Combined use of HPC, Cloud and HTC systems
- Enol Fernandez (EGI.eu)
The combined usage of HPC, Cloud and HTC systems is a common requirement for multiple research disciplines. EGI recently started an HPC-integration working group that brings together user communities, HPC providers, and technology providers to assess and plan the technical integration among various types of EGI services and HPC systems. This session will provide an overview of integration experiences, integration possibilities and some of the use cases that motivate the work itself. The session will be interactive, and will collect further input from the audience into the technical integration plan for EGI. The session will be relevant for scientific communities that need to run workloads across heterogeneous compute systems; and for compute service providers who want to support such heterogeneous workloads.
Main target audience
User communities with HPC requirements, HPC providers.
Currently 33 National Competence Centres (NCCs) for HPC (and associated technologies) are set up within the frame of the EuroCC project. This is accompanied by the CASTIEL activity to support this on a European Level by maximizing the synergies between the nations and thus boost the evolutions of the National Competence Centres (NCCs).
This presentation gives a short overview of the the transcontinuum concept that influences the current European roadmap design for HPC and HPDA future cyberinfrastructure.
This concept captures the need to provide application developers an end-to-end abstraction of the infrastructures (HPC centers, Cloud, Fog, sensors) needed to combine HPC computations, data life cycle management, AI...
This talk introduces the Ophidia High-Performance Data Analytics (HPDA) framework for scientific multi-dimensional data analysis, its multi-layered architecture, as well as the main challenges related to its deployment. In particular, it presents the solution developed for dynamic deployment of Ophidia over different infrastructures (i.e., HPC and Cloud clusters).
Furthermore, the talk...