Skip to main content
    An Interview with Richard Heyns, CEO of Brytlyt
    Written by Brytlyt

    An Interview with Richard Heyns, CEO of Brytlyt

    As published on Superb Crew:

    Brytlyt’s GPU Database and Analytics platform has been making the rounds lately after benchmarking results showed them to be the fastest GPU Database on the market. They’ve recently partnered with MariaDB to integrate their product and enable millions of MariaDB users to analyze billions of rows of data in milliseconds.

    Brytlyt has also partnered with IBM to offer their GPU Database software on IBM’s POWER hardware for those Enterprise customers looking to get even more from their GPU Database. To find out more about Brytlyt’s product and understand their vision, we sat down with Richard Heyns, CEO & Founder of Brytlyt.

    Q: Could you provide our readers with a brief introduction about Brytlyt?

    A: Yes of course. Brytlyt is a GPU Database & Analytics platform that is transforming the way businesses are using data. As companies are collecting, consolidating and combining their data from diverse sources, they want to analyze this data quickly and efficiently, so they can stay ahead of market demands, understand their business needs and get more and more from their data. With Brytlyt, multi-billion row datasets can now be queried and analyzed in milliseconds. In fact, independent benchmarking has found us to be the fastest GPU Database on the market today.

    Besides harnessing the astonishing performance of GPU’s, Brytlyt is the only GPU database to have properly solved the problem of doing JOINS on GPU, and we are the only vendor with patent-pending IP for doing JOINS. Brytlyt is originally a PostgreSQL fork and so offers easy integration with existing systems and users can add or remove GPU resources at will to quickly respond to fluctuating workloads. On top of all that, we are very proud that Brytlyt has exceptionally rich functionality and is easy to use.

    Q: Can you tell us a little bit more about the IP behind Brytlyt and how does Brytlyt do JOINS vs. your competitors?

    A: Sure. Traditionally, the most significant hurdle for GPU Database vendors has been to figure out how to achieve parallel processing when joining data. Joins are incredibly important because they establish the relationship between different sets of data and are crucial for meaningful analytics. The traditional approaches were designed years ago for single-core CPUs and are not well suited for the hundreds of thousands of cores in a GPU system.

    Brytlyt is the only vendor to solve this hurdle using our patent pending method that recursively separates rows of data containing essential information from rows that do not. This IP allows data processing operations that are not important to be identified and immediately excluded from the process. The result is a super-efficient, super-fast approach that can be run in parallel and on GPU.

    Q: You talk about collecting, consolidating and combining data from diverse sources, how is Brytlyt’s technology helping with that?

    A: There are many ways Brytlyt can collect and consolidate data sources. Since Brytlyt is based on PostgreSQL, existing investments in code and visualization dashboards can be immediately accelerated with very little additional effort. What is more, Brytlyt is a PostgreSQL fork and so has native connectors that allow it to connect directly to virtually any existing data sources. All this means it is very easy to integrate Brytlyt with their current investments, lowering time-to-value and improving ROI.

    Q: Talking about MariaDB, what was the primary reason for partnership with MariaDB? When can users preview Brytlyt’s capability on MariaDB?

    A: Our mission at Brytlyt is simple. We want to put Brytlyt’s technology into the hands of anyone who wants millisecond query response times. MariaDB with its commitment to open source and community collaboration gives us an excellent platform to put our technology in the hands of millions of users who can immediately benefit from the performance gains.

    Q: Can you give us some use cases where your customers have gained benefits from using Brytlyt’s GPU Database or Spotlyt, your visualization platform?

    A: Sure, let me start by telling you a bit about SpotLyt. SpotLyt is our analytics workbench, designed for real-time visualization of massive data sets. We built SpotLyt because although Brytlyt works with all the major visualization tools, we found none of these tools are able to handle large amounts of geospatial data or queries very well.

    The most interesting use cases we’ve been working on recently are in mobile and retail and also with companies looking for real-time insights on geospatial workloads.

    For telcos, the issue is polling smartphones, assessing customer metrics, optimizing network infrastructure or monitoring network usage and capacity. We also see our Telco customers using Brytlyt for real-time fraud detection, monetizing subscriber data, and even improving customer retention.

    Within retail and their huge datasets, the main issue has been extremely large SQL joins, and aggregations and Brytlyt has a very innovative approach to this. Brytlyt can do more than find the needle in the proverbial haystack; it can process the entire haystack. Counting the distinct number of customers is a great example of the kind of workload that is truly important to the retail industry.

    Q: That is all very impressive! Can you share some of the new features in the pipeline?

    A: What we are most excited about is what we’ve done with SpotLyt Workbench. SpotLyt 2.0 brings together immersive real-time visualizations, a powerful SQL editor, and the Torch AI framework – all running on GPU and accessed via a browser. Simple, easy to use, but immensely powerful.

    We’ve also added a raft of new stuff to BrytlytDB. For starters, we are moving from being GPU centric towards a heterogenous solution that brings more CPU resources into play and means Brytlyt can address even larger datasets. We’ve improved indexing, added range join, extended datatypes, improved stability, enhanced COUNT DISTINCT not to mention a bunch of other cool stuff.

    Back to Blog