HPC facilities face the challenge of a diverse user base with different skill levels and needs. Whether targeted at developing applications with good performance or simply using a precompiled application, training at these facilities must incorporate a variety of topics at different levels to cater to a range of skill sets.
As the day-to-day computing experience of both the public at large, and STEM students in particular, moves farther and farther from that of most HPC systems – tapping on handheld devices instead of typing at the Unix command line – and as many non-CS STEM disciplines de-emphasize programming as a key component of their students' education, HPC facilities increasingly need to provide training, because much of the knowledge necessary for effective use of high-performance computing resources must be learned outside the traditional classroom.
HPC centers are motivated to provide this training to best serve the needs of their users, and often they are faced with the stark choice of either training users to be productive on systems that often are hugely different from what the users are accustomed to, or investing enormous sums in making hundreds of community codes straightforward to use via familiar interfaces, on a constantly shifting landscape of rapidly evolving domain applications and interface preferences. With limited time and resources available, HPC centers choose to train users, creating a foundation that can be built upon even as technologies change. But as HPC architectures grow in complexity, training gains even more importance.
Training benefits both users and HPC facilities. For novice users, training can help them find out what resources are available at the HPC center, learn what they need to do to use the resources, and empower them to use the resources. Advanced users also benefit from training to develop their skills with advanced development and debugging tools, and to build better applications that use the facility’s resources efficiently. Training helps all users to assess their current workflow and entertain the adoption of new resources, and to put a human face on the HPC experience.
From the HPC facility perspective, training can lead to a reduction in workload on firstline support staff by creating a better informed user base; lead to more efficient use of facility resources; and build a larger user community. In-person training creates a closer interaction between facility staff and users, enabling staff to gain a better understanding of the needs of the user base, which is valuable not only for further interactions but also in establishing center policies. Yet due to the geographically spread user base, centers must also create convenient accessible training such as documentation examples, self-paced tutorials and live/recorded webinars, for those that cannot attend live training.
Because of the broad range of experiences amongst the participants of this workshop, they bring unique perspectives on the challenges and opportunities that HPC training offers. The intent of bringing this unique body of knowledge together is to foster information sharing and to facilitate a deeper understanding of effective strategies for providing training to the HPC and broader STEM communities.
In this workshop, we seek to answer the question of how HPC centers can best leverage each center’s limited resources available for training activities. By using training best practices, they can maximize the impact of training offerings. And by developing collaborations with other facilities, they can prevent duplication of effort and develop training materials with expert contributions. This workshop will facilitate both of these goals and develop a foundation for gatherings at future Supercomputing Conferences.