Tuesday, November 13

Galaxy: Asynchronous Ray Tracing for Large High-Fidelity Visualization
Presenters: Paul Navratil, Greg Abram, Texas Advanced Computing Center
Length: 30 minutes
Time: 10:30am-11:00am

In this talk, a hands-on demonstration of Galaxy, a fully asynchronous distributed parallel rendering engine for full global illumination for large-scale visualization built using the Intel Rendering Framework, will be provided. Galaxy provides performant distributed rendering of complex lighting and material models, particularly those that require ray traversal across nodes. Our design favors tightly-coupled in situ scenarios, where data remains on simulation nodes. We will demonstrate how to generate Cinema image analysis databases using Galaxy on TACC's Stampede2 supercomputer, experience which will directly apply to the upcoming TACC Frontera leadership system as well.

High-Fidelity Rendering for Large-Scale Tiled Displays
Presenters: Joao Barbosa, Paul Navratil, Texas Advanced Computing Center
Length: 30 minutes
Time: 11:30am-12:00pm

In the age of big data, increasing simulation accuracy, we often need more pixels to show the minute details that traditional single display system cannot provide. Industrial designers such as car manufacturing design teams rely on high-fidelity photorealistic rendering, as supplied by the Intel OSPray framework, to study the interaction of materials and shapes in their designs. The Display Wall module built on top of Intel OSPRay framework addresses these problems by providing an interactive tool that exploits the power of Intel-based clusters like Stampede2 to achieve interactive high-fidelity model and material rendering with contextual surroundings. In this presentation, we will show the challenges of building and implementing the module, the use cases scenarios, and the current results.

Color Strategies for Large-Scale Volume Rendering
Presenter: Francesca Samsel, Center for Agile Technologies, UT Austin
Length: 30 minutes
Time: 1:30pm-2:00pm

HPC has enabled scientific domains to study increasly large data in detail. Color, specifcally color contrast, is the medium through which we interpret the data. Here we present a workflow, strategies and tools for increasing the scope and detail revealed within one's visualization by employing expanding contrast ranges, allocating them to areas of high interest and aligning hue intensity with data importance. This methodology enables scientists to explore and communicate the needles in the haystacks of data via targeted use of color contrast theory. We also present techniques for expanding the luminance ranges and allocating the contrast to the areas of greatest interest.

Lmod: What's New with TACC's Environment Module System
Presenter: Robert McLay, Texas Advanced Computing Center
Length: 30 minutes
Time: 3:00pm-3:30pm

This presentation will give a quick overview on the Environment Module system called Lmod. It will discuss the features of Lmod compared to other module systems. Finally, it will cover the new ways that Lmod allows for hiding modules without removing them as well as better ways to handle dependent modules and other new features.

XALT: How to Track and Understand What Your Users are Running on Your System
Presenter: Robert McLay, Texas Advanced Computing Center
Length: 30 minutes
Time: 3:30pm-4:00pm

XALT is an extremely lightweight solution to track what your users do. XALT tracks both MPI and non-MPI programs. Come see how easy it is to better understand what your users are doing. Let's learn how to drink from the firehose of job data that your site is generating.

Wednesday, November 14

Frontera: Computing for the Endless Frontier
Presenter: Dan Stanzione, Texas Advanced Computing Center
Length: 30 minutes
Time: 10:30am-11:00am

The Texas Advanced Computing Center (TACC) will be deploying a new system, Frontera, in 2019, with support from the U.S. National Science Foundation as the first phase of the new NSF "Towards a Leadership Class Computing Facility" program. This talk will provide a brief overview of the plans for the new system and the broader project that surrounds it, review the architectural design choices made, and discuss the application community that will run on it. The talk will also provide an update on the current landscape of systems at TACC, how they are used, and how usage is changing.

Exploring Technologies for Data Intensive Computing
Presenter: Chris Jordan, Texas Advanced Computing Center
Length: 30 minutes
Time: 11:30am-12:00pm

TACC has been a leader in data intensive computing across multiple platforms, but particularly through support for large-scale shared collections on Corral, and through high-performance storage subsystems on Wrangler. We will discuss process and specific examples of testing and planning scenarios for investigating new technologies in shared and high-performance storage to support data intensive computing, including configurations and benchmarking topics. We present the concepts of technology investigation with a focus on application requirements based on experiences with Wrangler and other systems, the challenges and benefits of utilizing real-world applications examples in an experimental context, and a brief overview of the current state of standards and markets in NVMe and related high-throughput/low-latency storage systems.

Containers@TACC: Portable, Reproducible Scientific Applications
Presenters: John Fonner, Greg Zynda, Texas Advanced Computing Center
Length: 30 minutes
Time: 2:00pm-2:30pm

As computing becomes increasingly important across scientific domains, the ability to compute anywhere with consistent results is a critical capability. We present TACC's approach to adopting software containers for both centralized and user defined software support across our HPC and cloud ecosystem. We discuss the lessons learned over the past year and how we envision the future of containerized software at TACC.

Machine Learning at TACC: Support and Examples
Presenter: Weijia Xu, Texas Advanced Computing Center
Length: 30 minutes
Time: 3:30pm-4:00pm

New methods and tools in machine learning and deep learning have fueled increasing adoption of data driven discoveries across many scientific domains in recent years. Domain scientists are embracing those methods as both standalone data science methods, as well as effective approaches to reducing dimensionality in traditional simulation. However, data driven analysis commonly has high computational requirements and is bound by the availability of computational resources. Supporting these new methods within high performance computing environment is a pressing need. In this talk, we will discuss how TACC supports machine learning and deep learning, and present selected projects from TACC.