Conveners
Workflowmanagement for Parallel Computing
- Michele Martone (Leibniz Supercomputing Centre)
Based on the precondition of the availability of Kubernetes as a Service, this talk is going to show why the migration of legacy software projects to a Kubernetes cluster is a worthwhile undertaking and how it can succeed, including the decisions made for the tools and the lessons learned on the way.
Starting from the current state with multiple deployment tools and environments (VMs,...
Task farms can be used to solve embarrassingly parallel workloads where a number of independent tasks need to be performed. This presentation introduces [taskfarm][1], a python client/server framework that was designed to manage a satellite data processing workflow with hundreds of thousands of tasks with variable compute costs. The server uses flask to hand out tasks via a REST API and a...
Soft matter systems often exhibit physical phenomena that resolve different time- or length-scales, which can only be captured in numerical simulations by a multiscale approach that combines particle-based methods and grid-based methods. These algorithms have hardware-dependent performance characteristics and usually leverage one of the following optimizations: CPU vectorization, shared memory...
