Two weeks ago, I left the warm confines of southern Utah and headed north to Salt Lake City. The Supercomputing 2016 conference was in town for the week.
I’ve attended conferences both large and small going back 20 years, but nothing could prepare me for what I saw at Supercomputing. If you’re not familiar with Supercomputing, it’s an annual conference where scientists, researchers, and engineers gather to discuss high-performance computing, network storage and related technologies.
This week I’d like to talk about a few of the highlights I saw at the conference. I spent the first day walking through the exhibit hall. Booths ranged from large companies such as Intel and NVIDIA to small companies I’ve never heard of. Quite a few universities had booths as well including Utah, Stanford, Notre Dame, and BYU.
I spent the next couple of days sitting through presentations, and was surprised to see so many companies offering large-scale storage solutions. Dell/EMC were there to showcase their NAS solutions optimized for HPC environments that include large-scale data sets. These are pricy enterprise solutions that combine hardware and software into a single solution. These products aren’t as exciting as others around the hall, but they play a critical role in HPC and machine learning.
On the other end of the spectrum, I saw through a presentation at the Red Hat booth about how decoupling the storage hardware from the software was a good thing. Western Digital and Seagate were also there showing off their mid-range storage appliances. Most products I saw focused on capacity over speed, but not the following product from PEZY Computing. It was all about speed.
All flash storage brick consisting of 1520 1TB M.2 SSDs from PEZY Computing
I sat through a presentation from NASA where the scientist explained how lens and capture technology has changed over the past decade. Some of the most powerful and expensive cameras are floating in orbit, capturing high-res images of the universe. The size and sheer number of images amount to petabytes of data which must be stored and scanned. This where machine learning comes in. It was all a bit mind-bending at times.
NVIDIA had the largest and most popular booth at the show. I could have spent the entire week listening to their numerous presentations. Those I did hear were fascinating. One man spoke about how companies are using NVIDIA GPUs to create more aerodynamic vehicles and aircraft.
NVIDIA has positioned itself and its chips to be the brains behind machine learning. Intel ushered in the first computing revolution with their ubiquitous CPUs, driving computing performance to new heights. But NVIDIA is the engine that’s driving machine learning today with their line of high-performance chips. Unlike the CPU which is designed to do a lot of different tasks, the GPU is optimized to perform just one or two tasks as fast as possible. What I saw at the show was Intel Xeon chips driving consoles and servers, but NVIDIA Quadro and Tesla doing the bulk of the heavy lifting in vast arrays.
I also saw new ways to cool hot running GPUs, especially those in large server arrays. 3M was showing off their line of two-phase immersion cooling. Part of the fun in attending Supercomputing is seeing products that might find their way into more mainstream products over the next few years.
On the last day of the show, I listened to two presentations, both of which had a big impact on me.
A student from Stanford University gave the first presentation. She explained how machine learning is changing healthcare at a rapid pace. She told how machines are able to read x-rays to detect cancer cells. Early detection drastically increases the effectiveness of treatment and the patient’s chance of survival, and machines can be trained to recognize even miniscule changes from one day’s chart to the next.
Near the end of the presentation, one man raised his hand and asked if she was referring to “pie in the sky” technology or products that were available today. Her reply: “This technology is saving lives today.”
The next presentation was about how machine learning is helping usher in changes to transportation such as self-driving cars. This really hit home with me because my mother has suffered four strokes over the past six years. Each stroke has done a little more damage than the previous one to the point where her memory is quite poor. Due to that and other factors, she’s not able to drive which means she doesn’t leave home very often.
I don’t know if my mother will live long enough to summon a car to take her to grocery store or to visit her children. I suspect she won’t. The technology exists today, but faces regulation and cultural hurdles before anyone can use it. Until that time, I’m glad FaceTime exists which allows her to interact with her grandchildren.
Supercomputing offers a glimpse into the future. That future is full of big technology bets. Bets on chips and storage and software. But it’s also full of advancements of a more personal nature. These advances save lives and improve day-to-day living for people like my mom. That’s what I’ll remember most about Supercomputing 2016.