the dark energy survey middleware lsst workflow workshop 09/2010
DESCRIPTION
The Dark Energy Survey Middleware LSST Workflow Workshop 09/2010. Michelle Gower NCSA o n behalf of the DES Data Management Team and DES Collaboration. DES Data Management. Observing from 2011-2016: 200 TB raw data, 4 PB of products 100 TB object catalog - PowerPoint PPT PresentationTRANSCRIPT
National Center for Supercomputing ApplicationsUniversity of Illinois at Urbana-Champaign
The Dark Energy Survey Middleware
LSST Workflow Workshop 09/2010
Michelle Gower
NCSA
on behalf of the DES Data Management Team and DES Collaboration
DES Data Management
• Observing from 2011-2016:• 200 TB raw data, 4 PB of products• 100 TB object catalog• 120+ CPU years/year of processing
• Processing includes:• Removing the instrument signature• Removing artifacts (e.g. planes)• Calibrating/registering• Feature detection• Feature analysis• Monitoring/quality assessment
Requirements
• Level of parallelism changes through processing• Easily modified by scientists• Local and remote clusters• Work with project’s Archive system• Monitoring of jobs while they’re running• Less research/More production
Processing Framework Overview
4
ArchiveNodes
Target (HPC)Machines
NotificationService
Database
Pipelines of AppModules
EventMonitor
ArchivePortal
Orchestration
DAF
Middleware
• Workflow• Condor DAGman
• Job submission• Condor-G to pre-WS GRAM for TeraGrid resources• Condor (vanilla jobs) for local machines
• File transfer• GridFTP using clients uberftp and globus-url-copy
• Runtime Monitoring• Elf/Ogrescript
Workflow
Crosstalk CreateCor ImCorrect
Masking AstroRefine
Remap PSFModel WeakLensing
Make Bkgd
• Run queries to get master input file lists• Stage input images to target machine• Generate Jobs: input lists, job descriptions, DAG• Setup target machine• Stage generated lists and files to target machine• Make timestamp on target machine• Target jobs run• Ingestion of new or modified files
Job Generation
Upcoming Orchestration Work
• Skipping modules• Repeatedly using set of blocks (Campaign
processing)• Let condor manage vanilla jobs per machine
• threaded vs serial
• Finer grain restarts
Notification Events System
AppModule (Elf/OgreScript)
Science Code
ActiveMQ Message Bus
Event StorePortal Monitor
DES Monitor Portal
• Current Technology• Drupal CMS / PHP, Javascript/Ajax
• Tools for Viewing Active Processing Status • High level Alert Monitoring
• Quick navigation to event logs
• Quality Assurance Profiles
• Histograms of QA metrics, outlier detection
• Processing Timing Summary Report
• Middleware level timing profile
Acknowledgements
• DESDM Team:• Jim Myers, Terry McLaren• Joe Mohr• Bob Armstrong, Dora Cai, Ankit Chandra, Greg Daues,
Shantanu Desai, Michelle Gower, Wayne Hoyenga, Chit Khin, Kailash Kotwani
• Past DESDM Team Members, DES Project Team and Collaboration Members
• National Science Foundation
• http://cosmology.illinois.edu/DES/