Loading…
RMACC HPC Symposium 2015 has ended
Welome to the RMACC HPC Symposium. Login here to create your own personal schedule and add your name to the directory of attendees. Follow us on Twitter @CUBoulderRC for Conference and Schedule updates

Sign up or log in to bookmark your favorites and sync them to your phone or calendar.

Sponsor Track [clear filter]
Wednesday, August 12
 

10:15am

NVIDIA Technical Track : Application Readiness for the Pre-Exascale Phase

Computational efficiency with increased focus on performance-per-energy-cost has become the overarching driver behind architectural considerations for Exascale HPC.  The development of scientific and engineering application software that achieves full potential from such Exascale systems will require a level of hardware/software co-design that is a relatively new approach in

conventional HPC.  NVIDIA GPUs are a centerpiece of the first pre-Exascale systems announced in the USA under the U.S. Department of Energy CORAL partnership, and are a key component of co-design collaborations that will strive for accelerated application readiness by system deployment in 2017.

 

This presentation will first examine the motivation and progress of GPU-based heterogeneous system architectures for the pre-Exascale phase of HPC. The second topic will introduce the requirements for extraction of fine-grain parallelism of application software and current state of GPU-accelerated modeling and simulation applications with review of the programming strategies deployed. Select examples will provide relevance to science-scale HPC practice that quantifies the benefits of heterogeneous vs. CPU-only computing. In addition to GPU use with x86 CPUs starting from the mid-2000’s, the POWER and ARM-64 CPU architectures have become available alternatives

starting in 2014. The third and final topic will provide roadmaps of GPU hardware and system software, and interoperability with these new host CPU platforms.


Speakers

Wednesday August 12, 2015 10:15am - 10:45am
Wittemeyer Courtroom

10:15am

Using Globus to Manage Research Data from Inception through Publication
Limited Capacity seats available

Globus is software-as-a-service for research data management, used at dozens of institutions and national facilities for moving and sharing big data. Recent additions to Globus are services for data publication and discovery that enable: publication of large research data sets with appropriate policies for all types of institutions and researchers; the ability to publish data using your own storage or cloud storage that you manage, without third party publishers; extensible metadata that describe the specific attributes for your field of research; publication and curation workflows that can be easily tailored to meet institutional requirements; public and restricted collections that give you complete control over who may access your published data; a rich discovery model that allows others to search and use your published data. This presentation will give an overview and demonstration of these services, as well as case studies that illustrate how institutions are using Globus for data publication and discovery.


Speakers

Wednesday August 12, 2015 10:15am - 10:45am
Wolf 304

11:00am

Scalable systems management with xCAT/Confluent

Scalable systems management might appear to be a new problem that needs to be solved as cloud and hyperscale solutions become more mainstream.  However, scalable management has been a problem that has needed a solution in high performance computing for almost 20 years beginning with the emergence of Beowulf clusters.  Several open source and proprietary toolkits exists today with each differing in their ability to quickly and easily scale.   One such toolkit, the Extreme Cluster Administration Toolkit (xCAT: http://sourceforge.net/p/xcat/wiki/Main_Page/) first came on the systems management scene in the early 2000s with version 1.x and has since matured to it’s current 2.9.1 version. xCAT is a scalable cluster management and provisioning tool that provides hardware control, discovery, and OS diskful/diskfree deployment for all types of scalable systems beyond HPC.  xCAT is open source today, will continue to be in the future and is strategic to our scalable systems management strategy at Lenovo.  Come listen to this presentation to learn how the xCAT team is working to solve tomorrow’s system management problems with the emergence of the new Confluent module.  This module is poised to simplify extreme scalability and ease of use issue so users can minimize downtime and increase productivity of their systems.


Speakers

Wednesday August 12, 2015 11:00am - 11:30am
Wolf 304

2:45pm

Intel: Crossing the frontiers of performance and programmability
Limited Capacity seats available

System engineers have to contend with a number of challenges as they design platforms with Exascale level performance. This talk examines how questions of programmability and performance may play out through the transition to next generation architectures.

Speakers

Wednesday August 12, 2015 2:45pm - 3:15pm
Wolf 304

2:45pm

Towards Carbon-Free Data Centers
Limited Capacity seats available

Our ever-increasing demand for data and bandwidth has resulted in an unintended consequence, a substantial increase in global energy consumption with its associated growth in greenhouse gas emissions. The Natural Resources Defense Council*, states that in 2013, U.S. data centers consumed an estimated 91 billion kW-hr of electricity, which is equivalent to the annual output of 34 large (500 MW) coal-fired power plants. U.S. data center electricity consumption is projected to increase to roughly 140 billion kW-hr annually by 2020, the equivalent annual output of 50 power plants, emitting nearly 100 million metric tons of carbon pollution per year. This path is clearly unsustainable, so it is time to rethink data center power delivery/consumption, and to take a more holistic view of the computing enterprise. This talk will focus on how a fuel cell/compute rack building block can be used to create the roadmap for the sustainable and carbon-free data center.


Wednesday August 12, 2015 2:45pm - 3:15pm
Wittemeyer Courtroom

3:45pm

The Future of Interconnect is already here
Limited Capacity seats available

Mellanox InfiniBand technology is the foundation for scalable and performance demanding computing infrastructures.  Delivering more than 100Gb/s throughput, sub 700ns application to application latency and message rates of 150 million messages per second has already placed ConnectX-4 EDR 100Gb/s technology in the Top500 list of the world most powerful and efficient supercomputers.  We will discuss the latest interconnect advancements that maximize application performance and scalability, multi-host technology and introduce the next generation of GPUDirect RDMA Technology.  We will review the superior capabilities of Mellanox interconnect solutions that are available today, as well as into future generations.

Speakers
SS

Scot Schultz

Director, HPC and Technical Computing Mellanox


Wednesday August 12, 2015 3:45pm - 4:15pm
Wolf 304