The Launch of X Cloud with ScyllaDB’s Dor Laor

Hey, everyone. Welcome back here to Techstrong tv. You know, I haven't had my friend Dora on here in months, may, maybe even a year.

Let me introduce you to, uh, do Leor Do is the CEO and co-founder of Steeler DB joins us today from, from Israel. And welcome, do, and thank you. And, uh, we appreciate you coming on.

How's everything with Steeler db Alan? Good to see you. Good to see everybody.

Thanks for hosting me. Uh, Silicon is doing great. Uh, we're working our behind off in order to, uh, improve our releases.

Uh, so all, all good. We re recently released, uh, X Cloud, which we'll probably talk about and, and variety of other things. Very cool.

Do just for people who maybe are not familiar with you and cela, give them a little bit of your history and a little bit of the history of CELA db. Sure. Um, so I'm an engineering.

My roots always been, even now that I, I don't write code. I involved in the product technical person. And, uh, in, in the past, uh, I built a terabyte router company and, uh, later on I partnered with my co-founder, Avi, and we created the KVM Hyper hypervisor.

And that was 20 years ago. Uh, mm-hmm. Long time ago.

We were involved with, uh, like we brought KDM from the beginning and we highly were involved in Linux after Red Hat acquired that startup. And, uh, more than 10 years ago, we created CA db. Uh, cila is a highly scalable database company.

It's a no SQL distributed database. Absolutely. com and everything is there.

You can, people can download and, uh, kick the tires. Absolutely. And for tho from you people at home watching it, it's spelled just the way it is under do's name on the video.

com. Alright, we got that all out of the way. Do you guys recently announced, as you mentioned, uh, X Cloud?

Correct. What, what's X Cloud? So, um, uh, we have, uh, our order release, which you consume, uh, eh, with a self-managed version, the, the enterprise version or so valuable version, and also a fully managed database as a service in the cloud.

It's called C Cloud. Uh, recently we released a new one, a new release called X Cloud. What's special about it, people love DX these days.

So we added X to the product, but, but really what the X does is, uh, allow us to be the most elastic database in the market. People need elasticity. Uh, why would you need elasticity?

Because, and, and number one, you first need the database that can perform and can scale with the high availability to be a mission critical, uh, database. This is the, the, the number one need afterwards when you, uh, start to use it, then, um, uh, if you scale to a very high extent, it costs money and we like to make money, but sometimes you, your peak level varies and the usage varies. Uh, there are cases where a Black Friday or similar event come around and, and then you need to scale your deployments.

Uh, this is more predictable scale. There are unpredictable scales that, so sometimes you, you have a really good success and then many users come to the web, to the websites and you need to scale a database. In many cases, the, um, the elasticity requirement is daily people come, uh, wake up, go to, uh, our customers websites and, and our customers mobile app and whatever they do, generate loads.

Uh, during noon they eat, so load decreases a bit. Then there's the second pick of the day, and in the evening they go home and rest. So there's, if you pro provision 100% for the peak, you are wasting on resources.

A on the other hand, uh, traditional databases and prior to this X Cloud release, uh, if it's very hard to, uh, always me with the, uh, number of, uh, servers and infrastructure related to, uh, a, a deployment, we, we have deployments, uh, at the side of the pet petabytes. So moving around petabytes, uh, within, um, within several hours is extremely externally difficult task. Uh, x cloud is release that allow us to do it faster and allow us to double or quadruple, uh, the throughput for particular customer within minutes.

Um, so that this is what's the, the value of X Cloud. I can talk about what is there under the hood as well. Excellent.

Um, you know, in many ways this whole idea of this elasticity, I feel deja vu. I was told this about the cloud in general, right? That was one of the things about in the cloud, you burstable, but it, but what it proved to be in the cloud door was like, if you blow up a balloon, you know, and it stretches out, and then you let the air out of the balloon, the balloon is still stretched out.

And, and for so many people, elasticity on the cloud is sort of a one way ray, a one way, you know, direction. Yeah, you can always make it bigger, but how many people really do make it smaller? Is this something also true in sealer or, or can?

Is it truly that elastic True A, absolutely true B, b, everything you, you, you, uh, you you've described. Um, and it's also true that even regardless of the infrastructure and how cool X cloud is, many times people just keep on adding more and more data. So if, if you just add and you don't delete, then it is what it is.

But, uh, that when we have different capabilities, uh, there are normally two reasons why you, you'd want to scale out and, and why you'd like to scale in. Number one is storage. So if your storage grows, uh, uh, will will scale out and add resources.

If you delete data, we'll automatically, uh, scale in back again. And, uh, our current release, uh, have two unique things. Number one, it has, uh, autoscale is the storage, uh, the, the storage automatically scales and, and, uh, our compute is bundled together with the storage.

At the end of the day. I'll tell you a secret. We run servers, the thing that's called servers, and the server has, uh, memory and disk and networking and compute altogether.

That's some of the reason we have really good performance. We bundle compute and storage. Um, so, so, uh, if the amount of storage people save decreases, we automatically decrease the amount of servers.

And, uh, we do it with 90% utilization. Uh, so the server, the, the disk capacity can go up to 90% of, uh, of the shared, uh, eh, cluster infrastructure. It, it's very high percent.

Now, previously we went, uh, from 50% to 70%. Now, because we're very elastic and we can move data very fast, uh, we can go up to the very end of, up to 90% and still allow people to use it as, uh, and, and it's all automated. So when you go, uh, eh, to 90%, we automatically provision more servers.

If you go below 90%, there is some threshold like 85, then we automatically decrease servers. And we also do it in a way where, um, let, let's say if you have, if you choose to have very big servers, le let, let's say you have three gigantic servers. We, we support in over servers up to 256 CPUs.

Uh, if you add, uh, if you run out of space and you'll add another gigantic server, then suddenly your utilization will be low. So it's, it's bad for you to add a big server if, if you just need like another one or 2% of utilization. So what X cloud is, is doing, it automatically selects the server size for you, and we mix server sizes.

So if you only need extra 5% along with the, these very big servers, we are going to add a small, tiny servers with, uh, two vcps next to those other big servers. And we will keep, replace it automatically for you without you knowing as long. And, and it translates to value.

So we need to pay the, the total cost of ownership will be low because we were targeting 90% utilization of this capacity. Uh, the, the other, uh, reason why to scale is sometimes, uh, throughput and CPU consumption, and this is even easier because we, we need to move less storage. So it's, uh, provision new servers with less storage, that that's easier for us.

And we do that and we allow our customers to, to scale in and out multiple times an hour. Got it. That, that's a great, very in depth.

Thank you very much for that. You know, do all the news today is with ai, ai, agentic ai, generative ai, LLMs underlying all of this though, is, is data, is is the, the, the, you know, that they're training on, that they're storing, that they use. How has this whole kind, you know, it's a whole world unto itself.

How has this affected the steel of business? Um, it, it's definitely drives our business. Not not only ai, but absolutely, uh, uh, it helps to drive our business.

We, we, we have three main, uh, pillars with, uh, related to ai. And number one, we have traditional machine learning. We have feature store, so everybody who needs to customize automatically their, um, uh, their infrastructure and, and fit it into customer use cases automatically.

And segment that. There is a industry standard called, uh, machine learning feature store, and CLI is used there. For example, trip Advisor is a customer of ours, and it is feature store in order to find the best recommendation and the, the best deals and the best, uh, advices to their customers.

This is number one. Number two is, uh, database usages, uh, AI usages that needs, uh, large scalable storage underneath. W we have a very large, uh, um, vehicle vendor, which, uh, I cannot say their name, but it's one of the largest.

And, uh, and they use, uh, LOC to train their AI model for, uh, for, uh, for, for, um, automatic AI sales driving cars. So that, that's another, it's not just, uh, vehicles, but it can be anything else that drives a lot of data. They need to access a lot of, a lot of data and a lot of objects, standard access pattern in, in databases, but it happened to drive ai.

And the third one is, uh, vector search. Uh, so vector search is, is a component that supposed, uh, a rug, um, automated retrieval to together with the agent, agent agent. So you can run your query, but, but your query, uh, your query, you, you are going to query a customer of ours.

And, uh, the, the query should automatically go to private repositories that, uh, chat GPT cannot access and augment the data retrieve, like, uh, um, when is my slides, uh, will going to land? Uh, it, it needs to fig to retrieve that, uh, your ID to figure out where is your particular flight is supposed to, what is the number? And then get, get to the predicted, uh, uh, landing, for example.

And, and this is what, uh, vector search, uh, allows databases to do and, uh, will GA or vector search, uh, product by the end of the, the year it's not yet. Uh, it, it's only, uh, close beta. You heard it here.

Okay. Thanks for sharing that with us. So I, I just wanna make sure we hit the, the main points of this X Cloud offering.

So with, with all of these things that you'd mentioned already, really what this results in right, is you, you, you've improved compression, improved streaming. So not only are you helping to reduce the storage, you know, cloud costs, but it it's the network as well, right? Your bandwidth and, and everything else that, that comes in that, and that's an important piece of the, of the equation as well.

Um, the other thing I want to make sure our audience understands if they're not familiar yet with CLA is all this is offered as a basically database as a service, correct. DB as a service. So you don't have to worry about your infrastructure.

When we talk about storage, it's, it's all part of the sealer offering, right? You don't need to do all that. Absolutely.

We, we, we have high amount of automation that everything is done on behalf of the end user. All, everything. We, we don't do anything.

Our, uh, a manual ourself, everything. It's, it's, uh, the, the elasticity level is supposed to be that, uh, uh, the cluster is breathing, can, can breathe all the time. You just need to set, this is my utilization, this is my ex expected set of, uh, SAP power, and that's it.

The, the cluster will continue from that point onward. Um, as, as a customer, you just need to be connected and everything runs under the hood, including a b backup. For, for, for example, uh, if a customer, for example, runs at, uh, 89% utilization, uh, and, uh, we run a, a, a daily backup.

Uh, so, so could be that the backup will, uh, generate a snapshot and the snapshot, uh, will generate some data and, uh, temporarily, uh, will go beyond 90% capacity. Uh, this, the, the second will trigger that automatically we'll provision more servers, uh, to the cluster. Uh, because it's just an event of crossing 90% utilization.

Uh, once the backup, uh, will be complete and, uh, the image, uh, backup, uh, snapshot will be loaded to S3, uh, then the image gets deleted, uh, the, the cluster will have less than 90% utilization and we can automatically decrease the amount of servers. Everything is automatic. It's uh, it's really, uh, fascinating to see all of those use cases run under the hood and to see the metrics and to see it happening.

Absolutely. Just wanna also make sure Dora X cloud, this x new release, this next generation available now It's available now. Correct.

Just go to the website if you saw. And what about existing sealer customers? Do they automatically upgrade to it or no?

Uh, it's, it's not like, 'cause the, the previous generation you used to, uh, um, did divide the data in, in using some older algorithm. And do we have, um, migration process in order to move between the old generation to the new one? Uh, it, it's, uh, an automatic process, but, uh, it, it's needs to be, uh, it is a procedure we, we do together with the customer.

Excellent. Alright, Dora, I think we're about outta time. Thank you for coming on.

Congratulations on this next generation of SEAL db. You guys always stay one step ahead. It seems of, of what's going on there at the leading edge.

So keep doing what you're doing. com, check it out. We're gonna take a break here on Tech Drunk TV.

Will be back in a second.