Kinetica, a provider of an in-memory database accelerated by GPUs, has announced the availability of in-database analytics via user-defined functions (UDFs).
This capability makes the parallel processing power of the GPU accessible to custom analytics functions deployed within Kinetica, opening up the opportunity for machine learning/artificial intelligence libraries, such as TensorFlow, BiDMach, Caffe, and Torch, to run in-database alongside, and converged with, BI workloads.
“What we are offering are UDFs that are accelerated by GPUs. This enables us to benefit from the magic that comes from some of the deep learning and AI libraries that are out there, as well as taking the UDFs that customers have been using for years and moving them into a distributed model with GPUs,” said Eric Mizell, VP, Global Solutions Engineering, Kinetica.
The reasoning behind this is not only that customers have asked for it, but also that there is a gap in the market, said Mizell. There are not many data scientists and companies are challenged to get the most use from their data. If organizations can “democratize” data access and make it easier for advanced BI users to call these libraries, then they can start to get more value from their data, which is easier than hiring a data science team.
“That is the next step,” said Mizell. “How do we make it easier for organizations with very strong business analysts, and bring the algorithms to them in an easier to use fashion? That is the vision of where we want to take this.”
Kinetica also introduced a new interactive visualization tool called Reveal for real-time data exploration.
With Kinetica’s Reveal data exploration framework, business analysts can make faster decisions by visualizing and interacting with billions of data elements instantly. Users do not need to know SQL and can simply drag and drop data tables to slice and dice data and start creating on-the-fly data analytics. Reveal has more than a dozen analytical widgets to choose from for creating interactive real-time dashboards with just a few mouse clicks. It also features enhanced mapping capability and integrates with major mapping providers, including Google, ESRI, MapBox, and Bing, to conduct interactive location-based analytics on massive datasets, and provides enhanced security with fine-grained multilevel access control for permission-based widgets, views, and dashboards.
“We are starting to see a very strong need for location-based analytics. Providing organizations with a tool that is easy to use with drag-and-drop widgets that let them dive in and out of location-based data is very useful to them,” said Mizell.
In addition, users of the database can now take advantage of Kinetica’s VRAM Boost Mode, which allows them to prioritize their data tables and can force datasets to always sit in very fast cluster-wide GPU Video RAM (VRAM) for fast query performance. This gives customers better performance, while also still being able to leverage cluster-wide system RAM to both scale up and scale out to multi-terabyte in-memory processing.
“We use system memory and GPU memory so it is a mixture of the two. For certain workloads, VRAM memory is faster because it is closer to the GPU,” said Mizell.
The new UDF capability, Reveal data exploration framework, and VRAM Boost Mode are available in version 6.0 of Kinetica’s GPU-accelerated database.
Learn more at kinetica.com.