P99 CONF – Dor Laor, ScyllaDB

This is Techron tv. Hello everyone. Welcome back here to techron tv.

I am, I'm really happy actually, to have do, Loor Do is the co-founder and c e o of c e o of Sila db. And, and Dora is joining us today from Tel Aviv in Israel. We were talking off camera.

They, they still are having sirens there. You know, this is just a few days after the, uh, most recent, uh, violence has erupted there. And I appreciate taking time out of what's going on in your life to come on Techstrong TV and talk with us today.

Thank you. And our thoughts and prayers and support, of course, go to you and Israel and all of our friends, you know, in the, uh, Israeli tech industry at least anyway. Right.

Um, Dordt, let, let's start a little bit with a little background. I, I think most of our audience is probably familiar with cdb cdb, but some people may not be. So why don't we start there?

Let's give 'em a little baseline to go on. Mm-hmm. Sure.

Thanks a lot. So CDB is highly scalable database. It's a distributed database, uh, building c plus plus, and it's supposed to scale as much as you can go.

So, uh, we're built for the largest, uh, workloads in the industry, like Web two oh companies, uh, social type of, uh, workloads that you need to store billions, uh, and billions of, uh, of rows up to petabytes. And it needs to be always on and highly available today. Customers cannot sell with, uh, downtime at any hour.

It's supposed to be 24 by seven, uh, and they need to store huge amount of data and be ready to, to overcome any faults in every server fault or any even, uh, data center failure. We're, we're supposed to have a replicas in other data centers, and that's what syllab is all about. Uh, and we can do that at the best performance in the market.

Excellent. And you know, Dora, look, when I was a younger man, databases were relatively easy. Simp not easy.

They were never easy, but they were simpler. Everything was basically kind of a SQL relational database, right? Relational database, then SQL database.

Today we have time series databases, we have vector databases, we have, you know, there's no SQL databases. Of course, SEALER is very scalable. Is that scalability?

It comes from, you know, what, what kind of database are we talking here? Definitely a good question. So cila is the mostl database.

Uh, we started after some of the, many of the competitors were out there and already were defined, eh, and we are part of the ER database family. Uh, the, the database is like, uh, Cassandra and Dyno DB and the Google Big Detailss are part of the same family of databases that SVA is. And we actually have a p i comparability with Cassandra and with, uh, a WS DB databases.

And, you know, one of the hallmarks of NoSQL databases is, uh, the ability to scale and distri because they're distributed many times what in your, and I realize you're, you know, you're prejudice, you're the co-founder and c e O, but what makes sealers scalability so much better than, you know, some of the other North SQL databases that we see out here? Mm-hmm. Uh, absolutely.

And will be the last to, to force someone to use our database. And usually we'll going to say that if you have a small data set without too much high availability requirements, go ahead and use one of the relationship relational databases. Oh, there, because they're more flexible and they're better than the NoSQL databases out there.

Mm-hmm. However, if your requirements are too big and, and you hit the wall with regard to scalability, performance, and high availability, then you do need to replace it to move to a better data, uh, better database that can sustain those. So already by model, uh, we're no, c o is usually, and today the boundaries get blurred, but there, there's a cap theory where you need to select two outta three, a consistency part, particular partition tolerance and, uh, and, and availability.

So databases can only test have two of the three and s prioritize availability and partition tolerance over consistency. And we have eventual consistency. Uh, the boundaries get blurred, and we're actually, uh, in at the end of, uh, improving our consistency to, to be transactional.

But, um, the, the boundaries get blurred even within the family of, uh, a databases like Casandra and Dynamo and, uh, and others. We will be faster because we have a unique architecture of sh sharp per core. Uh, everybody, like when you have a very large data set, it doesn't fit a single server.

So everybody shard and divide the big data set into servers. We do that too, and everybody do that. Uh, the thing is that we have another le level of charting, and we, within every server, uh, in more servers, you'll have many cores, many virtual CPUs, and, uh, usually when a single C P U takes a lock, all of the other CPUs cannot touch it.

So this is the intent because, so you have consistency. Uh, the problem is that, uh, if you have a, a server with 40 cores, a hundred cores, then there's a lot of idle time in the waste that goes down the drain when you use a normal server with normal locking. We have a, uh, a, a very different architecture of one charter core.

So every single core is a separate chart within the server. Let, let's say you have a hundred cores, each of them has shared nothing, and we divide the data set in the server to a hundred pieces, and every piece is being run by single V C P U, uh, gets its own ram and also direct access to io. So we will be sometimes a hundred times faster than a traditional server.

That's the secret behind the technology, which is open sourced, and people can just go and, and watch the source forward. It's not that secret. Absolutely.

And, and that's another great thing about it as well. Good. So we've kinda laid the groundwork.

I wanted to now turn a little bit to what's, what's going on, what's new with Sila db. You guys have some news and announcements and uh, I wanted to give you a chance to share with our audience. Thanks, Dan.

So, uh, just now we closed, uh, another round of funding of additional $43 million of funding, uh, so that, that's, uh, news and, uh, we're, we're happy to Do congratulations Not only, uh, at this time, but also because it's, it was our, our choice we're near profitable, and this is some type, it's even a financial optimization. We like optimization both in the code, but also financially. Uh, so this is a good progress and it brings the total amount of funding we got, uh, ever since we started to a hundred million dollars and Congratulations.

That's, that's nice. That's a, especially in this climate, right? That's a, a very healthy race.

Yeah. Thank you to today. Investors are kind of, uh, giving money to companies that don't, may not need it.

And, and we are happily, after lots of years of hard work, uh, we we're in a relatively good position, uh, and, uh, we we're going to use that money to, uh, facilitate additional growth. Uh, we're, we have a new architecture that we we're deploying that we allow the database to be much more elastic. Uh, today we, the database has the best performance in the industry, uh, and we also like it to be really easy to use.

So let's say if you deploy the workload with, uh, a a hundred terabytes and you need to double it, then we'll be able to double it, uh, within an hour, which is really, uh, short amount of time to double a a hundred, a hundred terabytes, uh, workload. Uh, so that's a significant, uh, change. And, and the next version of Sila that we are going to launch in February, uh, we will support it.

And we also have really good results. Uh, we have, uh, good traction and good inbound coming to our platform from variety databases can be, uh, Dyna db orand, and, and lots of MongoDB, uh, workloads are co going that our way. Monga is a fantastic database for small workloads, uh, and it gets you up and running super fast.

com mm-hmm. Will go found out that MongoDB limits their growth and they needed to replace it with a faster database. And today we have customers like from, from gaming or from, from, uh, social, but also Strava.

I, I'm, I'm an avid skier, but also a mountain bike rider, and I use Strava to measure my, uh, progress and everything mm-hmm. That you measure is, uh, still is a good, uh, at scale, still is a good platform for it. You know, we, we full circle, we started the conversation talking about all the different kinds of databases out there.

And, and I think what the, the lesson is you gotta pick the right tool for the job, right? There are no necessarily good or bad databases. There's the right database for your mission, right?

As you said it earlier, if, you know, a, a SQL database for certain things is still much more flexible than, let's say a nons, seql a NoSQL database, right? Uh, if scalability and distribut are you, I don't even know if distribut iss a word, but if scalability and highly distributable are, are, you know, big, big points for you, SEAL is a great solution. And so I think that's the lesson to our audience too, is so many organizations, they get boxed into one database solution across an organization.

In today's day and age, we have so many different uses of data, and data is so important. You can't afford, I think, to just be locked into one database type of solution across an enterprise. I think you gotta use the right solution for the job.

And, and I think that's the lesson here at Sealer as well. So, well, congratulations on all of those. Um, I wanna also mention P 99, the P 99 conf you guys have coming up.

I think it's tomorrow by the time people see this, it'll be tomorrow, but it is a virtual event, so it's never too late, uh, never too late for people to sign up. Tell us a little bit about that. Um, yeah, that's a great event that, uh, we organize and it's a free event, uh, and a virtual event so people can log into it from around the world, both during the event and also afterwards.

Uh, we, we record everything. Uh, P 99 was created by us because we, we have a dedicated cinema summit event, yearly event, but, but that, that event is about set a P 99 event. Uh, before we did Thes database, we also created the ca hypervisor from scratch.

So we, we like open source operating system, virtualization and other technologies, and we were hungry to kind of, uh, spread the word about fantastic technologies. Like, like those that, uh, some people we worked in the past developed or new development in rust in rough consensus in, in go in, uh, in Linux. Uh, so, uh, we, we have, uh, more than 50 speakers.

The, the best engineers in the industry. It is an, it is a conference, uh, made by engineers towards, towards engineers, the small term managers, only technical content. And, uh, we have more than, uh, 15,000 people registered to it.

Really? That's fantastic. 15,000.

That's amazing. Can you just give us the U r L again there for people who may want to go register and attend? Uh, P 99 stands for the P 99 percentile, uh, eh, for latency, but not, it's not just about latency.

So P 99, eh, conf io. Okay. P p number nine.

io. You can check it out. That's October 18th and 19th, but you can also see it on demand after that.

Great. Do I want to thank you for taking the time to come on here. I know it's little unusual times you're living in there and our thoughts and prayers and good wishes to you and, and the rest of the team as, as well as everyone there, um, come back and see us again.

Keep us posted. We haven't had you on in a while, and it's always good to hear what's doing with our friends at SEAL db. Thanks a lot.

Absolutely. All right. Dole co-founder, c e o of CILA db, just sort of new raise, bringing their total raise to a hundred million dollars, continued momentum, some great new features and scalability.

io conference tomorrow and, and the next day, 18th and 19th of October. But we're gonna take a break here on Tech Trunk tv. We're gonna be back here in just a minute.