InterviewsTechnology & IT

SQream Technologies – Removing Limits of SQL Databases

We caught up with Ori Netzer, VP Product Management at SQream Technolgies to discuss the latest news on the product and technology they are building.

Who are you and what are you doing?

We are SQream Technologies, a Big Data startup from Israel. We have developed a patent pending technology for Big Data applications that boosts analytics performance of traditional SQL database systems through Massive Parallel Computing using GPUs (graphics processing unit). Our product is currently the fastest SQL-based analytics engine on the market. Our solution can be used with any type of relational database and standard hardware.

Let’s look at the technical details. We use existing hardware but instead of running SQL statements with a server’s CPU we use its GPU. A GPU works similar to a CPU but it’s more specialized and based on a totally different architecture. While a CPU has up to 8 cores, a GPU can have up to 2,880 which makes it up to twenty times faster compared to a CPU. You can think of a CPU like an Einstein, very smart and able to do everything. On the other hand a GPU can be thought of as a team of Usain Bolts who are world class, but at their respective disciplines only. In short this opens up a massive opportunity for parallel computing.

While GPUs are normally used for video rendering we have found an efficient way to run SQL statements on them. Therefore we break down the computing tasks into smaller subtasks and assign them to the individual cores of the GPU and maybe also across different GPUs in a cluster.

For the database user there is no apparent difference compared to their existing solutions. They can continue to work with their tables, schemas and queries. They only run up to a 100x faster.

A typical usecase we see a lot is that customers want to implement Big Data applications but are hesitant to change their existing infrastructure. In that case our software can be used with the customers’ existing hardware or they can add inexpensive standard hardware and continue working with their data structure and the tools they are used to.

How did you come up with the idea?

The idea came from Kostya Varakin our CTO. He is from Russia and used to work in high performance computing and while he was developing trading algorithms he found an efficient way to do Cartesian Mathematics on a GPU.

Then he met Ami Gal, our CEO who was then looking for companies to invest in. Since he had done some work in the field before he was amazed by the idea and they joined forces the build SQream Technologies.

What is your business model?

We sell software. Our analytics engine runs on any standard hardware setup like Dell, IBM, HP or any other hardware vendor our customers may choose.

How are you different from your competitors?

You can say that today nobody other then us is able to efficiently run generic SQL statements on a GPU. Only a few Algo Traders or Oil & Gas companies are using GPUs for their highly specialized algorithms. For each application they need to write code specifically tailored to their algorithm and the GPU. But writing code to run efficiently on a GPU is very complicated and many big companies have tried and failed at it already. At SQream we have filed 8 patents for our technology which are in different stages of approval.

If you look at the performance side, we’re in essence removing the performance limitations of relational databases that competing solutions have. Take for example a leading RDBMS which will work with roughly up to 10 Terabytes of data. At the same time we can work with up to 50 Terabytes. Recently we ran a test doing a joined of thirty million subscribers with thirty five billion revenue records. With a traditional SQL-based solution this would have basically been impossible.

A recent external benchmark also compared a Star schema on Hive with the same schema on Red Shift. Red Shift was ten times faster and our analytics engine was four times faster than Red Shift.

What are you looking for? Investment, talent or customers?

Basically all of the above but in particular female talent. We have a great team right now of 26 people and we would love to bring more female talent on board.

Also, from our experience good developers are especially hard to find. Since we are developing specialized and cutting edge technology, C++ is probably the only better-known technology we use. We also rely on Haskell and CUDA a lot where it is much harder to find exceptional talent.

Why do you think there are so little women in tech?

Rachel Wolfson, Marketing Manager at SQream, jumped in on this one: “I don’t know for sure, but I think it is very hard for women in Tech today because the industry appears to be dominated by men. However the opportunities for women have really increased, especially in Big Data and Data Science and therefore I highly recommend it as a field to get involved in. Just don’t be intimidated by it! I love working with SQream as a woman, Rachel added.”

Where do you see SQream Technologies in one year?

We are aiming to be a 9 Billion dollar company which will be very hard work. You won’t find us drinking mojitos on the beach any time soon but working hard with our customers to deliver value and continue building a unique product.

Thank you very much for your time!

The interview was conducted by Daniel Nippes.

Image credit Sqream Technologies

Previous post

Six Data Scientists You Should Know

Next post

Map-D Wins $100k in Nvidia Emerging Companies Contest