December 20, 2010
Zircon Adaptive Software delivers 700 times speed-up and a near-linear scalability of two financial applications on SGI Altix UV 1000
WAYNE, NJ, Dec. 20 -- Zircon Computing today announced the results of two recently completed benchmarks in which Zircon collaborated with SGI on a key initiative to optimize the performance of two applications on the SGI Altix UV 1000 platform comprising 512 cores, 1,024 threads, and 4 terabytes (TB) of memory.
"As part of Zircon's latest software release version 3.3, Zircon delivered enhanced support for task and data parallelism, thereby reducing the programming effort needed to parallelize application business logic," said Alexander Mintz, CEO of Zircon Computing, LLC. He added, "we also transformed the SGI platform into a powerful environment for delivering high performance elastic data analytics applications that can be distributed easily and with more flexibility on the multi-core Altix UV 1000 system."
To scale up a massive calibration exercise of a path-dependent financial trading strategy that does not yield itself to simple parallelization, the trading strategy aspect of the application was split into a set of dependent sub-tasks and run through a pipeline pattern. The SGI Altix UV 1000 platform fits uniquely into this domain of historical evaluations by providing access to a large pool of shared memory resources and offering 512 cores, 1,024 threads all in one box, to process these subtasks. The net result is the ability of Zircon Adaptive software to execute thousands of strategies concurrently on the available pool of threads resulting in a 712 times speedup on SGI Altix UV 1000. The other financial benchmark conducted on an SGI Altix UV 1000 system was an extensive and complex Heston model option pricing calibration with an optimization routine implemented via the NAG minimization function. This classical case of task parallelism allowed massive concurrency of thousands of individual calibrations producing near-linear scalability and showing more than 600 times speedup on a hyperthreaded SGI Altix UV 1000 platform.
"These benchmark results demonstrate how the Zircon software can accelerate complex data analytics applications in a near-linear scalable manner," said Christian Tanasescu, vice president of software engineering at SGI. "Acceleration benefits for data analytics applications using Zircon software are enabled by the NUMA architecture of SGI Altix UV 1000, with up to 2,048 processor cores and 16TB of memory running a single image of the operating system, resulting in an extreme scale-up capability for numerous data-intensive applications."
"Zircon's achievement of these record-breaking results stands to offer significant benefit to financial services companies for whom milliseconds matter," said Rajeeb Hazra, general manager of high performance computing at Intel. "By combining the power of Intel Xeon processors 7500 series with the Altix UV 1000, Intel and SGI can support Zircon in continuing to optimize applications with high degrees of task and data parallelism."
The SGI Altix UV 1000 system used for this benchmark was powered by 64 Intel Xeon 7560 processors, 4 TB of memory and running SLES 11 SP1. The Heston calibration application ran 34,400 financial models in 3 minutes, which is over 500 times faster than a serial run taking 1,464 minutes on a 512-core SGI Altix UV 1000 system, showing a near-linear scalability. With hyperthreading, the improvement was more than 40 percent higher, yielding over 666 times faster calibration time with 1,024 threads. Similarly, the trading strategies analysis application showed a 479 times speedup on a 512-core SGI Altix UV 1000 system compared to a serial run and a 712 times speedup using all 1,024 threads.
About Zircon Computing
Zircon Computing, LLC develops and markets the Zircon Software Product Suite, providing ultra high performance grid and cloud computing solutions for organizations of all sizes and in many industries. The company delivers its easy-to-use, platform-independent solutions directly to enterprise clients as well as through an international network of certified partners. Founded in 2005, Zircon Computing is based in Wayne, NJ, and has operations worldwide. For more information, visit www.zircomp.com.
Source: Zircon Computing, LLC
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?