Fuzzy Thinking about GPUs for Big Data

There are plenty of companies, big ones and startups alike, that are trying to bridge the gap between big data and fast data. These same vendors are also trying to move analytics from reactive to predictive as well trying to provide the power of supercomputers on a simple desktop.

Fuzzy Logix, profiled here by NVIDIA in their startup series, is one of those vendors who are emphasizing GPU-based computing in order to achieve those goals.

GPU has been an exciting topic in high performance computing of late, with Datanami Managing Editor Nicole Hemsoth writing the following based on a talk with NVIDIA senior manager Sumit Gupta: “The GPU and MapReduce combination has been the subject of some notable research and is finding its way into more popular use in GPU environments, says Gupta. Azinta Systems founder and advocate for using GPUs to boost large-scale data mining, Suleiman Shehu agrees that GPUs can revolutionize large-scale data mining by significantly speeding up the processing of data mining algorithms. He points to the K-Means clustering algorithm as a prime example, stating, ‘the GPU-accelerated version was found to be 200x-400x faster than the popular benchmark program MimeBench running on a single core CPU, and 6x-12x faster than a highly optimised CPU-only version running on an 8 core CPU workstation.’”

Specifically, the North Carolina-based startup is focused on building the foundation for GPU-based analytics models. According to the profile, the company has developed code for over 500 in-database models. From those, programmers would be able to build higher-order applications on top.

Further, the models are developed in all common programming languages, such as Java, C, .NET, and others. The intention is to quicken the process between IT application development and usage by the normal end user. The fact that these applications are GPU-based make them hypothetically faster to run than on regular CPUs.

The use cases here represent a decent amount of potential. For example, according to the profile, Fuzzy Logix is helping call centers by determining which representative would be best to handle the customer, not by asking the customer to wander through a various slate of menus but by taking their location, demographics, and past experiences, aggregating that data, and thus assigning a representative analytically.

As a company, Fuzzy Logix is focused in an area that is highly competitive right now in big data: taking the applications from IT guys and data scientists to users on the business end. Chief Operating Officer Michael Upchurch believes GPU technology will be key to that effort.

“Working with our clients, we are consistently identifying general business challenges that can be best solved by leveraging GPU-based analytics,” said Upchurch. Of course, it is one thing to talk about how GPU will be key, it is another thing entirely for a vendor to deliver.

As of right now, the results are relatively impressive. Their appliance, which runs on four NVIDIA Tesla cards, is reportedly able to perform a billion calculations in just 13 milliseconds. According to the report, this efficiency surpasses that of CPUs.

Related Articles

The GPU “Sweet Spot” for Big Data

IEEE Tags Top Tech Trends of 2013

20 Lessons Enterprise CIOs Can Learn from Supercomputing