The High Performance Computing Special Interest Group for UK Academia
DaCe Background: Optimization: Tiling for complex cache hierachicies on multicore GPGPUs, task fusion. FPGA DaCe: a way to solve this? Map application to a work and data-flow graph Have a ‘performance engineer’ to rework the workflow to make it more efficient in terms of data usage, cache, etc. Creates a template for that problem to …
Penguin The element that particularly struck me here was ScyldWorkstation – a 4k virtual desktop, which might be another solution for visualisation, GUIs, and other workstation replacements and instant-access use cases. Univa The two elements here were support for instances in the cloud (see mention of CycleCloud in this context from earlier) and, in a …
Survey https://tinyurl.com/system-test-bof Survey of responses. Will find results later to copy in. Lightning Talks KAUST ~200k cores No special priveleges for tests [different to Lboro acceptance tests] Component tests (scheduler, etc) Synthetic tests (performance) VASP, WRF, etc used as well as SPEC tests. Minimised software tickets to basically zero in 24 hours following release to …
I had a talk to people from Globus. This is now using a subscription model to allow software sustainability although it is still part of the University of Chicago. Its focus is now not on access to compute resources as of old but facilitating data transfers, handling credentials at each end, retrying transfers etc. The …
Introduction Interesting discussions, especially concerning CycleCloud and the Microsoft partner UberCloud, led by Wolfgang Gentsch, original Sun Grid Engine developer. CycleCloud This was bought in by Microsoft and provides a dashboard that allows a researcher or IT department to do things such as provision a Sun Grid Engine cluster, or a Condor pool or specific …
Explosion of data, but how can this lead to effective decision making, especially high-consequence ones? Can’t just get more data as data cannot always be easily acquired, and may not always be required. Multiscale, multiphysics an issue for modelling. High dimensionality. Imperfection in the data and the models – need to indicate the uncertainty of …
Vision: Browse the Universe – high quality scalable graphics. It took a while before this could even be started as need the technology. Spatio-temporal representations. Binary search tree and octrees to divide-and-conquer the data efficiently – allows a choice between high resolution in time or space as required. Globebrowsing. Using quadtrees to cope with the …
Eurolab4HPC supporting HPC innovation. Expert view: Martin Schulz Power bound computing – use more nodes, etc. and find best configuration (makes more sense with the slides) Network contention: harder to predict execution times but cheaper in terms of hardware. Manage complex workflows Monitor the application and hardware stack at all points. [Seems more comprehensive than …
Read More “Challenges and Opportunities in Academic HPC Systems Research in 2030”
Panel Introductions Open OnDemand Vision: web based access, easy to use. In use since 2016. Good use for the Ohio machines through this. Can also create GUIs. SGCI Associated with XSEDE. Science gateways project member. ‘Full Stack’ approach. Apache Airavata. Federate multiple clusters, clients. ActiveEon HPC + AI solutions. Portals. SLURM, PBS. LSF, on premise, …
This will get expanded on with more comments and images. Due to the sheer size of the exhibitor hall it is only possible to scratch the surface without several people and coordination. Lenovo: discussed the cloud platform from Nimba (cloud in Dallas but also can use on-premise and multi-cloud) and purchases of kit. Atos: brief …