Search for: All records

Open OnDemand (openondemand.org) is an NSF-funded open-source HPC platform currently in use at over 200 HPC centers around the world. It is an intuitive, innovative, and interactive interface to remote computing resources. Open OnDemand (OOD) helps computational researchers and students efficiently utilize remote computing resources by making them easy to access from any device. It helps computer center staff support a wide range of clients by simplifying the user interface and experience. 

The need for distance and virtual learning platforms has been emphasized by the COVID-19 pandemic. With the closure of campuses in Spring of 2020 and many classes moving to online only in Fall of 2020, platforms for facilitating computationally oriented curriculum have had to be quickly adopted. Open OnDemand offers a familiar web-based portal to computational resources such as high-performance computing and cloud. Through OnDemands customizable dashboard, students can be offered an interface tailored to the course schedule giving them a just what I need view. Advantages to instructors include a web accessible, platform agnostic interface leading to less time for troubleshooting local student platforms and more time for discussion of the core course curriculum, a fully customizable course page, access controls, and more. Here we present Open OnDemand as a platform for developing, deploying, and presenting software and course material to software-oriented classes as used at Ohio Supercomputer Center and Virginia Tech. 

Abstract The Precipitation Occurrence Sensor System (POSS) is a small X-band Doppler radar that measures the Doppler velocity spectra from precipitation falling in a small volume near the sensor. The sensor records a 2D frequency of occurrence matrix of the velocity and power at the mode of each spectrum measured over 1 min. The centroid of the distribution of these modes, along with other spectral parameters, defines a data vector input to a multiple discriminant analysis (MDA) for classification of the precipitation type. This requires the a priori determination of a training set for different types, particle size distributions (PSDs), and wind speed conditions. A software model combines POSS system parameters, a particle scattering cross section, and terminal velocity models, to simulate the real-time Doppler signal measured by the system for different PSDs and wind speeds. This is processed in the same manner as the system hardware to produce bootstrap samples of the modal centroid distributions for the MDA training set. MDA results are compared to images from the Multi-Angle Snowflake Camera (MASC) at the MASCRAD site near Easton, Colorado, and to the CSU–CHILL X-band radar observations from Greeley, Colorado. In the four case studies presented, POSS successfully identified precipitation transitions through a range of types (rain, graupel, rimed dendrites, aggregates, unrimed dendrites). Also two separate events of hail were reported and confirmed by the images. 

High performance computing workloads often benefit from human in the loop interactions. Steps in complex pipelines ranging from quality control to parameter adjustments are critical to the successful and efficient completion of modern problems. We give several example workflows in bioinformatics and deep learning where computing decisions are made throughout the processing pipelines ultimately changing the course of the compute. We also show how users can interact with the pipeline using Open OnDemand plus XDMoD or Plot.ly. 

Summary High performance computing (HPC) has led to remarkable advances in science and engineering and has become an indispensable tool for research. Unfortunately, HPC use and adoption by many researchers is often hindered by the complex way these resources are accessed. Indeed, while the web has become the dominant access mechanism for remote computing services in virtually every computing area, HPC is a notable exception. Open OnDemand is an open source project negating this trend by providing web‐based access to HPC resources (https://openondemand.org). This article describes the challenges to adoption and other lessons learned over the 3‐year project that may be relevant to other science gateway projects. We end with a description of future plans the project team has during the Open OnDemand 2.0 project including specific developments in machine learning and GPU monitoring.