by Cathy Chang

The value of business data is increasing dramatically. So, it was timely that this month’s Silicon Valley HPC/GPU Supercomputing Meetup Group (see below for more information about them) featured a talk by Ren Wu, Research Scientist at HP Labs, who spoke about using GPUs to accelerate business intelligence (BI) analytics.

BI analytics, also known as advanced analytics, refers to computer technologies used to mine relevant business information from raw data. BI analytics can give a company insight into historical, current or future-looking business operations.

During his talk, Ren illustrated the technology trends and challenges surrounding BI analytics and how he led a team to study the viability of GPUs as BI accelerators. He proposed that the BI space is moving towards using massively parallel accelerators, GPUs (where he believes that the “G” really could stand for “general purpose”), for BI analytics.

According to Gartner, as of 2010, advanced analytics is a top technology that can’t be ignored, trailing only cloud computing. The BI market is at $11 billion and Forrester predicts it will grow to $14 billion by 2014.

Back in 2007, an IBM exec spoke about how organizations that are able to sift through data quickly (some in real time) are gaining a competitive advantage over others. For example, the NYC Police Department was able to solve crimes and capture suspects more quickly with a real-time Crime Information “Warehouse” that helped them more quickly gather, share and act on updated reports and evidence. Many examples exist in the business world, as well.

Ren further illustrated experimental designs and results that showed an order of magnitude performance advantage for GPU over their optimized CPU implementations. This led his team to conclude that GPUs are viable accelerators for BI analytics. Ren’s presentation on this topic at GTC 2010 can be found here (with downloadable video here), his supporting discussion slides are here (PDF link) and his earlier interview via CUDA spotlight is here.

While advanced analytics is an attractive space for acceleration, there are also challenges. Large datasets are a given for advanced analytics. The question of whether GPUs would help accelerate BI analytics over CPU-only solutions is intriguing: How compute-intensive are the usage scenarios? Would using GPUs yield much faster response time addressing much larger datasets in advanced analytics? What does it take to transform massive-scale, multi-channel, multi-modal data into pervasive, always-on BI?

We would love to hear your experience using GPUs to accelerate analytics. What were your challenges? How did you overcome them?

About the HPC/GPU Supercomputing Meetup Group:
As this is just one of many stimulating talks and discussions at this Meetup group, here is some background: Jike Chong initiated the Silicon Valley group in February 2011, the group has attracted 200+ members, many of them pioneer parallel software practitioners who are passionate about high performance computing and GPU supercomputing. If you find topics in these areas interesting, join the discussion on Monday, Sept. 12.