Distributed User Support and Troubleshoo4ng for an e‐Infrastructure for Bioscience Research Silvia D. Olabarriaga Eduard Drenth Mark Santcroos Simon Dalmolen Antoine van Kampen Wico Mulder e‐Bioscience Group Collabora4ve Network Systems BioInforma4cs Laboratory Academic Medical Center (AMC) Logica www.bioinforma4cslaboratory.nl www.logica.com S.D.Olabarriaga@amc.uva.nl
Summary • Context • Problem • Approach • Pilot • Final remarks 14 September 2010 EGI Tech Forum 2010 2
Virtual Laboratory for e‐Science This work is supported by a BSIK grant of the Dutch Ministry of Education, Culture and Science and is part of the ICT innovation programme of the Dutch Ministry of Economic Affairs 14 September 2010 EGI Tech Forum 2010 3
CNS @ Logica • Collabora4ve Network Systems • Exper4se team to set collabora4on between organiza4ons with IT solu4ons – Agent technology – Seman4c reasoning – Self‐organiza4on 14 September 2010 EGI Tech Forum 2010 4
e‐Bioscience Group @ AMC • AMC: – hospital – medical (informa4cs) school – research ins4tutes – spin‐off • e‐Science for Biomedical research • e‐infrastructure for biomedical research – e‐BioInfra – Dutch NGI (BiG Grid) 14 September 2010 EGI Tech Forum 2010 5
e‐BioInfra: Layered Architecture 14 September 2010 EGI Tech Forum 2010 6
e‐BioInfra: Usage • Supported by eBioscience group – Applica4on por4ng, workflows, experiment monitoring • Applica4ons – Neuroimaging • MRI and func4onal MRI • Diffusion Tensor Imaging (DTI) • CT Angiography – Bioinforma4cs • DNA next genera4on sequencing • Proteomics • Metabolomics 14 September 2010 EGI Tech Forum 2010 7
e‐BioInfra: Actors From DNA Sequencing Platform Schaik, Luif et al, EGEE UF 2010 14 September 2010 EGI Tech Forum 2010 8
Problem: monitor large experiments • Grid‐related errors… – Span mul4ple domains – Span mul4ple 3rd‐party so[ware components – Are o[en indicated or detected in log files • Troubleshoo4ng is difficult... – Requires much manual interven4on – Knowledge is distributed – Consumes significant amount of man power – Workflow vs. job level 14 September 2010 EGI Tech Forum 2010 9
Grid workflow execu4on 14 September 2010 EGI Tech Forum 2010 10
Opportunity: Agents From GMAC 2009 From EGEE TF 2009 14 September 2010 EGI Tech Forum 2010 11
Approach • Agent framework – Autonomous elements – Independent, loosely coupled layer – Communica4on across domains – Intelligence • DUST – Distributed User Support and Troubleshoo4ng – Pilot project 14 September 2010 EGI Tech Forum 2010 12
Pilot Project • Develop intelligent assistants to (semi) autonomously monitor the execu4on of workflows on the grid using MOTEUR. • Monitor logs on the server (and grid) side, • Detect (relevant) events and • No4fy the user and support team 14 September 2010 EGI Tech Forum 2010 13
Current Situa4on 14 September 2010 EGI Tech Forum 2010 14
Target Situa4on 14 September 2010 EGI Tech Forum 2010 15
Technical Implementa4on 14 September 2010 EGI Tech Forum 2010 16
FIPA DS: Directory facilitator AMS: Agent management system (JADE implementation) 14 September 2010 EGI Tech Forum 2010 17
14 September 2010 EGI Tech Forum 2010 18
14 September 2010 EGI Tech Forum 2010 19
Preliminary evalua4on • Proof‐of‐concept – Agents take care of communica4on – Simple func4onality (job status, error no4fica4on) – Flexible implementa4on – Minimally invasive • Need to assess: – Scalability – Complexity – Prac4cal value 14 September 2010 EGI Tech Forum 2010 20
Discussion • Engineering grid applica4ons is challenging • System complexity is bound to increase • Problems are bound to occur in such dynamic and complex systems • Informa4on and exper4se is bound to remain distributed (in produc4on grids) • Troubleshoo4ng needs to be approached properly (which framework?) – Applica4on vs. grid level – Pilot job frameworks 14 September 2010 EGI Tech Forum 2010 21
Thanks for listening! 14 September 2010 EGI Tech Forum 2010 22
Recommend
More recommend