eResearch workflows for studying free and open source software development
Download as PPT, PDF6 likes596 views
1. The document discusses using scientific workflows and tools like Taverna for distributed collaborative research on free and open source software development using large datasets, computational resources, and reproducible analysis. 2. Taverna is presented as an example of a scientific workflow tool that allows modular development of analysis through reusable components with input and output ports, offering advantages over scripts. 3. An example workflow is shown that calculates network centralization in dynamic networks and generates time series and CSV output for further analysis.
Ad
More Related Content
Similar to eResearch workflows for studying free and open source software development (20)
More from Andrea Wiggins (20)
Ad
Recently uploaded (20)
Ad
eResearch workflows for studying free and open source software development
- 1. eResearch workflows for studying free and open source software development James Howison, Andrea Wiggins, & Kevin Crowston Syracuse University School of Information Studies 9 September 2008 ~ IFIP 2.13 - OSS 2008
- 2. eResearch Scientific practices and technologies permitting distributed research collaborations using: Large data sets Computational resources Analysis tools and workflows Replicable research with provenance metadata FLOSS research community is starting to move in this direction
- 3. Current FLOSS Research Practices Data increasingly available in “repositories of repositories” FLOSSmole FLOSSmetrics SRDA (Notre Dame) Not much sharing of analysis or methods for calculating measures; mostly bespoke scripts
- 4. FLOSS Research Repositories Java projects plus user contributions Pilot: KDE Partial: Sourceforge, ObjectWeb, Apache, GNOME Sourceforge Sourceforge, Rubyforge, ObjectWeb, Debian, freshmeat Sample collected In Debian Downloads & pageviews Downloads, pageviews, ratings, in Debian, partial: actual use Use & popularity Size & complexity metrics Planned/pilot: size & complexity metrics SVN/CVS full, size, packages, releases & dates SVN/CVS count, packages, releases & dates SVN/CVS count, packages, releases & dates Software venues Planned/pilot: mail lists & trackers Forums & trackers Releases; in progress: mail lists, forums, trackers Communication venues Memberships & roles Memberships & roles Memberships & roles Developer demographics Basic data & confirmed locations Basic data Basic data Project Demographics Source kibitzer Qualoss & SQO-OSS FLOSSmetrics & CVSAnalY SRDA (Notre Dame) FLOSSmole
- 7. Scientific Workflow Tools Tools for scientific analysis (e.g. Kepler, Taverna) Self-documenting analysis Analysis conditions recorded at time of execution Steps in workflow executed by components with multiple input and output ports Components linked by joining input and output ports Supports modular analysis development, associated with easier collaboration and higher quality products Represented as flow diagram, stored and shared as single XML file
- 8. Why Workflows Instead of Scripts? Differences and advantages over scripting languages Wider accessibility Programming skills helpful but not necessary to start Compatibility Easier integration of heterogeneous components Mash up and reuse existing scripts Standardized metadata Out-of-the-box operation Requires workflow software to execute, but little other configuration is required Example: BioPerl evolution from Perl scripts to libraries to workflow suites
- 9. Taverna Workbench Open source stand-alone desktop application in Java Can also be run via “headless” server application Two main interface modes (v 1.71) Design mode: workflow definition, automatically rendered diagram, available component types Execution mode: process monitor, intermediate values, results, XML reports of status and all values Local and remote components are connected through input/output ports in MIME typing Can also be grouped into subworkflows
- 12. Workflow Component Types Abstract & notification processors String constants: file locations, threshold values, etc. Beanshell: simplified Java Rshell: R statistical program running through Rserve Java widgets/shims for common operations i/o, lists, text manipulation, JDBC, etc. Command-line Web services WSDL with SOAP Can be “scavenged” from URLs or other workflows
- 13. Example Workflow Teal: inputs Light blue: outputs Other light blue: string constant Green: web services Purple: Java shims Yellow: RShell Calculates weighted network centralization in dynamic networks, generates both numeric & graphical output
- 14. Example Output Inputs Sliding window size Project venue Date range Outputs Time series graph of centralizations with summary stats CSV of dates and centralization values for additional analysis
- 15. Benefits of Using Workflows Modular design yields benefits in flexibility, transparency, and ease of reuse Easier to co-develop and debug components, and to integrate independent efforts Can quickly change strategies with minimal adjustment to existing workflow structure Can reuse existing scripts and workflow components Can also conduct exhaustive sensitivity testing Multiple ways to achieve analysis tasks
- 16. Conclusion Combination of growing large-scale data sets and workflow tools present great opportunity for eResearch on FLOSS Work needed for eResearch infrastructure: Access to data Ontologies for naming data and defining relationships Incorporating metadata & social science data, e.g. content analysis schemes Trade-offs involved in standardizing on tools to benefit collaboration, but much potential gain
- 17. More Taverna demo screencast Long version (24 minutes): floss.syr.edu/Presentations/tavernaDemoScreencast.mov Short version (14 minutes): floss.syr.edu/Presentations/TavernaDemoRedux.m4v MyExperiment FLOSS group www.myexperiment.org/groups/64 16:30 - 17:30 presentation today: Social Dynamics of FLOSS Team Communication across Channels Tomorrow: Workshop on Public Data about Software Development (WoPDaSD 2008)