March 22, 2013
Sixth graders Emiliano and Connor, like most boys their age, love sports and hanging out with pals. They excel at academics and equally hold litanies of accomplishments that are testaments to their devout, devoted and dedicated family and faculty. Soulful breaths of innocence and inspiration, the boys bring fresh perspectives to a complicated science that even the most seasoned veterans have the opportunity to stop and take note.
Every year, starting in earnest in the 4th grade, the boys participate in a rite of passage, the school science fair. Whether we remember our adolescent experiments or present day experiences as parents with fresh memories of our own children's entries, school science fair's can be as daunting as they are exhilarating. And where some of the biggest surprises and future promise is revealed.
During the thick of this academic ritual a passing comment stopped me in my tracks. A simple question, did I happen to see the sixth grade entry "How Super Is Your Computer?" What? Marveling, reveling and giggling much like a school girl in the days that followed I was honestly giddy. The mere notion that two sixth graders endeavored to attack such a complex subject, with a set of problems and an incredibly steep learning curve that a broad community of their elders have dedicated entire educations and lifetimes on solving, for an elementary science fair no less, seemed incredulous.
Their inspiration was as innocent as their experience. "There's nothing more annoying than a slow computer," they stated! From that simple statement they first had to determine what makes a computer fast and then define a problem to solve. Immediately after thinking about "slow computers" they tasked themselves with learning about the fastest systems today and their importance to society? And what makes a computer "super".
Readers intimately understand that this seemingly simple hypothesis is not so simple. But there amongst all of the elementary entries, methodically laid out in tri-fold glory, these two dauntless young men took on Linpack, the TOP500, and represented the mystique and mathematics that are "supercomputing".
And they delivered a super science project as a result. So impressive in fact that in a juried competition, at their St. Theresa's parish school, the boys entry is one of six projects going to a diocesan-wide competition later this month.
"I was very excited to see Connor and Emiliano exploring a non traditional science fair topic using the scientific method to address a problem in technology," says Todd Council, the middle school science teacher and an acknowledged inspiration to the boys.
"It is inspiring that students at this age can see the strong relationship between science, engineering and technology and that the three judges that reviewed their project also recognized the merits of their work," Council concluded.
The lessons learned as a result of their efforts, that take a good number of us years (decades and even centuries) to internalize and rationalize, were stunning. Their research was extensive and exquisite, down to venerable sources such as industry tracker TPM, IEEE to the font choice for their display and analysis report. Their analysis was contextually rich, from both a usage and historical reference to Titan's top ranks in the November TOP500. Situational fluency in addition to expertise, as we all know, is a critical benchmark in the discipline. The curiosity is how they intuitively seemed to "know" this, highlighted by their base premise that a key ingredient to proofing super performance starts at a microprocessor level.
Their problem statement is as crisp and imaginative as it is logical. "If micro processing is so critical, shouldn't a computer reflect this speed when doing calculations?" Yes. Theoretically. "If we do our own benchmarking, comparing two computers, and one has 70 percent more processing power, shouldn't it be able to calculate the same number and type of problems at least 70 percent faster?" Indeed. Now, a veteran tendency is to jump to the conclusion and give the answer away at the offset. But remember, these are sixth graders. Fresh and shiny and new with no preformed knowledge or bias – try as marketing and hype might. What's equally important to internalize is that "most" people, i.e., everyone outside this industry, would logically presume the same. That the base performance would be at least 70% better stands to reason.
Their testbed was not dissimilar to a typical benchmark environment, just a remarkably smaller scale, testing between a dueling desktop versus a laptop, differed by a couple of ticks and tocks across processor generations, and some sizable memory. What was most impressive was their heavy duty bring-up – learning everything from variables, dependent variables, integers and floating point, Java, processor architectures and performance measurements, to making T Charts and running spreadsheets. They did it all. From scratch. A remarkable testament to their education, curiosity and perseverance.
What their research equates to is traversing about 70 years across the complex technology spectrum, which in elapsed time equates to post-doctorate eduction to master 'supercomputing.' In just a few short weeks they set about learning what has taken us years to begin to understand and rationalize. Most insightful was that for everything they learned, they footnoted the total experience with the idea that there was still much to learn. Exactly! You can't help but be enveloped by their inquisitive spirit.
According to both of them, and from an excerpt from their presentation, they sum up what for many of us rings close to home, daily, "We are still working on analyzing the results even more. This is hard and takes a lot of time to compare the percentages and make conclusions about each one. But we made some interesting conclusions based on what we saw." Indeed. And then some!
Any benchmark aficionado will understand their testing matrix well. Six different tests, twenty times each, on each system. Translated they benchmarked 100 million calculations on data-sets with three layers of complexity; short and long integer and floating point addition and multiplication. They recorded and graphed the operations and compared test averages in order to determine percentage differences, patterns and changes in speed. Then extrapolated their own performance data in context to Linpack and the TOP500.
Conclusion? "Our hypothesis was wrong and our results were surprising," they state.
They reasonably suppose that if computer A was 70 percent faster it would calculate problems 70 percent faster than computer B. Their unexpected surprise came when the results showed as much as 415 percent gains for floating point and 130 percent increase in integer performance. They determined that system A's microprocessor had been designed to run advanced addition and harder calculations.
Most intriguing was their ongoing interest in performing more analysis than time would allow for their science fair entry, including analyzing the speed between both systems in more depth and understanding more around memory's impact on overall performance. While a good portion of their research is testament to their expertise equally impressive is their acknowledgments which every good researcher knows is critical to complete works.
What's more is that ingenuity seems to run in the families, such as Connor's four year old brother, who holds a promise of his own. This budding little flower's pre-kinder entry into the very same science fair was just as amazing as that of his big brother. He starts with a profound hypothesis "Can You See Heaven From a Telescope?" As delicious as his theory is his conclusion ... simply that the amateur telescope was not powerful enough to glimpse divinity. Divine!
Whether the boys are on a STEM path that leads them to be future PhDs with careers as PIs or otherwise reveals itself in time. What is clearly proven is that they are already well on their way in an incredible journey of learnings and experience that does seem to be beyond a norm. Fingers crossed, it remains to be seen how they fair in the next heat. Regardless, they're undeniably an order of magnitude superior in my book of supercomputing heroes!
10/30/2013 | Cray, DDN, Mellanox, NetApp, ScaleMP, Supermicro, Xyratex | Creating data is easy… the challenge is getting it to the right place to make use of it. This paper discusses fresh solutions that can directly increase I/O efficiency, and the applications of these solutions to current, and new technology infrastructures.
10/01/2013 | IBM | A new trend is developing in the HPC space that is also affecting enterprise computing productivity with the arrival of “ultra-dense” hyper-scale servers.
Ken Claffey, SVP and General Manager at Xyratex, presents ClusterStor at the Vendor Showdown at ISC13 in Leipzig, Germany.
Join HPCwire Editor Nicole Hemsoth and Dr. David Bader from Georgia Tech as they take center stage on opening night at Atlanta's first Big Data Kick Off Week, filmed in front of a live audience. Nicole and David look at the evolution of HPC, today's big data challenges, discuss real world solutions, and reveal their predictions. Exactly what does the future holds for HPC?