Production high-performance computing (HPC) systems are adopting and integrating GPUs into their design to accommodate artificial intelligence (AI), machine learning, and data visualization workloads. To aid with the design and operations of new and existing GPU-based large-scale systems, we provide a detailed characterization of system operations, job characteristics, user behavior, and trends on a contemporary GPU-accelerated production HPC system. Our insights indicate that the pre-mature phases in modern AI workflow take up significant GPU hours while underutilizing GPUs, which opens up the opportunity for a multi-tier system. Finally, we provide various potential recommendations and areas for future investment for system architects, operators, and users.

This Innovative Practices Work in Progress describes how serious games can be used to help teach High Performance Computing (HPC). Serious games provide pathways for learners to develop intuition while learning new concepts. We define serious games as those that incorporate learning objectives, educational content, and assessment and whose purpose is learning rather than entertaining. Such games are valuable in an educational context because they engage students in active learning and help students develop mental models from their experiences. These student experiences generally lead to deeper questions, allowing instructors the chance to clarify misunderstandings, and reinforce learning. This work describes two games used in informal HPC courses to provide students with tangible hands-on experiences with HPC concepts.

This Work-In-Progress paper details the process and lessons learned when converting a hands-on engineering mini-course to a scalable, self-paced Massively Open Online Course (MOOC). Online courseware has been part of academic and industry training and learning for decades. Learning activities in online courses strive to mimic in-person delivery by including lectures, homework assignments, software exercises and exams. While these instructional activities provide “theory and practice” for many disciplines, engineering courses often require hands-on activities with physical tools, devices and equipment. To accommodate the need for this type of learning, MIT Lincoln Laboratory's “Build A Small Radar” (BSR) course was used to explore teaching and learning strategies that support the inclusion of physical construction and real world data collection in a MOOC. These tasks are encountered across a range of engineering disciplines and the methods illustrated here are easily generalized to the learning experiences in engineering and science disciplines.

This Work In Progress focuses on student perceptions of the effectiveness of three content delivery modes; a) traditional, residential in-class b) class capture for asynchronous online delivery, and c) modularized targeted content videos for online and blended or flipped classroom mode. Despite the growth of MOOCs and the concomitant shift from long lecture videos to learning modules, many online courses still rely on a class capture method. A recent study on video use by students in online courses recommends the use of short videos, six to nine minutes in length, based on student viewing habits. [1] The student behavior was inferred from click-stream data captured within the LMS but there was no direct interaction with students to gauge impressions of the content delivery or impact on learning. To improve both the residential and online experience requires a deeper understanding of those features which promote learning in each approach. Some of this understanding can be gleaned from traditional academic course evaluations, but student perceptions yield an additional level of understanding. For this study we analyzed data from multiple offerings of a 7-week graduate level numerical analysis course; Application of Finite Element Analysis. Over three years this course was offered residentially and online using multiple delivery modes. We categorize the delivery modes as Traditional, Class-Capture, and Modularized/Blended. Each student studied the material using one delivery mode and was then asked to review the material using another delivery mode. The students were asked to evaluate the perceived learning attained based on their original content delivery mode and the alternate mode. The feedback from this analysis, combined with traditional course evaluations will be used to develop a more comprehensive and objective survey to be used in the next round of courses.

Communication in multi- and many-core processors has long been a bottleneck to performance due to the high cost of long-distance electrical transmission. This difficulty has been partially remedied by architectural constructs such as caches and novel interconnect topologies, albeit at a steep cost in terms of complexity. Unfortunately, even these measures are rendered ineffective by certain kinds of communication, most notably scatter and gather operations that exhibit highly nonlocal data access patterns. Much work has gone into examining how the increased bandwidth density afforded by chip-scale silicon photonic interconnect technologies affects computing, but photonics have additional properties that can be leveraged to greatly accelerate performance and energy efficiency under such difficult loads. This paper describes a novel synchronized global photonic bus and system architecture called P-sync that uses photonics' distance independence to greatly improve performance on many important applications previously limited by electronic interconnect. The architecture is evaluated in the context of a non-local yet common application: the distributed Fast Fourier Transform. We show that it is possible to achieve high efficiency by tightly balancing computation and communication latency in P-sync and achieve upwards of a 6× performance increase on gather patterns, even when bandwidth is equalized.