Dr Werner Scholz, Mr Jakub Szarlat, Mr Jon Tinberg1
1Xenon Systems, Springvale, Australia
Cloud computing providers have a range of offerings relevant to eResearchers, including Infrastructure-as-a-Service (IaaS), High Performance Computing as a Service (HPCaaS), Artificial Intelligence and Deep Learning as a service (AL/DLaaS) as well as data storage as-a-Service (STaaS).
Rather than purchasing compute, networking and storage resources, housing them and managing them at your facility – all of this is outsourced in the as-a-Service world.
This has the potential to free up personnel to focus on research and core tasks, rather than IT management. The opex monthly fee model can help match IT resources to grant cycles. The smaller configurations also allow for discrete IT resources to be allocated to smaller projects rather than competing in a shared pool.
However, eResearch IT projects are often the most demanding for compute and storage, requiring specialized High Performance Clusters, powerful GPUs for AI/DL, and large amounts of data storage. This is a much different set of requirements compared to the cloud providers’ usual enterprise customers. Configuring these resources in a cost effective and performant manner is key to successful as-a-service projects.
In addition to specialized IT requirements, eResearch must be mindful of a range of issues around research data, including: sovereignty, privacy, the longer term management and preservation of the data (and associated cost).
Join us for this informative session as Dr Werner Scholz and Jakub Szarlat from XENON explore the issues around HPC/DL/AI-as-a-Service and how the unique demands of eResearch fit in the world of as-a-Service.
Biography:
Werner has more than 20 years experience with high performance computing systems – from individual workstations and storage servers to massively parallel HPC clusters and large storage systems. He is also the developer of an open source finite element simulation package, which uses MPI, OpenMP, and GPU parallelization techniques. It is in use by academic and industrial research organizations around the world.
Werner has a PhD in physics from the Vienna University of Technology in Austria, where he specialized in computational physics and magnetic materials. He is the author of more than 80 journal articles in the area of computational physics and magnetic nanostructures and co-inventor of more than 25 patents related to magnetic storage technologies.