Computing power to answer NASA's complex science and engineering questions
Past Webinars Available for Playback
After a webinar has taken place, the recorded presentation and the slide deck (PDF) are available below.
Introduction to the New Athena Cluster
This webinar introduces you to NAS' new CPU cluster named Athena which is scheduled to be released to production soon. Athena is composed of 1,024 AMD Turin nodes interconnected with HPE/Cray Slingshot 11 in a Dragonfly topology. We provide a brief illustration of this interconnect and describe the use of the Cray Programming Environment for porting applications; and share some experiences from NAS staff and Early User testing. Athena Webinar Slide Deck (PDF-1.3MB)
Introduction to Cabeus, Part 1: Overview
This session provides information on accessing Cabeus; describes its hardware resources; defines a virtual node and illustrates its effect on PBS jobs sharing; recommends a proper way to use some of the host memory as a temporary filesystem for I/O; and helps you to understand and check how each job is charged against your SBU allocation. Cabeus Webinar Part 1 Slide Deck (PDF-1.7MB)
Introduction to Cabeus, Part 2: Building Applications
This session gives a high-level view of a few programming methods recommended by Nvidia, namely CUDA, OpenACC, OpenMP and standard languages; weaves these programming methods with two important concepts including CPU offloading work to GPU and explicit vs. implicit data movement between CPU and GPU; and highlights the roles of compute capability, CUDA toolkit and CUDA driver for building your application. Cabeus Webinar Part 2 Slide Deck (PDF-1.3MB)
TOSS4 Migration: A Retrospective and Current Status – June 27, 2023
This webinar provides a brief summary of issues seen in the recent TOSS3 to TOSS4 migration from the users' perspectives and the resolution to those issues. A Q&A session followed the main presentation. Recorded Presentation Slides (PDF-1MB)
How to Use the Cloud at NAS – June 27, 2023
This pre-recorded webinar covers how to use the Amazon Web Services (AWS) through the HECC Cloud service, including how to get an account, where to find documentation, and how to access and navigate the HECC Cloud environment. A topic index with timestamps can be found in the PDF below. Recorded Presentation Slides (PDF-1MB)
Webinar/Virtual Office Hour on TOSS 3 Migration – January 11, 2022
This webinar + virtual office hour will provide information about best practices for migrating to TOSS 3 and provide an opportunity for you to voice your concerns and discuss any issues with the migration. Recorded Presentation Slides (PDF-1MB)
2021 NAS Data Science Webinar – September 15, 2021
This webinar provides an overview of Machine Learning software stacks provided by NAS and some of our latest tools and utilities available to you. This is also an excellent opportunity to meet the Data Science team members and learn how to get started with machine learning at NAS. Recorded Presentation Slides (PDF-4MB)
HECC Data Science Platform – June 3, 2020
This webinar provides a brief, high-level overview of the data science conda environments provided for use on HECC systems, with a focus on the TensorFlow 2 environment. Environments provided for other versions of TensorFlow and PyTorch will be mentioned, but not discussed in detail. Other topics covered include setting up Jupyter Lab and using TensorFlow with multiple GPUs, along with a brief discussion of use cases. The goal of this webinar is to help you understand the software environments we provide to assist with your AI/Machine Learning and Data Science work on HECC systems. Recorded Presentation Slides (PDF-6.4MB)
Overview of HECC Pay-for-Use AWS Cloud – August 7, 2019
After a year of development, HECC is now capable of and ready to provide Cloud as an alternative to on-premises resources for NASA HECC projects on a pay-for-use basis. This webinar will go over the reasons behind offering AWS Cloud and how PIs can work with HECC to create an environment tailored for each project's need and budget. A high level description will be provided on how AWS is integrated into the HECC environment with an implementation that is cost conscious, flexible and familiar for HECC users. This initial offering is valid only for using AWS Public Cloud. AWS Gov Cloud will be included in the near future. Recorded Presentation Slides (PDF-1.7MB)
Compiler Directive based GPU Programming in C and Fortran – April 25, 2018
The purpose of this webinar is to provide a brief, high-level introduction to GPU accelerated computing using compiler directives. In the past, programmers had to rely on language extensions provided by CUDA. Programming models such as OpenACC and OpenMP aim to ease the task by providing support via compiler directives. In this webinar we mostly discuss the OpenACC API. We explain basic OpenACC concepts and constructs, and show how to compile and run accelerated codes on Pleiades GPU nodes. The target audience for this presentation are users with very little or no experience in GPU programming, who would like to explore a quick path to getting started. Understanding of parallel programming is assumed. Basic knowledge of OpenMP might be helpful. Recorded Presentation Slides (PDF-4.9MB)
Memory Usage on NAS Xeon-Based Systems – November 8, 2017
In this training session, we review multiple aspects of memory usage for applications running on NAS Xeon-based cluster systems. We distinguish between phsyical memory and virtual memory and talk about what consumes memory. We then outline a few scenarios when the physical memory is not enough and what tools are available for monitoring memory usage. In particular, a new user-friendly tool called vnuma which allows you to collect and visualize memory usage is introduced. Recorded Presentation Slides (PDF-1.6MB)
Using IOT for I/O and MPI Performance Analysis – September 6, 2017
I/O and MPI communication are two important aspects of high-performance computing applications. This webinar provides an introduction to IOT, a licensed toolkit available to all NAS users, for I/O and MPI performance analysis. In addition to basic usage instructions on Pleiades, a few examples utilizing production codes provided by NAS users will be used to demonstrate some of the capabilities of IOT and how it can be useful in multiple scenarios. Recorded Presentation Slides (PDF-4.7MB)
Using OpenMP 4.0 on Pleiades – July 26, 2017
OpenMP is a standard API to support shared-memory multiprocessing programming that is based on compiler directives. In this training session we briefly review some OpenMP programming basics and explain how to compile and run OpenMP and hybrid MPI+OpenMP codes on Pleiades. We then discuss OpenMP SIMD vectorization support, which was introduced in OpenMP 4.0. The training concludes by addressing general issues such as performance considerations and common problems. Recorded Presentation Slides (PDF-1.3MB)
Performance Analysis Using Open|SpeedShop – April 27, 2017
Open|SpeedShop is a collection of open source tools maintained by Krell Institute and installed on NASA's Pleaides and KNL platforms. Jim Galarowicz and Greg Schultz present and demonstrate the latest features of the Open|SpeedShop application performance tool. The talk covered the general use and capabilities of Open|SpeedShop, focusing on its usage on the NASA Pleiades platform. Slides (PDF-4.9MB) This webinar was not recorded.
Modules Under SLES12 and Using Software in pkgsrc – April 12, 2017
This webinar provides information that (a) helps you with SLES 12 migration, and (b) describes pkgsrc in general including what packages are included and useful commands for getting details about different packages. Recorded Presentation Slides (PDF-1.2MB)
Getting a NAS Account – October 19, 2016
This webinar covers the steps required for Principal Investigators (PI) and new users in getting an NASA Advanced Supercomputing account to use HECC resources. An overview of NASA’s identity process for remote IT access will be discussed. In addition, we will touch on NASA’s IT Security Training policy and overall account maintenance required to retain supercomputing access. Recorded Presentation Slides (PDF-14.6MB)
Using SciCon Tools to Help with Productivity – August 24, 2016
There are multiple tools created by the SciCon members over the years to help users with various tasks. An overview of these tools will be provided in this webinar where users can learn how to check available licenses, status of PBS resources, status of user's PBS jobs, amount of memory used in a job, debug a running job, pin processes for better performance, and analyze performance for MPI applications. Slides (PDF-1.6MB)
Using MPIprof for Performance Analysis – July 20, 2016
Understanding performance characteristics of applications on high performance computing (HPC) systems is important for reducing compute time and improving HPC resource utilization. There are software tools available for analyzing performance of serial and parallel programs. This webinar focuses on MPIProf—a profiling tool for analyzing MPI applications. The tool has a simple and common interface to work with many MPI implementations, such as SGI MPT, Intel MPI, and OpenMPI. We will give a basic overview of the tool in gathering and reporting statistics about communication time, message size, I/O calls, and memory usage in applications. We will explain the collected results using Overflow, a computational fluid dynamic (CFD) code, as an example. Other capabilities, such as user-defined profile using MPIProf, will be discussed. Finally we will touch upon the accuracy of MPIProf results in comparison with other methods. Slides (PDF-4.3MB)
Simplifying and Optimizing Your Data Transfers – June 15, 2016
In this webinar, we discuss the transfer of data both within the NAS/HECC environment and between NAS and external sites. A variety of frequently recurring transfer tasks are covered, along with how they can be easily achieved using the Shift automated transfer tool. In addition, we present the factors influencing transfer performance, together with how you can assist Shift in maximizing performance. Finally, we discuss some common usage mistakes—and how to avoid them. Recorded Presentation Slides (PDF-1.6MB)
Performance Evaluation of Pleiades Broadwell Nodes Using NASA Applications – May 5, 2016
The Pleiades supercomputer now contains five generations of Intel processors: Broadwell, Haswell, Ivy Bridge, Sandy Bridge, and Westmere. This webinar reviews the architectural evolution of these processors and highlights the salient trending features such as decreasing core frequency, increasing memory latency, decreasing memory bandwidth, increasing number of cores and threads, wider Single Instruction Multiple Data (SIMD) registers, and Turbo Boost. We examine the impact of these features on NASA applications, and present a model to explain why it is hard to get double-digit floating-point efficiency even on BLAS 1 routines like AXPY and DOT. Note that this webinar was not recorded. Slides (PDF-4.4MB)
Effective Use of the Lou Mass Storage Cluster – December 11, 2013
This webinar covers the use of the Lou mass storage system, in which Lou1 was recently integrated. Standard best practices and data management techniques are reviewed. You will learn how to get high transfer rates between Lou and Pleiades using the Shift file transfer tool, as well as some lesser-known options to speed up the rsync copying tool. Some methods and tools to check the integrity of your data transfers, and instruction on the use of PBS-scheduled nodes for post-analysis of data on Lou and Pleiades are also provided. Recorded Presentation Slides (PDF-963KB)
Using the TotalView Debugger and MemoryScape – August 21, 2013
This webinar provides an overview on how to launch and use TotalView, a GUI-based debugging tool that gives users control over processes and thread execution, with a focus on TotalView's integrated memory debugger, MemoryScape. You will learn how to find memory leaks, identify heap memory buffer over- and under-runs, detect dangling pointers, and more. Recorded Presentation Slides (PDF-17.80MB)
PBS Queues and Scheduling on Pleiades – June 11, 2013
This webinar provides an overview of the Portable Batch System (PBS) queue structure used at NAS to manage the computing workload on NAS systems, including a general description of how job scheduling works and a review of resources and job monitoring tools. It also provides a better understanding of how to choose the best queues for your work and will teach you how to get more details about a specific queue. In addition, you will get guidelines for requesting assistance under special circumstances that are beyond the typical queue offerings. Recorded Presentation Slides (PDF-1.16MB)
Advanced Features of the Shift Automated File Transfer Tool – April 10, 2013
This webinar covers newly added and existing advanced features of the Shift automated file transfer tool to help you get the most out of Shift and to reduce turnaround time of your file management tasks. A recap of basic features and usage is presented before delving into advanced functionality, including high-speed local/remote tar creation and extraction, rsync-like synchronization, parallelization of single and multi-file transfers across clients and/or hosts, and a revised retry capability, among others. Recorded Presentation Slides (PDF-1.3MB)
UnInit: Fix your code! Finding computation with uninitialized data – October 31, 2012
This webinar addresses a classic error made in many codes, where floating point computations are performed using data that has not been initialized. With modern compilers and the current crop of exotic methods by which data can be allocated on new systems, this common error can introduce bizarre and non-reproducible results in users' codes. HECC code experts have developed a methodology called "UnInit" that works in conjunction with the Intel compiler and some simple runtime libraries. Recorded Presentation Slides (PDF-1.25MB)
Basic Debugging – October 10, 2012
During the code development or porting process, debugging is often needed to pinpoint the cause of either wrong or different results. In this webinar, you will learn how to use a variety of debugging tools, including compiler flags. We introduce Totalview, an advanced debugging tool, to demonstrate ways catch some common programming bugs such as exceeding array bounds, floating point exceptions, uninitialized variables, and improper algorithms in serial or parallel programs. Instructions on how to use Totalview on Pleiades are provided. Recorded Presentation Slides (PDF-13MB)
Advanced OpenMP Topics – September 12, 2012
Second in a two-part series. This webinar focuses on more advanced OpenMP topics, including the newer tasking model, performance considerations, and future extensions of OpenMP. Participants are encouraged to view the recorded presentation and slides from the first webinar, "Shared-Memory Parallelism and OpenMP." Recorded presentation unavailable due to technical issues with audio track Slides (PDF-2.84MB)
Shared-Memory Parallelism and OpenMP – August 22, 2012
The first in a two-part series, this webinar gives an introduction to shared-memory parallelism, major components, and use of the OpenMP model. It also introduces the hybrid MPI + OpenMP model. Usage examples are provided throughout the materials. Recorded Presentation Slides (PDF-3MB)
Simple Automated File Transfers Using SUP and Shift – April 4, 2012
This webinar covers the use and features of the Secure Unattended Proxy (SUP) and a new automated file transfer tool called Shift. The SUP allows you try transfer files and perform other operations for up to a week without using your RSA SecurID fob. Shift supports automated local and/or remote file transfers with on-demand status and advanced capabilities such as data integrity validation and automatic transfer parallelization. Through the use of these tools, you will be able to simplify, automate, and improve the performance of your file transfers. Recorded Presentation Slides (PDF-1.6MB)
New User Orientation – November 9, 2011
This webinar gets new users started with NAS resources. It introduces the NAS systems, logging in, transferring files to/from NAS, setting up the module environment, compiling, and running jobs with PBS. The emphasis is on using the Search capability in the Knowledge Base and common issues faced by new users. Recorded Presentation Slides (PDF-1.8MB)
