Data Challenges in the Cloud Era – Ajay Kulkarni, Timescale

This is Techstrong tv. Hey guys, thanks for the throw. We're here with AJ Kearney, and he's the c e o for timescale.

And we're talking about how databases are gonna be reinvented in the age of the cloud because well, things are changing. Aj, welcome the show. Hey, thanks for having me.

We have had time series databases, Postgres databases, all kinds of stuff. Being around the cloud, we can all go back to relational databases. But explain to us, if you would, as we enter this new era and there's data everywhere in the cloud, what exactly do we need from a database these days?

It is so fascinating. Uh, when we started this company, we thought we were just focused on, uh, IOT data and then, uh, time series data. And then over time we've, we've found people using us in more and more ways, uh, that we didn't expect.

And, uh, in particular, what we found is that there's just been like a whole new set of, uh, of sources of data that like didn't exist before. Uh, obviously sensors are an example of that. You think about like, you're, like a Tesla or a Lucid, which is one of our customers, or you think of, uh, you know, industrial manufacturing, the sensors people have deployed or, or transportation logistics, you know, sensor data, this one whole category.

But we have another of our customers who's in the, uh, who's one of the largest, uh, music labels in the US and well in the world. And we're like, why do we have an IOT customer over here then a music label over here? He's like, that seems weird.

And then realize it was the same trend that in this world there were also new sources of data coming online. And, and for the, for the music industry, it was, I think, a classic form of digital transformation that has like music consumption moved from offline to online, from, you know, tapes and CDs to, you know, SoundCloud, Spotify, you got all these new signals, uh, of what people were listening to in real time. Or, you know, maybe back in the day, like Billboard would compile like a weekly list of what's trending by calling radio stations.

Uh, but now, you know, companies can do this in real time by just reading feeds of SoundCloud and Spotify events, which is what, what this customer does. So I think what we've seen is that like, like on the, on one side you have all these whole new sources of, uh, data that are coming online. And the other side, uh, compute keeps getting cheaper, uh, storage keeps getting cheaper, uh, compute also gets more powerful.

And you put all these things together and you, you give 'em to developers, right? And the developers say, oh, I can, I can build a lot with this. Let me, let me, let me build, you know, and, and so what we're finding is that in this modern world, you know, people often talk about software in the world.

We like to talk about data. He is eating software that more and more applications are, are not like your credit applications of the past, like, you know, create, read, update, delete, but these modern probably more data intensive applications that are collecting these really like, uh, high volume new sources of data and making sense of them to provide and deliver really innovative and new customer experiences. So I, I think the role database in the cloud is that like, uh, also all, all companies are becoming software companies or they're becoming eating replaced by software companies.

Uh, you see this in the car industry, like I just mentioned, like, you know, you have Tesla and Lucid, but on the other hand, you might have, you have other folks like GM and Ford trying to become more software oriented. And anyone who's not doing that, it's gonna get replaced by one of those first two. So every company's becoming a software company.

Uh, software is powered, uh, a great software is powered by great databases. Uh, and in particular, uh, these applications that developers are building are not the simple applications of the past. They're really data intensive.

And so the, all of a database has only increased and it's increased not just like in terms of, uh, like the number of companies, but also like per company. The use of the database has also increased. So it's kind of this, kind of this geometric expansion of the role databases.

And I think that's one thing that makes us really excited here at Timescale. And we get to a single database architecture of some type because we have seen relational and columnar and time series and document databases. And it seems like everybody's kind of struggling with how to manage all of those different, um, flavors of a database.

So is that gonna somehow consolidate as we go along? Uh, it's a good question. I, I feel like that's the question we've all been asking for the past decade, decade plus.

Uh, I think, I think the reality is that as you move to the cloud, it doesn't really matter. Cuz I think in the cloud, like, especially what we're finding with new cloud architectures, so what used to happen in the past is people would say, oh, like I need a best in class database. So if time series data, I need timescale.

If I have, uh, you know, geospatial data, maybe I use post JS and Postgres, relational data Postgres. Uh, but what we're finding is that in the cloud, people just say, Hey, like, here's my data and like, like help me, like build a application on top of it. Um, so the model that we personally are moving towards in timescale is, hey, like, yeah, we can handle your time series data, we can handle geospatial data, we can handle your relational data.

Uh, we can handle workloads that, uh, require, uh, really heavy compute require workloads that require a little compute, a lot of storage. Um, but the key thing is like, we will handle it for you. So, uh, to answer your question, uh, I think the, like, we'll see, not a consolidation, but we'll see this question kind of go away, uh, as more and more workloads move to the cloud.

And, and if I get one more thing to that, Michael, I, I actually, one thing we strongly believe at Timescale is that Postgres is actually an incredible versatile foundation. Uh, and, you know, Postgres can handle relational time series, geospatial document json, you know, the whole thing. Vector databases now is a new hot thing.

Postgres can handle it. Uh, but you do have to make some tweaks behind the scenes. But yeah, so I, you know, I I I think the cloud, it kind of changes the equation in a really exciting way As we go along.

And we have so much data, you hear people talking about cardinality, which seems to be, you know, a very long word to describe the fact that there's just a lot of data that we're trying to correlate and trying to get some sense of the insights that we can, um, surface. But the question then becomes, um, are we reaching some point where we're having a bottleneck of some type because there is too much data. I mean, what is the ability of the current database architectures to scale up into the range of what seems to be coming petabytes of unstructured and structured data?

Yeah, I mean, cardinality is typically defined as how many unique things can you track, uh, you know, if you're, are you tracking like a thousand devices or are you tracking like a million users? Uh, uh, we've never had a problem with cardinality, uh, I think other database companies have, because the way they've architect, it's never been an issue for us. Uh, so I don't think cardinality is the bottleneck, but I, but, but there is a bottleneck.

And I think the bottleneck is really like, um, as these data volumes increase and you ask these complex questions of data, the database seems to be able to handle those types of queries. Um, so put another way, like, you know, I think compute is, you know, you know, I forget the latest graphs, but, but Moore's law is essentially slowing down, right? And, you know, computer's only getting so fast and, you know, GP is useful for some use cases, but not all.

But I think the trick then becomes, if you really want to unlock, you know, faster performance, and faster performance really means just a better customer experience. It means screens that load instantly as opposed to, you know, start having to wait. Um, I think the bottleneck becomes how can you, uh, architect the system so that you're thinking smarter?

And that's, that's actually one thing we're trying to do with the, as we rethink the database in the cloud. Well give us some insights as to what that means. I mean, is it a different schema?

Is it a different, uh, set of engines? What's going on? Yeah, there are a lot of things.

You know, I think, uh, one question we've been asking ourselves is when someone provisions a database, they often think about, Hey, how much storage do I need? Uh, how much compute do I need? Uh, now obviously that's, that's a friction point, that's a headache that creates some risk of like, oh, hey, did I provision enough?

And there's been this, the rise of serverless to kind of, you know, get around. But serverless has a its own set of problems where it's really expensive. Um, what we found and what we're starting to launch over the next few months is something that essentially gives you kind of true consumption.

Um, so the ability to actually, uh, not think about how much storage you're provisioning, how much compute you're provisioning, how much memory you're provisioning, um, and let the system kind of automatically handle it for you. And then you only get built for what you use. Um, the advantage of this is that as a developer, then you never have to think about like, Hey, did I provision enough?

Or, Hey, am I spending too much money? Because you only get billed for what you use. And, and this might seem, I don't know, this might seem like, uh, I mean, it's a really interesting change because on one hand it really li like eliminates a lot of the mental, he like, uh, gymnastics around trying to figure out the right provisioning so you have enough but you're not overpaying.

Um, but it also leads to like cost savings where you can save, you know, 20 50% of your, of your cloud spend the database that's architected the right way. Um, what I say, like it's architect in the cloud, I mean, there's some things we're doing in our cloud offering, uh, which is now called timescale. That would be a lot harder to do if you're trying to ship software.

Um, and I think, I think by kind of deciding, hey, there are gonna be some features that we ship via software and some features we just deliver via the cloud, I think it's unlocked kind of new exciting innovations like those. If that's the case, then do we need to essentially go into our developers and change their cultural way of thinking? Cuz they're so used to over provisioning infrastructure because of all the issues you just described.

Can we get to a new way of thinking about how we use infrastructure? It it is a new way of thinking, but I actually find it's incredibly liberating. Uh, databases are interesting where, yeah, some developers really nerd out on them like we do, but most developers don't want to think about their database.

They just want something that works where they can write data, read data, analyze data, um, that's, you know, easy, fast, scalable, cost effective, and worry free. And, and if the data, if the developer has to think about their database, usually means something is wrong. Like something in that, in the model is not working.

Uh, and we often talk about like, databases are kinda like your wifi. Like if you and I are worried about our wifi right now, then I mean something's wrong. Like we just want it to work.

The databases are different than the wifi. Wifi is ho muchas databases are highly heterogeneous. Um, and so one thing we've found is we've been kind of rolling this out with some of our, you know, beta users, is that yes, it's, it's kind of like, you don't, don't have to think about this anymore, but then it's like, oh wait, I don't have to think about this anymore.

Like that's, it's pretty liberating. And, and you know, and some people are like, I don't believe you. Like, how does this work?

And we kind of share, you know, some information with them and we let them test it. Um, but I actually don't think this is a, it's not a shift in thinking. It's, it's kind of like a, hey, this thing you thought you had to worry about, you don't have to worry about What happens to the role of the DBA in this world then.

Yeah, I mean I, I I, I think the role of DBA has always been, I I think it's been changing ever since. You know, the cloud first came about, I don't know, 20 years ago. Um, I think, I think there's still a role for people who, uh, you know, are maybe database engineers or data engineers or kind of managing the infrastructure.

But I think the role is shifting. Um, and I think it's, you know, not every company can afford a dba. Not every developer gets time, enough time with their dba.

And so we, the way we view this is we're kind of, you know, enabling the average developer to do more of their application. Um, but the reality is that, you know, if you're a dba, you know, I think you're, you're, you're the industry's evolving, your role's evolving and you know, I think it's incumbent on you to evolve with that. So as we come together and we all say, you know, data's the new oil, but it seems to me that part of the problem with that whole thing is we don't really have any refineries that can handle all the different types of that oil.

So, you know, are we looking at the wrong problem these days and we're not focused enough on how to process the data more efficiently in a way that we'll realize all these dreams that we have about digital transformation and fact driven decision making. Uh, yeah, I mean, I, I think, uh, I think there often there's too much emphasis on, uh, the specific technologies and the trade offs and, and I, I think the key is like, Hey, what is the customer experience you're trying to deliver? Uh, and in some cases, um, the challenge isn't at the infrastructure level, it's at the application level.

And there are often people who come to me, you know, and they say, Hey, should I use timescale? And I say, Hey, well are you happy with your current database? Because if you are, you know, don't, don't break what's not broken.

You know, don't fix what's not broken. Um, but I do think, like, I think, I think as an industry like it, it's, it's often hard to really have empathy for the developer cuz we just love the work we're doing around databases and data infrastructure. And the thing I really have to emphasize with our team is like, our average customer doesn't get here.

You know, they want this to just work and the way we nerd out on these things, they will not nerd out, but it's our job to nerd out on these things so they don't have to worry about it. So I I I, I do think it's shifting. I, I think with the cloud, I think it allows us to take on more of the, you know, job to be done, so to speak, you know, of the, for the developer.

Um, yeah, but I, I think it is shifting, but I think it's shifting in a healthy way. I, I think at the end of the day, like it's all about the impact we're having and you know, and hey, like what are we powering? We're powering a, a customer experience that works or is this a science experiment?

Cause it's a science experiment, who cares? You know, From an ROI perspective, are we now approaching a point where, because I'm not over provisioning, that my cloud bills may be, um, more acceptable because I'm processing more data per se and I'm getting more value out of that investment? Uh, yeah, I mean I think, I think the graph kind of looks like this, you know, like, it's like, like, uh, uh, we think the, the, or maybe like this, you know, we think the cloud has its ability to really unlock, uh, a lot of efficiencies that I think we're gonna deliver on the next few months that I think most people are not delivering on.

But I think over time, like more companies gonna spend more and more of their budget on their, on software, which will mean on their database. So, uh, I think we can do better than we are today, but I think over time you're gonna spend more money, but you're gonna spend more money to make more money, um, to build a better business. Um, so yeah, so I, I, I think, I think to answer your question, you can have a much better ROI than we have today.

Um, and I think what that'll mean is that, you know, as software reads, the world data is software that it'll enable companies to get more with their existing budgets and be happy with the growth of those budgets as they serve their customers. Do you think that also has implications for sustainability because I am processing data more efficiently and that has carbon emission implications? Yeah, no, that's, that's a great question.

Uh, to be honest, uh, I personally have not fully thought through the, the energy footprint cuz the, the cloud isn't really a cloud, it's like the physical data centers, right? And, and, and I think that point is painfully obvious when we ask Amazon to, you know, provide X number of new VMs and they can't, you know, in a given region, you're like, oh, it's, it's actually not elastic. You know, these are actual physical things.

Um, you know, I don't know. I mean, I, I suspect, and this is totally an amateur layperson's, uh, perspective, I suspect that there are some really smart people thinking about, Hey, how do we make these data centers more green, um, by, you know, better leveraging renewable energy, you know, having them located closer to, uh, renew renewable energy sources like wind and, and, uh, you know, hydro power. Uh, but I'll be honest, this is a little bit above my pay grade and, you know, is not what I spend my time thinking about.

But I'm glad people are thinking about this. All right, folks. Well, you know, no matter how much things change, they tend to stay the same on one thing.

It all starts with the data, and then you work your way up from there. Aj, thanks for being on the show, Michael. Pleasure.

Thank you. All right. Back to you guys in the studio.