This month I embarked on a bit of an experiment by attempting to answer the question, “Can I successfully complete a nearly three-week, international, technical business trip without a laptop?” So, off I went, iPad in hand. No keyboard. No mouse. No safety net.
It all comes down to the fact the computer age is evolving into the mobile age right before our very eyes. While my little experiment hasn’t been without issues, it’s gone significantly smoother than expected. It seems I now have apps for everything I commonly do (including SSH). In fact, I’m writing this blog entry using my iPad’s soft keyboard right now.
So, why does this matter to HPC? Surely, supercomputing is wholly entrenched in “real” hardware. We sold our souls to the hardware manufacturers many years ago, didn’t we?
HPC Collaboration Through Open Source
One of the volunteer activities in which I am involved is Berkley’s BOINC. Started in February 2002, BOINC provides an open-source mechanism for distributing certain classes of HPC workload to a variety of different, geographically disperse compute “devices.” While some of these devices are enterprise-class hardware, a large percentage are normal, everyday desktops that quietly use excess cycles to work on different problems. Those wishing to employ the power of BOINC simply need to code to its specification, launch a BOINC project and recruit users (i.e., nodes). BOINC will even help with this last step.
But, is this any way to build a supercomputer?
Supercomputers Using Unconventional Hardware
On January 1, one of the BOINC projects for which I volunteer cycles, Einstein@Home, reached a milestone. This project, which is searching for spinning neutron stars (pulsars) with the hope of eventually proving some of Albert Einstein’s gravitational wave theories, now has a BOINC network with more than one petaflop of compute power. Theoretically, that puts it in 24th place in last November’s Top500 list. That’s not too shabby for a supercomputer made up of a hodgepodge of desktops, servers and the occasional gaming system.
But, what happens to this form of technical computing as desktops go the way of the dodo?
HPC Moves to Mobile
Already, projects like Boincoid and AndroidBoinc are coming online to move it mobile. Issues surrounding battery life and user experience impact are being addressed. Maybe these distributed supercomputers will simply and smoothly evolve into this next age of computing, and my phone will someday find the cure to cancer.
Certainly, there will always be a need for the large, dedicated HPC clusters. Certain workload types require the low latency and predictability such supercomputing centers provide. For everything else, maybe cloud is the answer, and HPC cloud computing already exists in the form of distributed, non-dedicated networks like BOINC.
At this point, many of these projects are originating from the universities. As many of the projects aim to only conduct research, there may be little incentive for corporations to get involved. However, there are us, those who live and breathe HPC. Let’s get involved and provide our expertise, even if it is simply becoming a volunteer node in a global supercomputing project.
I am an HPC node. I am on the Top500.