University of California, San Diego; San Diego Supercomputer Center, United States of America
XSEDE’s Extended Collaborative Support Service (ECSS) helps researchers and students at U.S. institutions to make effective use of NSF-funded cyberinfrastructure. Activities include performance tuning and optimization, advanced visualization, data analytics, software parallelization, development of science gateways and workflows, I/O optimization, efficient use of accelerators and outreach to communities that have not traditionally used HPC. ECSS staff also play a key role in the creation and delivery of training materials in collaboration with XSEDE’s workforce development team and the mentoring of Fellows from the Campus Champion program. One of the biggest challenges facing ECSS is that is it needs to support a diverse collection of architectures (x86, ARM, GPU, novel AI processors), computing paradigms (batch, high throughput, notebooks, workflow, cloud) and scientific domains (physical, biological and social sciences). Being able to meet these demands requires that ECSS continues to train its own workforce, especially in new and emerging technologies. In this talk, we describe our best practices for delivering training to the XSEDE community. We also discuss our strategy for internal training to ensure that XSEDE staff are kept up to date in the latest hardware, programming languages, tools, libraries, compilers, machine learning techniques and computing frameworks so that they fulfil their core mission.