Fixing Legacy Data: How Quesma Reinvents Database Migrations and Insights

[00:00:00] Welcome back, but today I'm excited to unveil the Tech Talks Network. It's a collection of my shows designed to explore every corner of enterprise technology and the impact that it is having on our life, work and even world. So, the Tech Talks Network consists of eight podcasts. So please check out techtalksnetwork.com. You'll find links to everything that you need right there. But back to this show, today I've got a great guest lined up for you.

[00:00:28] So, if you or your organization are trying to modernize messy stacks while still needing speed and performance, I think this conversation will offer something useful to you. So I invite you to join us as we explore the broader market need, how Quesma is positioning itself in the crowded marketplace, and what's coming next from this venture-backed startup that isn't afraid to get its hands dirty when it comes to fixing those real-world data problems.

[00:00:57] But enough from me. Time to get today's guest on. So, thank you for joining me on the podcast today. For everyone listening, could you tell them a little about who you are and what you do? Hello, I'm Jacek Migdalm. I'm founder and CEO of Quesma. It's an early-stage startup doing database gateway. We try to make changes in your data stack easier as well as bring better insights.

[00:01:21] And thank you so much for sitting down with me today. And one of the things I love doing on here is finding out more about the origin stories. Can you walk me through that founding story of Quesma? And also, what led you to identifying this market need in the database gateway space? There's got to be a story there, right? Yeah, the story starts really early. I was blessed because my dad is an engineer, so I was introduced to computers super early when the whole programming environment was on floppy drive.

[00:01:51] And back then, one of the cool technologies was databases. So, SQL databases, especially Oracle's ones, were like a miracle. Now you can organize your data. You can do a lot of stuff. And my dad did his living introducing digitalizing parent ecosystem, telco, energy utilities, all sorts of companies. They moved from pen and paper to pretty much, okay, everything is digital.

[00:02:16] And over the years, I made my own careers. I did like competing programming, but I was always close to database world. There's so, so much advancement in this technology, while, like, you know, my dad is soon to be retired, still work on the same database. And those nice vendors start to be really aggressive, really bad, while the majority of the world is very, very far away with, like, you know, the best modern data stack.

[00:02:43] And I'm curious, what were some of the most eye-opening moments during your research phase, which I think was in 2023? Because I believe that was, like, the pivotal moment that helped shape the company's direction, right? Like, I was looking for a lot of, like, cool ideas, but they end up not really working. But this idea was, like, something that was inside me that I rediscovered. I was, like, surprised how easy it is to get to find some of those pains. Like, if you ask people, okay, what's your data migration story?

[00:03:12] Many people have, like, some nightmares. They are, like, you know, horror stories. I was, like, surprised by horror stories because many of them are not reported. And those people are, I'm not going to touch it out. Or they have, like, horror story about some licensing issue. So, and I was, like, surprised how easy it was to say, okay, I'm going to fix your ugly legacy database to get some executive attention. Because all of them got scarves. They, like, worry to touch it.

[00:03:40] So, the problems seem to be really, really there. Like, it was not clear for them that there's, like, solution yet. But I like when people go for many means just rambling how bad it is, how terrible, how they don't like it, how somebody is trying to squeeze them and they want to get away from that. And it was, like, very professional people saying things like that.

[00:04:04] And as a result of that, I was reading that you positioned yourself as a smarter gateway for database interaction. So, how exactly does your proxy work? And what advantages does it offer over some of the traditional migration tools or observability platforms that are out there at the moment? Yeah. Traditionally, most people do very, like, migration, like, very service-based. So, though there is, like, you know, some tools to help you with that, it's a lot of manual work. Right?

[00:04:30] And that's, like, how traditionally that people do that system and they do maybe three more, maybe three-year project, usually three years, sometimes got even for more years. And they change everything from one place to the other and they move it. And I also saw that this approach seems to be very costly and very risky. You uncover risk at the very last moment and those projects often get derailed because there is a lot of undocumented use case. So, what I propose is to be this proxy in the middle layer.

[00:04:59] So, you could, like, transparently add it at the beginning of your migration and test it, like, with both new and the previous systems along the way and gradually change it transparently to your application. So, like, rewrite in fly, like, piece by piece instead of, like, a big bang migration. And when I first met you at the IT press tour in London, you mentioned significant use cases in industries far and wide from telecom to travel and media.

[00:05:29] Are you able to share some of those for people listening? Just to help them understand how you address performance, cost challenges, etc. So, the first use case we are targeting is Elasticsearch for, I would say, observability. Because what also we found is conservative markets. So, the use case is like, okay, I have, like, a very big system. I really deeply care about performance, problems, bugs. It's, like, a mission critical. But, on the other hand, this scale is, like, huge.

[00:05:58] And I could not afford some shiny tool like a data doc because it's costing, like, a hundred times too more something that I wish to pay. Right? And to give you, like, examples, like, one of the customers we have was, like, a telco. And in telecommunications space, it's very common to network troubleshooting. So, you know, things break. There could be a frequency. There could be hundreds of things why you cannot take phone from one location.

[00:06:24] It's, and there is, like, an army of people every day working tirelessly to make your phone work. Right? And you may not appreciate. You may just disappreciate why I don't have reception. But there are a lot of people employed to make that work. And they have a lot of challenges. But the first challenge is to even understand what's going on. And to understand that, they collect all of the data records whenever you go to any location from your phone.

[00:06:49] So then when there is, like, some issue, they use this, like, usually egg-baked system to search for the logs, figure out what was the root cause. Maybe there was, like, some connectivity issue. Maybe it just happened in some users during roaming. And once they figure out, they repair it. So our solution was, like, to address this use case. Because, as you know, these days, Telecom spends sometimes, like, even millions on a single system like that. And it's, like, you know, it's not keeping up. Like, you know, people are not paying more for their cell phone. There are no more people.

[00:07:19] So the company tries to squeeze them. But for those people, it's, like, one of the most crucial tools. Right? So our solution is to modernize that. Make it 10 times faster, cheaper. But also, like, keep all of the ecosystem. Because it's usually very hard to change those systems. And we want to keep them because they like the UI. They like their debugging tools. So they want to take advantage of those. So we are kind of, like, replacing the engine of a car. But keeping rest of your car the same. Fantastic.

[00:07:49] A few of the other industries, travel as well. Anything else you could share around there? So in travel industries, like, also, like, so travel is, like, kind of, like, similar use case. It's travel industry. It's also, like, a big Germany travel booking company. So this is, like, a business when there's a lot of integration. It's not that you are dealing with one hotel. You are dealing maybe with hundreds of different hotel systems, flight systems, things like that. And the margins are not that big. You think you pay a lot for your plane. But actually, brokers get, like, a tiny fraction of that.

[00:08:19] But what happens is also, like, you know, things break. One integration, something added. So they also use it to monitor what is breaking. Why it's breaking. It's usually not the whole system. It's usually, okay, maybe people from Greece buying tickets from that airline could not buy them for some reason. And then somebody backed that. And again, the system was very inefficient, like, too slow. It was taking a lot of time to travel through those outages. While on the other hand, they were also paying a lot of money.

[00:08:47] So we had to move them from Elasticsearch to Clickhouse. And we already proven, like, several times gain performance gains across many dimensions. And from the outside, looking in, you've built this gateway between Elasticsearch, EQL and SQL-based databases. Claiming that EQL users can actually use it faster, cheaper, can access it faster and cheaper. Store data sources. Can you expand on that? Because that's the big ROI here. Yeah. Yeah.

[00:09:15] So the story is, like, usually the biggest gains are not just because of the better technology, but also, like, if the technology was designed for the wrong use case. Right? And so if you want to tighten your screws and you don't have, like, a screwdriver, it's going to be much more painful. Right? The database Elasticsearch got stored that was designed for Google study searches. So kind of like, oh, give me top 10 results.

[00:09:38] And there's a lot of theory, science, how to rank documents, how to support that better. If you search for car, you should also, like, BMW, things like that. A lot of this theory. Which, you know, it's sometimes very important. And it turned out that people don't have that many products, so they could pay extra performance price for that. But in the log data, you don't care about that. You just care, is that your cell phone number or not?

[00:10:06] It's like a very much simpler queries on that. But the scale is, like, enormous. Right? So the traditional Elasticsearch are document-based, like, no schema, usually mostly row-based. They were written in Java. While the most efficient engines are kind of columnar. So they kind of store one column after the other. So it's kind of like, instead of the grocery and getting one beer and repeating it eight times, why not just buy a pack of them? Right? So that's like a vectorized database.

[00:10:35] So that's why we are seeing those gains for these types of use cases. And we're talking about, of course, your first product in market. But not only that, you've also got two more launching this one. So can you tell me a little bit more about these new products that are being demoed this year? Yeah. So the first one we drew a demo is also, like, there was, like, this Google release, like, a pipe extension to SQL. So we found also, like, when people are using observability, we are having, like, extensions

[00:11:02] that could, like, enrich your data, find, okay, I have IP addresses. Where are these users from? I'm seeing your email. From which company you are? How big is this company? So we are doing this abstraction on top of databases because we found people were using a lot of products, not sure, like, why they need it. And this abstraction, this SQL with pipe syntax that Google invent is getting more traction and could allow to add this extra functionalities inside your database.

[00:11:29] Another one which is going to be launched, like, next week, like, kind of, like, preview is Quezma charts. So we also found, like, many companies, and that's coming in our research, that, you know, those people, if you're working, sometimes they are not able to present data well enough. They have, like, some generic BI tools, but we found, like, a really nice way to generate insight in form of, like, visualizations, and we call this product, like, Quezma charts.

[00:11:56] And again, for everyone listening, what would you say that makes you stand out from all those other incumbents and service-heavy solutions in this $10 billion data migration market that we're talking about? What makes you different? Yeah, I believe the biggest in this market, we are taking very different, like, a pragmatic approach. We already assume in all of our products that we have, like, a huge amount of data. You have a lot of legacy. A lot of people were not designed ground. Okay, you start from scratch.

[00:12:25] And this is, like, a huge differentiator for us that we like messy environments, which sometimes make things trickier. But for us, we already work in air-gapped environment, even as, like, one-year-old company. So this is, like, something uniquely differentiator.

[00:12:40] Another way what we are doing, I'm not, like, calling my AI, but sometimes we use, like, AI like, 10 times cheaper. But at the same time, combine the best word that is, like, soft us and the other word that we have, like, everything written as, like, a rule, well-tested. So it's live.

[00:13:10] It's not going to have, like, any mishap. So I believe we figure out this balance really well. We are still, like, thinking whether to communicate on our website more about this insight because, on the one hand, it's, like, kind of, like, implementation detail. On the other hand, it let us do things that previously would be prohibitively expensive. And I was also reading before you came on the podcast that you're exploring a new SQL syntax with extensions and, of course, AI integration.

[00:13:38] We had to get AI there somewhere. So how do you see this evolving to support interoperability and richer query experiences for your users? So as we have Pypack, it's same, like, a proxy, same gateway. It's a little bit different use case. So it's adding more new functionality to this. Here, so, yeah, I'm seeing sometimes, like, people like this abstraction or they don't know exactly what schema I have, but they want to combine, like, a free...

[00:14:05] We have, like, clients in one system, like a CRM, in our own database, in the third. I would like to figure out what's the best way to combine those data. And they really like sometimes, like, this AI guidance that you could, like, almost, like, generate schema for them how this should work. That people like a lot. But I'm also, like, seeing that you have to be sometimes, like, careful to make sure where

[00:14:28] you change things from, okay, it's very productive, but also, like, not leaking your data or not, like, you know, making some mistakes when you decided on one approach. And I think I'm right in saying, for everybody listening here, wanting to find out more information about you, am I right in saying you've also built observability pipelines at Sumo Logic and optimized databases at Facebook, obviously now Meta and Dynatrace. Is that right?

[00:14:55] Yeah, so other, like, founding team members, but I was doing both Sumo Logic and Facebook. I changed database, and that's also, like, part of my life story, that the biggest changes in my life when I either saved or earned the most amount of money were if you change how you structure your data. Because for a lot of companies that are, like, a huge either cost item or huge profit item, if you could make better decisions, that's sometimes, like, a biggest lever.

[00:15:23] And I found that, like, having, like, a big production system is, like, a course because sometimes, like, very hard if you break something, it's, like, enormous disaster. But on the other hand, if you optimize something there, that's, like, enormous gain. And you're also a venture-backed startup, of course. Tell me a little bit of information about that, too. Yeah, we're venture-backed because we are in category when this is, like, kind of, like, a hard and expensive product to build.

[00:15:53] It's kind of like, okay, you want to be correct, you want to be accurate, you want to be performant. You know, there's, like, a 10 different checkbooks that you need to hit even on, like, a minimal version. So because of that, yeah, we need to hire, like, a lot of people if you work at the database for last 10 years. Right? And that's, like, you know, they earned something. So that's why we started with $2.5 million US dollars, both from VC from Hardcore. It's, like, a pan-European investor from, primarily from Denmark.

[00:16:22] They baked several companies, like, Neo4j, Tink, and also, like, Polish investors like Innovo, as well as five US-based angels. Wow, incredibly cool. And obviously, April is a big month for you. You've got two new demos. But if we look further ahead, what else is on your roadmap, especially as you expand across open source adoption, observability pipelines, AI-enhanced querying? I suspect it's going to get even busier for you, right?

[00:16:48] Yeah, currently, we are, like, showcasing, we are showcasing and dealmen, like, charts everybody can use. This pipe is, like, we are collaborating with very big companies. So this is, like, you know, it's kind of, like, if you'd want to do standard, it's, like, a consortium play. So there's a lot of, like, also, like, you know, aspects of finding this group of people interested in collaborating and implementing the standard. And we have, like, a server good conversation. So with this standard, I'm seeing it may take a while, like, I would say, to form this consortium.

[00:17:18] But I'm seeing, like, pretty much the biggest USA companies interested in doing that. So this is, like, I would say what we want to start effort. And this is ideally this year we would like to have something that would be, like, live in few systems among those companies. That's, like, what we are looking forward. We are going to do this. We are going to do this, like, a lot of things that we can do on the start with these charts. That would be easier to spot because anybody can use it. I would say late April, I would say a data console, I believe April 22nd, we are going to release it.

[00:17:47] And soon after that we are going to be releasing the world. and this is going to be too nice to that you could like generate charts in any style like you know any guidebook any company style with like a few lines of text or like a line of text you could like do a lot of like data analytics wow so much to do there and not only that obviously you're talking to me today where you're based in poland we met in london a few weeks ago and not

[00:18:13] only that i suspect you're going to quite a few tech conferences around the world yeah where can people listen we do have a lot of us listeners listening as well where can people find you on the road any anything you can talk about there in oakland like this april like that's like a san francisco bay area i will be also like in kubecon in usa later that year and probably a few others but i'm still waiting for some responses so that's like a two us-based conference in europe probably

[00:18:41] it'll be also like monitorama yeah awesome and for anybody listening wanting to find out more information dig a little bit deeper on how you might be able to help them where would you like to point everyone listening to stay up to speak with some you know if you have some ugly legacy system that you know people are screaming and you're thinking about change or you're seeing just use case and either it may be cost or either you try to leverage like reach out even if it's not today

[00:19:10] like you know we work with few people early on so like reach out that's like a very good use case to talk about the other use case is like this immediate ready product that you could try yourself even without talking to us it's like elastic search to click out well i'll have links to everything to make it nice and easy for people to find you who are your typical customers you mentioned legacy tech there is that a big driver of the kind of people that are coming to you and asking for your help

[00:19:35] with yeah so currently i would say the best conversation we have are companies which got some history so they're like five ten years old there is something like enterprise but sometimes even like a company that you think they are modern they still been around and the primary currently we are focusing on analytical use cases so queries that you know okay they query a lot of data people to figure out what's the optimal price i would say the beauty and bits of the data that's almost every industry

[00:20:04] have some data but currently we have like a better like in people running on bigger scale like telcos like a cdn companies that they're running on very big scale and they i would say they have incentives to optimize so you know they are either spending a lot or they are not getting enough from the system they could not troubleshoot their issue they could not find the insight fast enough and for anybody

[00:20:31] listening there anyone with uh who is battling that challenge that incentive to optimize i think it's something that every business must be going through but let me know what you thought about this i'll add links for everything but just thank you for coming on the show and shining a light on this great work you're doing thanks again thanks so much name i think my guest approach today is somewhat of a timely reminder that innovation doesn't always mean starting from scratch sometimes the real

[00:20:59] breakthrough lies in meeting companies where they are no matter how tangled their tech stack might be or how old it is they simply need a more pragmatic way forward whether that be easing painful database migrations or adding ai powered insights through smart charts and sql extensions etc and most of all the big takeaway for me is that they're proving that modernizing legacy systems a victim of so much technical

[00:21:29] debt that has built over the years can be achieved without breaking them so big question is your organization still wrestling with clunky data workflows expensive migrations or tools that seem to promise everything but end up delivering nothing but frustration are you building on top of a house of cards or just trying to keep your current systems running without losing ground love to hear your

[00:21:55] thoughts on everything we covered today and what you thought about the conversation that we had and what's been your biggest data pain point lately and what are you doing to solve it it's not just about me and today's guest i'm bringing you into the conversation so drop me a line join the conversation email me tech blog writer outlook.com linkedin x instagram just at neil c hughes i've given you a lot to think

[00:22:21] about then i'm going to give your brains a rest and we'll return tomorrow with another guest on another problem and hopefully a solution speak with you all then