XSEDE Symposium March 6 2015

PRACE-XSEDE Interoperability projects

Presenter(s): Morris Riedel (Juelich Supercomputing Centre) Sandra Gesing (Notre Dame University) Shantenu Jha (Rutgers University)

Presentation Slides Gesing Presentation

Presentation Slides Jha Presentation

1) "Smart Data Analytics for Earth Sciences across XSEDE and PRACE", speaker Morris Riedel, Juelich Supercomputing Centre.

The ever-increasing amount of earth science data arising from measurements or computational simulations requires new ‘smart data analytics techniques' capable of extracting meaningful findings from ‘pure big data'. XSEDE as well as PRACE provides excellent resources that enable efficient and effective data analytics when several technical frameworks and data analysis packages would be available. While we assessed tools and technologies for a couple of earth science case studies, the scientific case in this particular Webinar is driven by one particular earth science analytics use case: automated anomaly/outlier detection of earth science time series datasets that require a parallel and scalable clustering detection algorithm that is able to take advantage of the interoperability between XSEDE and PRACE systems today. As one of the key results of our technology assessment project, we present our parallel and scalable Density-based Spatial Clustering of Applications with Noise (DBSCAN) algorithm implementation and how it can be used across different infrastructures using open standards to decouple architecture from concrete implementations. Solutions will be outlined that can be used today in production if associated resource allocations in XSEDE and/or PRACE are granted for the user. More details will be presented at the Research Data Alliance (RDA) 5th Plenary Big Data Analytics Group Session at San Diego in March 2015 and at the European Geosciences Union (EGU) 2015 Big Data for Earth Science session in Vienna in April 2015.

2) "Unicore Use Case Integration for XSEDE and PRACE", speaker Sandra Gesing, U of Notre Dame

European Team: Molecular Simulation Community represented by MoSGrid, SCI-BUS and ER-flow and the computational radiation physics community represented by the Helmholtz-Zentrum Dresden-Rossendorf (HZDR), US: Center for Research Computing, University of Notre Dame

The project focuses on the integration of two UNICORE use cases for the joint support of XSEDE and PRACE: the first one targets the molecular simulation community, the second one the computational radiation physics community. The project MoSGrid (Molecular Simulation Grid) offers a web-based science gateway supporting the community with various services for quantum chemistry, molecular modeling, and docking applying UNICORE as grid middleware.

The main technical challenge in MoSGrid has been to extend the portal infrastructure for the intuitive use of the XSEDE and PRACE infrastructure via UNICORE and according credentials. The members of the computational radiation physics community involved in this project focuses on the generation of advanced laser-driven sources of particle and X-ray beams. They aim to "simulate what is measured" to reproduce experimental measurements and connecting them to the fundamental plasma processes on the single-particle scale. The main goal wasto make the according tools on both XSEDE and PRACE systems via UNICORE available to allow for the exchange of common workflows, which can be applied on both infrastructures. The talk will go into detail about the goals, the lessons learned and the accomplished steps.

3) "Interoperable High Throughput Binding Affinity Calculator for Personalised Medicine", speaker Shantenu Jha, Rutgers Team European: Prof Peter V. Coveney (University College London) and Prof Dieter Kranzlmuller (LRZ/LMU), US: Prof. Shantenu Jha (RADICAL, Rutgers)

To improve the ability calculate drug-binding affinities the CCS at UCL has developed the Binding Affinity Calculator (BAC), which allows the building of the necessary patient specific models required to simulate drug performance, a process which requires a complex number of steps in order to customise a generic model with patient specific information, and then run the calculations As part of this XSEDE-PRACE project, BAC has been interfaced with RADICAL-Cybertools to interoperably utilize XSEDE and PRACE resources. We will present some preliminary results

JOIN WEBEX MEETING
Meeting number: 849 368 891
Meeting password: PRACE-XSEDE