A Q&A with IBM Research’s Hillery Hunter
IBM’s latest Power servers come with an AI twist. Optimized for deep learning, a new so-called PowerAI toolkit will “help train the systems to think and learn in a more human-like way, at a faster pace,” as announced at SC16, the International Conference for High Performance Computing, Networking, Storage and Analysis.
I spoke with Hillery Hunter, IBM Research’s director of Systems Acceleration and Memory and Memory Strategist, about her team’s contribution to the software behind the world’s fastest deep learning servers.
How was the AI added, and what does that mean for how a Power system functions?
Hillery Hunter: In this case, AI refers to a collection of deep learning frameworks and algorithms. Today’s launch represents our first public offering in hardware-software co-optimization for deep learning. Researchers have worked closely with IBM’s systems engineers to create code that is optimized to the Power S822LC (IBM’s highest-performing OpenPOWER server).
We used the S822LC’s unique design – including high-bandwidth NVLink interconnect, not just between pairs of GPUs, but also between the GPUs and the CPUs – to deliver higher deep learning training performance.
This means that the co-designed PowerAI hardware and software can build learned models from images, speech, or other media in less time than prior generations of hardware and software. Deep learning training time is a key metric for developer productivity in this domain. It enables innovation at a faster pace, as developers can invent and try out many new models, parameter settings, and data sets.
IBM Power S822LC for HPC servers support of the Human Brain Project, a research project funded by the European Commission to advance understanding of the human brain. IBM, NVIDIA and the Juelich Supercomputing Centre delivered a pilot system as part of the Pre-Commercial Procurement process. Called JURON, the new supercomputer leverages Power S822LC for HPC systems.
What does this mean for those using Power systems, today?
HH: We’ve worked to make access to PowerAI as simple as possible by including all related and dependent packages into a single distribution. Anyone who would like to take advantage of the benefits of our high-speed deep learning training can download the popular deep learning frameworks from the PowerAI landing page to run on their Power System. You can also try PowerAI, including the popular deep learning framework Caffe, through IBM’s Power HPC Cloud partner, Nimbix, which is offering S822LC servers in the cloud.
What did IBM Research contribute to the “AI” of PowerAI?
HH: IBM Research has a long history of contributions to hardware, software, and solutions for Power Systems – from processor microarchitecture innovations through system power, packaging, and cooling. This was a global effort, including engineers and scientists from our teams in Tokyo, India, and here at the Thomas J Watson Research Center.
For PowerAI, we have contributed new deep learning framework algorithms and performance tuning to help highlight the features of the S822LC server. The advantages of high-bandwidth NVLinks between the GPUs and to the CPUs, combined with the IBM Caffe deep learning framework optimizations, result in lower deep learning training time.
What implications does PowerAI have for Open Power?
HH: Open Power opens up POWER-processor based system designs, and so provides an opportunity for our partners to build domain-optimized systems. We hope that PowerAI inspires many others to continue to use the CPU-to-GPU NVLink capabilities of the POWER8 processor, both in algorithm co-optimization with hardware, and in novel system designs.
Where will the AI be most-readily apparent?
HH: Deep learning is at the heart of the way many of us are already experiencing cognitive in our everyday lives, most notably in voice and image recognition. Moving forward, deep learning will fuel technology in many more parts of our lives, from cancer detection to human-like conversational capabilities for computers and mobile devices.
PowerAI further democratizes deep learning for our users such as banks, to improve upon tasks like fraud detection. Or within our cars, especially as they become more autonomous. These advanced systems need a system to develop AI that’s just as advanced so they function seamlessly in the real world.