons, whether it's operational issues or changes in the underlying data or human error, many times.

Lior Gavish:

And data observability is an idea we're taking from the world of DevOps of basically how to introduce reliability into software systems. And we're applying that to the world of data engineering and data pipelines and data products, right? And the idea is that by instrumenting the data infrastructure and the data that's flowing through it, you can actually measure the health of data and the health of the data products that are being used downstream and do things like create proactive learning when things go wrong, where things break so that the people building the system can fix it.

Lior Gavish:

And you can actually help more rapidly fix things and you can even help prevent issues from happening in the first place. And all this should sound very familiar if you're a software engineer, but I think this is a new concept that's coming into data engineering and into the data practice, and we're very excited to help with that. So yeah, that's data observability.

Eric Kavanagh:

Yeah. It's important stuff to keep the trains running on time, right? Because people get, they rely on these services, they rely on these data services. And so if they're not working, bad things happen and you're an early indicator of what's going on there and you also I'm guessing help optimize information architectures and information systems, is that about right?

Lior Gavish:

Right. Right. So we're talking today about cloud databases and about all the advantages in terms of speed and agility that they bring into play. One of the bigger or one of the welcome side effects of that is just the complexity of these systems is growing exponentially, right?

Lior Gavish:

If people can build things faster, there are more products, there are more underlying data assets, there are more things that can break. And so using cloud databases has really accelerated the complexity of these systems and the need to know the things the are working as expected. So just like you said, knowing that things are coming in on time, knowing that the data is correct, knowing that all the data that you need is there is critical in terms of making sure these systems realize their value and earn the trust of whoever is using those systems of the business, essentially.

Eric Kavanagh:

Yeah. The whole world is data driven now, as we're seeing, and really the whole world is analytics driven. I have a good buddy who says, "We're now not just in the information economy, it's the insights economy." And the meaning there folks is that you need to be deriving insights from your data, whether it's real-time, whether it batch, offline, weekly reports, things of that nature.

Eric Kavanagh:

But the closer you can get to now, the better, depending on the industry, of course. And that real-time nature, to your point, Lior, I think is even more complex, right? So you really have to build those pipelines, know when they're working, know when they're not working and be able to fix them quickly when they are broken, right?

Lior Gavish:

Right. Yeah, absolutely. It wasn't too long ago where the way data was used as someone would pull reports from the data warehouse, put them in a binder and present them to the business. And there was a gate there that would make sure that the data is right, that all the data is there, that everything is right. And now when we're talking about real-time use cases like what Venkat mentioned, you no longer have that luxury of a human doing that work.

Lior Gavish:

You need to think about it in a more scalable way and make sure that if you as a data architect, as a data engineer are serving data in close to real-time, you need to build better confidence in the reliability and health of that data. And to do that in real-time, you actually need a whole different methodology and a whole different set of tools to make that possible, right?

Lior Gavish:

And this is why we believe it's important to bring some of the learnings from building software services, from building web applications, from building other types of software systems where that real=time element exists in bringing that to the world of data and implementing it over the data stack and the data development workflow, if you will.

Eric Kavanagh:

Yeah. No, that's so interesting. And I'm glad you're bringing up some of these historical constructs and you think about DevOps and how DevOps really revolutionized development for the organization. And I think in a very positive way too, because we had, you still have this, but in the old days, there was a pretty clear divide between business and IT, right? And the business people didn't know how the IT worked and the IT people knew the business field, didn't know how the IT worked and then created a bit of a discordance, I suppose.

Eric Kavanagh:

And then DevOps comes along where you got developers working right with operational frontline people to get stuff done, but it was almost like they got it done too fast. And it's like, "Well, how did you do that?" "I don't know. It just works, leave me alone." And now we have all this data ops stuff, and it's all flushing out and helping the automation of that entire end-to-end process. And that's when you can really start screaming. That's what Amazon has done so well, it's why they're this gargantuan company now, because they nailed down those processes, made them Bulletproof and automated everything they possibly could, right, Lior?

Lior Gavish:

Yeah, absolutely. Like you said, the data space has made a lot of progress and especially the cloud environment enables better workflows where data is processed and transformed in a way that's more consistent with how you do software development, with how you do DevOps, right? So there's tools that will allow you to manage your data assets like code, with source control, and with testing and with proper review and deployment processes and a lot more automation.

Lior Gavish:

And these things are really helping both accelerate the pace of innovation, right? And also introduce more repeatability, more consistency into it. And through there more reliability and more trust, more velocity, and even more security, right? You touched on it. Some of the things that are really hard to do in the past in terms of enforcing proper governance, proper access control process, proper auditing are now baked into the systems, into the development workflow and are enabling that in the data space and that's something that I'm very excited about.

Lior Gavish:

And data observability is also part of that, right? We're essentially automating the part of monitoring your system of proactively managing its reliability and proactively fixing issues as they happen.

Eric Kavanagh:

Yeah. That's such good stuff. And maybe I'll jump into the round table here. I bring Venkat back into the story from Rockset. Lior referred to agility and speed and the need to be nimble, and I think that is such a critical success factor today when markets can change very, very quickly. We saw with COVID, consumer behavior changed dramatically overnight.

Eric Kavanagh:

The companies that were prepared for that, that had their systems instrumented, that had their data, they had their analytic pipelines built. They were able to pivot much faster than those that didn't. And I think the ones that didn't, they probably aren't even around anymore. So that agility is absolutely crucial today, and that's what you guys focused on too and how you built out Rockset. Do you want to talk about that and maybe compare yourself to a Snowflake, for example?

Venkat Venkataramani:

Absolutely. I think we talked about cloud databases. You look at Snowflake data breaks, you look at these giant data lakes, lakehouse, data warehouses, what have you. But so the world is going from batch to real-time. What Snowflake and data breaks did to batch analytics and business intelligence, we're doing it to real-time analytics and operational intelligence.

Venkat Venkataramani:

It's not about the quarter is over, what worked in this quarter, what didn't work in this quarter, how do we go and win the next quarter better? This is the last week of the quarter and revenue operations, dashboards are coming in, what can we do to go and win the game before the game is over, the middle of the third quarter.

Venkat Venkataramani:

An NFL game, you see people running around with Microsoft Surface tablets and stuff. So the world really demands real-time analytics. But if you look at the history of databases, for the most part, they fall into either transaction processing, which are fast and they give you speed. But it's very, very hard to scale them. And then you need to do very rich, complex, 17-way joint SQL analytics on it. And then you get these warehouses that are scalable, but they are slow and don't expect them to be covering your applications or these kinds of operational intelligence. So this is really why Rockset exists, which is Rockset is a real-time analytics platform and that basically it gives you fast analytics on real-time data.

Venkat Venkataramani:

And another thing that you mentioned, early on we talked about is all the connectivity that is much more available and easily accessible. And we're the only real-time analytics platform in the cloud that comes in with built connectors, for example. It doesn't matter which system you're using to manage your data, you can bring that data in real-time to Rockset for fast analytics on that, and building real-time dashboards, real-time applications, APIs and what have you.

Venkat Venkataramani:

So I think all of this essentially accelerates this movement. If you go and interview a hundred people, do you want slow analytics on stale data or fast analytics on real-time data? I don't know how many people will ask for the former. Then you should ask, "Why is that not the new default?" And I will tell you why, because fast analytics on real-time data has always been not possible without a lot of cost and complexity. And so this is why we're re-imagining what a real-time analytics platform should look like in the cloud and really bring it and make real-time the new default.

Eric Kavanagh:

Yeah. That's a good point. And we've got about two minutes before our next break, but I'm going to throw a proposition out to you, Venkat and then our other guests can comment on it in the next segment. But when people get data, when a business person gets data, especially insights, they're driven by data, they use it and then they don't want it to go away, right? Once you get that score in someone's hand, so the salesperson can see, "Hey, this is actually a low value customer. This is very high customer." Anytime you get that kind of insight that is germane to your decisions day to day, you're going to want to keep using it, right? Very quick, go ahead.

Venkat Venkataramani:

Absolutely. One of our customers coined this phrase, and this is straight out of their mouth, which is, "If my internal analytics tools are slower than Instagram, no one in my and uses it." Right? And so they came to us, they had built a web app on a cloud data warehouse that shall not be named. These tools were taking 30 seconds, 60 seconds to render. They moved to Rockset and it runs in 18 millisecond for every equity response time. So speed matter. Not only that you need to give data into the hands of your people, it has to be fast and it has to be fresh.

Eric Kavanagh:

Yeah, that's a good point. And I remember talking to an analyst on the show a while ago, we were talking about Facebook in the early days and why Facebook won. And one of his arguments was that it was sole performance. It was so snappy. You click on something and happens. We're really spoiled by the speed of cloud based systems these days, no one wants to wait.

Eric Kavanagh:

If you click on some link and it's like five seconds have gone by. Man, you are like 10,000 dots away from where you were five seconds ago. That's how quickly it transpires in your brain these days so you have to get that speed. You must get that perform to get the value out of your technologies. But folks, don't touch that dial, we'll pick it up after the break. You're listening to DM Radio.

Eric Kavanagh:

Back here on DM Radio talking all things cloud databases with our guests. We've been talking to John Kreisa of Couchbase, Venkat Venkataramani of Rockset and Lior Gavish of a company called Monte Carlo. And John, I'll throw it over to you from Couchbase. We've kind of talked about this already in the show, but the developer focus for enterprise technology companies, databases, other kinds of applications.

Eric Kavanagh:

It's a real big deal these days because you want to capture the imagination of the developers. You want to give them what they want and give them a seamless experience because they're the ones at the cold face, as you say, hacking away building stuff. So you want to cater to them, get them on your side. And I think that's attack that you folks have used pretty well too. Is that right?

John Kriesa:

Yeah. That's right. Developers are, again king makers in terms of choosing technology and it has to be easy to get going, but then ultimately have the power that they need to build the applications and have some of the future proof. I'll call it capabilities that you start simple, but build out that that application that they're trying to build to meet the needs there.

John Kriesa:

It has to be flexible, fast and have familiar capabilities that make it so they can going in terms of language support, tooling so that they can swap. And we were talking about before DevOps, related workflows and how that's getting closer and closer to the business side of the house and just really that has that barrier has gone away. And so the developers have to work directly with and be smart about the business and they have to have the flexibility to do what they need to do.

Eric Kavanagh:

Yeah. Right. And I'd be curious to know from your perspective, working at Couchbase, is it primarily net new projects where folks will look to a Couchbase or are you actually seeing migrations off of older systems like a Db2 or an Oracle database? What do you see in terms of that shakeout?

John Kriesa:

Yeah. Now, for us, it's both. We see relational technologies that are not scaling and don't have the flexibility to upgrade the application. So we see quite a bit of what we call legacy offload. Customers who are looking for a new infrastructure with the familiarity of a relational database, but with the flexibility of a document oriented JSON database. And so we see quite a few moving to our model because of that and because of what we provide and that multimodal capability, which we talked about before which is, "Hey, I'm going to build the application as a key value use case, but ultimately I need to add search and add SQL query and other capabilities into that application." And then of course, if you're just looking to build a new, highly capable, end user, highly interactive customer application, then we see plenty of that kind of use case as well.

Eric Kavanagh:

Yeah. And I'll bring Venkat into this too. It's funny how much things really have changed over the past, let's call it 10 years, if not 20 years about how you would go about building applications. And we used to go buy all this off the shelf software, you deploy it on either on your server or on your data center or whatever. And now there's just a whole different ballgame in terms of how you can build out these apps that are very finely focused on what your business does well, very fine grained.

Eric Kavanagh:

And you're getting insights into the hands of new people who had not traditionally done that kind of stuff, right? Business intelligence and analytics was in the hand of a handful of people at the top of the company only, and it was pretty much historical. Whereas now you're actually feeding insights the people who are running the business and you want to talk about shortening the time to value that's where the value occurs immediately every second of every day, right, Venkat ?

Venkat Venkataramani:

Absolutely. Look, I think we live in the information era. We just gave these devices that can hold a lot more information than what a single human brain can hold and can help you make decisions. We just gave it a boring name called a database. I think that's what happened.

Eric Kavanagh:

That's funny.

Venkat Venkataramani:

And the digital transformation started in the '80s when the first time instead of paper and pen records and ledger rooms, we went and picked relational databases and say, "You know what? We're going to store off records there." In fact, Oracle calls their rows records for that reason. So that don't worry, we got all of your records, don't worry about it, right. And then what happened was the next advent of warehouses and whatnot, where I have all this information now, can I make better data driven decisions?

Venkat Venkataramani:

And this is the advent of warehouses and lakes and whatnot. But again, Monday morning quarterback, video cameras are invented or maybe I can watch these game videos on Monday morning and plan my workout schedule so that I go win the next game coming this Sunday. But now, every helmet, every player is wearing a Fitbit, every helmet has information. You can analyze this information in real-time.

Venkat Venkataramani:

Now, people are demanding, well, in the middle of this third quarter, the game is not over yet, we're behind by eight points. What plays are working, which players are injured in my team, which players are injured or are not having a good day on their opponent's team, what players are working? I want to change the result of the game before it's over. And that is really what's happening now.

Venkat Venkataramani:

So real-time analytics is the future. Now, people are using it only in these cases where they cannot live without it. It's almost like a requirement, but this is going to become the new default as the world transitions more and more over away from batch towards real-time.

Eric Kavanagh:

Yeah. That makes a lot of sense. And Lior, maybe I'll bring you back into the conversation. Multi-cloud is a reality now, I think a lot of companies have figured that out. We talked about hybrid cloud, I kind of joke sometimes. We went from using hybrid as the word to multi after two weeks because we realize, yeah, that's not just going to be hybrid. It's going to be multi. Do you have a strong play there and being able to serve as that early warning system for multi-cloud environments? What do you think about that?

Lior Gavish:

Yeah, absolutely. We're seeing customers adopting a multi-cloud strategy. They basically want to use the best tool for the job for different use cases, which is amazing and it enables a lot of new use cases and it allows people to do it in a cost efficient manner, in a redundant manner so they're more resilient. But it does also introduce complexity and it brings up the need to essentially create a layer that helps people understand what's running where, how it's performing and whether the applications that they're running on top of it are actually trustworthy, right?

Lior Gavish:

It's all about trust in the end of day. Can you trust the data that's there? Can you let the end user click and refresh? Like we mentioned earlier and know that they're getting the right answer and this is where in general adopting a DevOps mindset is critical. That's why we believe it's critical to bringing those principles that we learn in web applications into the world of data applications. And that's where also observability plays a big part and helps. And so from our perspective, multi-cloud brings a lot of power, but with it comes a lot of responsibility to manage that and to understand how it's working at any point in time.

Eric Kavanagh:

Yeah. And I think John Kreisa I'll throw it over to you. I think it's wonderful that we have such competition now in the cloud, you've got Microsoft who saved us from the AWS monopoly, which I always find amusing. But Google has gotten very serious with GCP and you've got some other lots and lots of niche players. And then we were talking in the beginning of the show, maybe we'll do this in the podcast bonus segment about the edge and what the edge means.

Eric Kavanagh:

Because now from an architectural perspective, there are so many ways that you can solve for these challenges. How much compute do you do at the edge versus the data you send over the transom? How much habits in the cloud, what do you keep on your data center? The beautiful thing is there are lots of options for your architects, let's say your cloud architects to really sit down and hammer out some bulletproof processes and make your business special. What do you think, John?

John Kriesa:

Yeah, I'd agree. There's enormous complexity when you start to solve problems on a global scale, there's distributed databases, there's multi-cloud and then as you said, it's out to the edge. And so all of that creates complexity yet you still have the need for speed. We've been talking about it. The interactivity doesn't go away and pushing things out to the edge, brings that data closer to the user, brings that compute closer to the user but it adds that complexity for the developers and for those that are creating the application.

John Kriesa:

So would agree 100% that's something that is the natural progression, but it adds that complexity. And certainly for tooling that like Lior and others bring, it helps with developers to bring and understand what's happening and understand that there's a problem.

Eric Kavanagh:

Yeah. I heard a fun stat the other day from an analyst on this show who said that every two years, the number of developers doubles as people come out of college and start getting into the ball game. And I think about that, that's very exciting. But then you're like, "Wow, that means half the developers out there right now are brand spanking new." Half of them don't have this history and sometimes that's good and sometimes it's not so good, right?

Eric Kavanagh:

There are some lessons that we can let go of from the data center days, I think in terms of some of those networking challenges, for example. But it's an interesting challenge to know that half the people out there are brand spanking new. Maybe they'll have new ideas, maybe they'll have new ways of doing things. One of my developer buddies is a great quote that will close our lives show on here which is that, "Busy is the enemy of creative." Right. I love that.

Eric Kavanagh:

There are always more interesting ways that you can solve problems by just taking a step back and realizing what are you outsource? We can use a service to do this. What do you build yourself versus what do you just rent from the cloud? Lots of fun decisions to be made. Podcast bonus segment is up next. Folks, you've been listening the DM Radio.

Eric Kavanagh:

Okay. Folks, time for podcast bonus segment here on DM Radio. We're talking to Lior Gavish of Monte Carlo, John Kreisa of Couchbase and Venkat Venkataramani Rockset. And maybe Lior, I'll throw this out to you first. The edge is such a fascinating frontier because you can do so much at the edge. You can capture information, you can do some processing to understand what's happening. You can send signals back alerts back, think about IoT and just the stunning complexity, but usability of IOT. We're seeing all sorts of fun stuff happen there. What are your thoughts on the conundrum of figuring out what to do in the edge and how to manage all that architecturally?

Lior Gavish:

Yeah. I think the key with edge computing is just to be really crisp about what you're trying to do. There's some cases where it makes a lot of sense to go to edge, right? If you're looking for low latency, if you're looking for an extreme level of privacy, right, in security. Maybe some other cases as well. And those cases makes perfect sense to go ahead and use edge technologies and invest in that.

Lior Gavish:

Having said that, also important to acknowledge the complexity that comes with it. It's an ASIN technology. It's not something that's well understood today and it's fully developed. And so you want to be careful to go there where it makes sense and to use more traditional methods when that makes sense. So a very powerful tool that comes a great responsibility, right?

Eric Kavanagh:

Yeah. Well, Occam's razor, right? If you're in the systems design business, you want to build systems that get the job done. You do want to future proof it, you want to open the door to be able to bring in new technologies. That's one good thing about open source and just one good thing about some of these frameworks. But You definitely never, ever, ever want to make something more complicated than it needs to be.

Eric Kavanagh:

Maybe I'll throw this over to John Kreisa of Couchbase. I'm surprised that so far, I don't see a dominant player in edge. I don't see one vendor just taking over that space and ruling the roost. What do you think about that? How big of a big deal is it for Couchbase as you guys look forward?

John Kriesa:

I think there's that probably the wide range of edge and IoT use cases that would indicate why there's no one single dominant vendor, just to talk to that point in particular. For us, it's a key core component of the platform. And a lot of the use cases we see from our customers do include that. So customer service, service oriented, related use cases, handheld devices, things of the edge that also need both online and offline.

John Kriesa:

Because one of the problems edge presents is you can lose connectivity. So you need something that can handle data due processing and not go down when it loses an internet connection, be it a WiFi network or cellular, or what have you. So that just goes to that complexity that we're talking about, about the devices and having that built into the platform is key. And that's something that we've done certainly, and as a key piece of many of the use cases our customers have.

Eric Kavanagh:

Yeah. Right. And Venkat of Rockset I'll bring you into this. Yeah. I keep thinking about operational analytics and the need to get those insights to these people because let's face it. When you think about a data warehouse, when the data warehouse was built, there were only so many data sets that you cared about. It was mostly ERP data. It was transactional data. It was not the preponderance of data that we see these days, which is all over the map. Click stream data, consumer behavior, all sorts of point of sale data can be leveraged and the world doesn't fit into a nice relational box, right?

Eric Kavanagh:

There are times, especially in operations when you're using all sorts of different data sets that traditional ERP systems just don't have or didn't provide. And so to me, that's one of the benefits of taking an approach like you folks have done at Rockset is that you're able to cater to these edge cases, let's call them. Not necessarily on the edge, but edge case, meaning little niches in businesses that are where the real meat happens, but they've never been able to quickly be able to pull data in and run an analytics. Is that about right, Venkat ?

Venkat Venkataramani:

Absolutely. In edge computing, I think the building solutions that work end-to-end and in edge computing is still in the early days. I think we know some of the pieces that are needed, but not all the pieces that are needed. It reminds me of a conversation I had with Nokia Bell Labs researchers that are doing cutting edge research on edge computing weeks ago, just like two or three weeks ago. And they were saying, "All right. We need to do video automatic inference and figure out, let's say an application will alert a company, whether or not somebody's wearing a COVID mask or not."

Venkat Venkataramani:

And then we started talking about, "Okay, how will you install it in a hospital?" Well, you really need to take a different action. Whether the person that is not wearing a mask is a visitor, somebody dropping a FedEx box, inpatient, outpatient or a medical practitioner, a medical professional. Well, the video stream that can translate and detect person wearing masks versus person not wearing masks, that computing can happen on the edge, but you still need all of this data.

Venkat Venkataramani:

The inference data, not the raw video data to come in and you need to be able to join that with your records for patient records, employee records, which is coming from a different database and whatnot. And only then you can actually build a solution where edge computing is very big component to it. But the solution, because who gets alerted when they detect that somebody's not wearing a mask and how they work on that is everything is different.

Venkat Venkataramani:

So that is what we do. We built in, this is why at Rockset we think a lot about this and have built in connectors. So you can bring in real-time data from wherever it is. It could be in the edge, it could be on-prem, it could be anywhere. But you can stream that in real-time and then get real-time applications built on top of that.

Eric Kavanagh:

I love it. It's just a whole new world of very, very focused, specific applications that you're not going to find at the best buy. You got to build this stuff. Well, folks, this does conclude our show. Big thanks to all of our guests today. Look these guys up online, we'll talk to you next time. You're listening to DM Radio.

Venkat Venkataramani:

Thanks, Eric.