Issue link: http://osercommunicationsgroup.uberflip.com/i/412890
S u p e r C o m p u te r S h o w D a i l y Th u r s d a y, N o ve m b e r 2 0 , 2 0 1 4 6 FINALISTS COMPETE FOR ACM GORDON BELL PRIZE With five technical papers contending for one of the highest honored awards in high performance computing (HPC), the Association for Computing Machinery's (ACM) awards committee will officially release the announcement of the Gordon Bell Prize winner at the 26th annual Supercomputing Conference (SC) awards ceremony this November in New Orleans. Financial support of the $10,000 award is provided by Gordon Bell, a pio- neer in high-performance and parallel computing. The Gordon Bell Prize is an annual award given to an individual or team who has demonstrated an outstanding achieve- ment in one of three areas: peak perform- ance, scalability and time-to-solution, or a special achievement. Solving an impor- tant scientific or engineering problem in HPC is important to demonstrate and jus- tify, but scientific outcomes alone are not sufficient for this prize. This year's finalists are: "Real-time Scalable Cortical Computing at 46 Giga- Synaptic OPS/Watt with ~100× Speedup in Time-to-Solution and ~100,000× Reduction in Energy-to-Solution," with research led by Dharmendra S. Modha, IBM Fellow and IBM Chief Scientist – Brain-inspired Computing, and team including members from IBM and Cornell Tech; "24.77 Pflops on a Gravitational Tree-Code to Simulate the Milky Way Galaxy with 18600 GPUs," with their research led by Simon Portegies Zwart and Jeroen Bédorf of the Netherland's Leiden Observatory and team; "Anton 2: Raising the Bar for Performance and Programmability in a Special-Purpose Molecular Dynamics Supercomputer," with lead researcher David E. Shaw, of DE Shaw Research, and team; "Petascale High Order Dynamic Rupture Earthquake Simulations on Heterogeneous Supercomputers," a collaborative research project co-led by Michael Bader (TUM, Germany), Christian Pelties (LMU, Germany) and Alexander Heinecke (Intel, United States); and "Physics-based urban earthquake simula- tion enhanced by 10.7 BlnDOF x30 K time-step unstructured FE non-linear seismic wave simulation," with research led by the University of Tokyo's Tsuyoshi Ichimura. The authors of "Real-time Scalable Cortical Computing at 46 Giga-Synaptic OPS/Watt with ~100× Speedup in Time- to-Solution and ~100,000× Reduction in Energy-to-Solution" developed a paral- lel, event-driven kernel for neurosynaptic computation, called TrueNorth, that tar- gets a broad range of cognitive applica- tions. This kernel is highly efficient with respect to computation, memory, and communication. This entry also demon- strates TrueNorth as a co-designed sili- con expression of the kernel. "24.77 Pflops on a Gravitational Tree-Code to Simulate the Milky Way Galaxy with 18600 GPUs," looks at the long-term evolution of the Milky Way Galaxy, which is simulated in this sub- mission using 51 billion particles. The simulation achieves impressive perform- ance on the Swiss Piz Daint supercom- puter using the N-body gravitational tree- code Bonsai. The paper, "Anton 2: Raising the Bar for Performance and Programmability in a Special-Purpose Molecular Dynamics Supercomputer," introduces a second-generation special- purpose supercomputer for molecular dynamics simulations, Anton 2. The authors' results indicate significant gains in performance, programmability, and capacity compared to its predecessor, Anton 1, with simulations running up to 180 times faster than on any general-pur- pose hardware. In "Petascale High Order Dynamic Rupture Earthquake Simulations on Heterogeneous Supercomputers," the authors present an end-to-end optimiza- tion of the Arbitrary high-order DERivative Discontinuous Galerkin (ADER-DG) software SeisSol. The opti- mizations target the Intel ® Xeon Phi coprocessor platforms, achieving impressive earthquake model complexi- ty of complex seismic wave propagation phenomena. And finally, the science in the paper, "Physics-based urban earthquake simula- tion enhanced by 10.7 BlnDOF x 30 K time-step unstructured FE non-linear seismic wave simulation," works to improve the reliability of urban earth- quake response analyses, using an unstructured 3D finite-element-based wave amplification simulation code, GAMERA running on the K computer. One of these papers will be announced the winner of the 2014 Gordon Bell Prize with the runner-up receiving Honorable Mention. The Gordon Bell awards com- mittee is selected by ACM and comprised of past Gordon Bell winners, as well as leaders in the field. THE GOLDEN AGE OF BSD UNIX By Michael Dexter, CTO, Gainframe For many of us, BSD Unix was our first Open System, but not every user is aware of how pervasive it has quietly become. Virtually every non-Microsoft operating system includes a BSD technology such as OpenSSH or a BSD libc, and you probably didn't know that Netflix is using FreeBSD to transport around a third of all U.S. Internet traffic each and every evening. Your Android or Apple phone relies on BSD Unix, and certainly every HPC system contains a BSD Unix component from its network infrastruc- ture though its management. These are all "stealth" success stories for BSD Unix, but it's time to consider the BSD family of operating systems for your HPC application. "Berkeley Software Distribution," or simply "BSD" Unix, began life as a DARPA-funded quest for a unified oper- ating system for government contractors. University of California, Berkeley researchers gradually replaced AT&T UNIX™ with a feature-rich operating system that was portable, open source and eventually included revolutionary features such as that TCP/IP networking protocol that gave us the Internet. With a strong legacy of networking, BSD Unix made its way into nearly every network router on the market, but by fate was beaten to the personal computer by GNU/Linux. Linus Torvalds himself said that "If 386BSD had been available when I started on Linux, Linux would probably never had happened." That quirk of his- tory has in no way diminished BSD Unix's capabilities, but only its populari- ty. Modern BSD Unix is performant, secure and as portable as ever. Primarily represented by the FreeBSD, OpenBSD and NetBSD operating systems modern BSD Unix has compelling features for diverse applications from embedded through HPC. FreeBSD recently celebrat- ed its 20th anniversary, and is broadly the most feature-rich BSD. Advanced features such as the bhyve hypervisor, on dis- play at SC14, and the OpenZFS filesystem make FreeBSD a solid platform for bare metal and virtual- ized applications of all sizes. bhyve breaks with tradition by only supporting the hardware virtualization assistance found in modern CPUs and VirtIO virtu- alized network and block storage devices. As a result, bhyve can run OpenBSD, NetBSD and GNU/Linux vir- tual machines at hardware-native speeds using only a small kernel module and support utilities. OpenBSD has long focued on proac- tive security, comprehensive networking, quality documentation and overall Unix correctness. The OpenBSD team has given us the ubiquitous OpenSSH secure shell and is developing LibreSSL to mit- igate the pervasive flaws found in the ubiquitous OpenSSL code base. They also produce the PF packet filter, OpenBGPD, OpenNTPD, OpenSMTPD and mandoc projects which collectively make for a very capable "batteries included" operating system with a consistent user experience. Finally, NetBSD has taken the BSD tradition of portability to new levels, and established itself as a solid platform for embedded and research systems with innovative features such as userspace kernels and devices. Collectively, modern BSD Unix rep- resents a comprehensive platform for building solutions both large and small with exemplary networking, security and virtualization features. Visit booth #3461 to learn more about mod- ern BSD Unix. For more information, go to www.bhyve.org or wiki.freebsd.org/bhyve.