TACC Presents Highlights from SC13
During SC13 in Denver, Colorado, the Texas Advanced Computing Center (TACC) has been presenting “live” commentary from the show floor with on-the-spot demonstrations and interviews with TACC representatives and their University of Texas partners. This third edition of TACC’s SC13 roundup provides an overview of the center’s Ultimate Science Environment for researchers and an introduction to the center’s two newest machines, Maverick and Wrangler.
Kicking off the third installment is HPC luminary and University of Texas at Austin Professor George Biros with a few words on TACC’s flagship supercomputer, Stampede. “Stampede has been a critical component in our research efforts to develop scalable algorithms,” notes Biros. “And we are working on fast algorithms for fluid simulation, medical image processing and geophysical applications with other groups across the country. It’s a great resource for us, and TACC staff has been very responsive. We are very lucky to have access to this national research center.”
The video also includes an overview of Wrangler, the innovative data analysis and management system scheduled to go live in January 2015 that TACC first announced in October. This 120 node Hadoop-ready Dell cluster is designed from the ground up to meet the needs of the data-intensive open science community. It’s distinction is a primary storage tier based on NAND Flash memory, which aims to deliver data at one terabyte per second and execute up to 275 million IOPS (input/output operations per second).
Dan Stanzione, TACC’s Deputy Director, states that “Wrangler will be the first purpose-built data-intensive supercomputer at TACC. We’ve always built very large high-performance computing systems, but we’ve really focused more on the simulation community. With Wrangler, we’re changing our direction to move more into the data-intensive applications and so as a result of that it’s configured to have a much higher performance IO system per node than our traditional systems. We’re doing some really innovative stuff with our partners in NAND Flash technology to deliver more than a terabyte per second of data and more than 250 million IO operations per second.”
TACC is also in the process of deploying its new high performance visualization and data analytics cluster, called Maverick. Developed in partnership with HP and NVIDIA, Maverick is headed for production status mid-January 2014. The open science and engineering community resource includes 132 HP ProLiant nodes, each of which is outfitted with dual ten-core Intel Xeon E5-2680 v2 processors and one NVIDIA Tesla K40 graphics chip. The five-rack cluster also sports 14 HP ProLiant management, login, and Lustre router servers.
In the video, Kelly Gaither, Director of Visualization at TACC, highlights the center’s three-pronged mission, which includes resources, services and user support, and a research and development arm that is critical for paving the way for solving tomorrow’s challenges. Maverick’s role will be to support that R&D effort.
Says Gaither: “Our hope is that we bring a new breed of system to the user community, where they can get on and they can feel like it’s an interactive resource that feels just like a laptop that gives all the capabilities with reduced barriers to entry.”