Supercomputing 2011 - Part Two - Inspiration & the Exhibit Hall

SC11 is nothing if not busy!  Far more sessions that you can ever attend, and all of them seem quite interesting (at least from the titles!)

I want to report back on the Education Program so far, and also the Gala opening of the Exhibition Hall.

Education Program first, today has been the highlight of the program for me so far, three sessions in particular stand out for me and I'll share a little about each with you.

First up is the Keynote that was a joint session for the Education and Broader Engagement Programs.  The speaker was Dr. Cecilia Aragon, Associate Professor at Washington University.  She spoke on two topics, one was her research which I will come too in a moment, but the second part of her talk was about her journey..

Daughter to immigrant parents from Latin America, Cecilia spoke of the discrimination & predjucies that affected her and her family.  Subtle and not so subtle things that meant opportunities were not made available that had & would be made available to a similarly performing non-minority student.  It became easy to see how the actions of one person, postive or negative, can affect participation rates and student achievement.  Cecilia gave two stories of her experiences when the family moved to Germany for a short while, the expectations that teachers have of students can have a huge impact on their performance.  A moral to her story, always have high expectations of students and don't write them off because they are starting from a position of disadvantage (real or perceived).
After university and dropping out of PhD study, she worked for NASA for around a decade as a scientific programmer, then went back to university with the help of NASA funding to complete her PhD.  She then worked as a staff scientist as Lawrence Berkley Labs before getting her academic position at Washington University, winning several awards along the way including a Presidential Early Career Research Award (the highest Early Career award in the US!).  However the story does not stop there.  Cecilia took up flying, this is despite having a fear of heights!  Not only did she take up flying, but she took up aerobatic flying which happens to include plumeting towards earth at up to 250Mph!  If that wasn't enough, she within 6 years made the US Aerobatic Team and also won medals at the world aerobatic championships.  Check out this video of her in action.

The research part of her talk looked at SUNFALL and application developed as part of the SNFactory project which helped prove the universe was expanding, a finding that won Saul Perlmutter the Nobel Prize for Physics in 2011.  SUNFALL is a "collaborative visual analytics system for supernova discovery and data exploration" and is an impressive piece of work, not just for the software but also the impact that it made on the science process.  In describing her research work, Cecilia talked about eScience (eResearch), which she didn't define as we have in NZ as using technology to improve research (i.e. science researchers do eScience) but rather the research about how research is done using technology (my interpretation & paraphrasing).  I think that this may be a useful way of looking at things; it that allows scientists to do research (rather than eResearch/eScience) and others to do research on how to do that research better using technology (eResearch).

Cecilia also had a number of other things to say, including the importance of being a mentor and mentee.  All of which provided lots of food for thought and a lot of inspiration.
I recommend that you take the time to hear Cecilia speak if you get the opportunity, failing that follow some of these weblinks to find out some more, Cecilia on Wikipedia and her staff page.

After that great start to the day, I attended two sessions in the Education Stream:
A talk on Parallel Paradigms and Decomposition and a Computational Thinking Session on Stimulating a Parallel Perspective with Bob Panoff.  These were both very good sessions.  One of my take home points from the Computational Thinking session was that scientific breakthroughts have generally come from thinking about Scientific Problems in different ways, not just doing what has been done for the past X years.  I won't go into much detail now, as I'm still digesting the information, but I'm coming to the conclusion that the technical aspects (MPI, OpenMP, Embarassingly Parallel) aren't necessarily the most important skills to teach, reasoning about the underlying problems is far more important.  That has quite large potential impacts not just for HPC/eResearch education but also undergraduate computing education given the fact that all CPUs are now parallel computers.  This is a topic that I will come back to in a later blog post.

This evening was the Gala Exhibit Hall opening, firstly wow!, the exhibit floor is huge, and in two hours I only went round part of it, I'll have to go and do the bit I didn't make on the 6th floor over the next two days.  The exhibit floor is filled with vendors and academic organisations running stands, much like a trade fair, a huge number of people were often at each stall so it has hard to get close to some.  I spent quite a bit of time at the Internet2 booth, and I want to focus on that for a bit.
Internet2 are rolling out a 100GbE core network (KAREN has a 10GbE core network) which is a world first, but that whilst impressive is not IMHO as interesting as their work with dynamically configurable network paths/circuits.  Based on OpenFlow technology that have implemented/built an network & software to control that network.  This means that you can define (via a GUI) a network setup (nodes you want to connect and the bandwidth to allocate) and in software this will be dynamically provisioned (in under 1 second) giving you a private network that you can do what you like with and not affect or be affected by anyone else.  This is a game changer.  It removes contention for a shared core IP network for applications that need/demand it, it removes the need to run an IP network even.  It allows things that could/would drown a shared resource (think Radio Telescopes and other large instruments) to be partioned off and given their own network paths.  Network Researchers can build weird and wonderful networks to run experiments quickly and then tear them down again when the experiment is over.
Their are a number of network related experiments being run over the week, I hope to see some of them in action and be able to report back, I may need to find a way to clone myself though to get round everything.

Tomorrow evening is bird of feather sessions, research posters and scientific visualization showcase.  All with some great things on offer, I'm not sure how I will decide!

Submitted by stuartcharters on