[FOSSRIT] Fwd: Scientific Computing Group Meeting on Monday 11/18: Talk by Justin Flory

Hi folks, I wanted to extend this invitation to the FOSS mailing list. Come join me on Monday at 1pm to learn about containers and supercomputers! Details below. -------- Forwarded Message -------- Subject: Scientific Computing Group Meeting on Monday 11/18: Talk by Justin Flory Date: Fri, 15 Nov 2019 16:41:27 -0500 From: Daniel Wysocki Hi all, Our next Scientific Computing Group meeting is scheduled for this Monday, Nov 18th, at the usual time (1pm) and place (Orange Hall 1350). The talk is by RIT's Justin Flory, a 5th year undergraduate student in Networking and Systems Administration.  He recently worked on the IT end of a supercomputing facility, and researched the performance of different "containers", which he will be talking about here.  For those who don't know, containers are a way of packaging up an already configured computing environment, in a way that makes it portable and reproducible.  They're becoming heavily used in scientific computing, both because scientists like reproducibility, and because it means you can avoid the headaches of installing complicated programs, and just grab a container that's already been set up for you. Talk details are below, hope to see you there! -Daniel [Link] <http://rit-scg.com/talk/2019/11/18/docker-containers-and-supercomputers/> *Title:* Docker containers + supercomputers = ?? *Abstract:* How do you deliver research software into research computing environments? Or supercomputers? Docker containers and related products are revolutionizing IT infrastructure, but where do containers belong in supercomputing / High-Performance Computing (HPC) infrastructure, on large distributed computing grids? In a world of proprietary hardware and drivers, large-scale distributed systems, and emphasis on bare-metal performance, are containers another virtualization fad to skip over supercomputing? Guess again. This session explores different container run-times built for supercomputing / HPC environments, what they offer, and some benefits/costs of implementing them alongside HPC job scheduling software. This session will be useful for you if you work with HPC infrastructure, you write code intended to be run on parallel systems, or you are interested for what the future of HPC technology might look like.