Galaxy Code and Storage Architecture from GCC 2013

Talk on Galaxy code organization from the 2013 Galaxy Community Conference given by Nate Coraor (@natefoo) and myself.

Transcript

1. Galaxy Code and storage architecture www.getgalaxy.org @jxtx / @natefoo / #usegalaxy
2. Galaxy application architecture
3. galaxy.util. pastescript.serve Middleware Stack galaxy.webapps.galaxy. GalaxyWebApplication WSGI Web Server galaxy.app. UniverseApplication Toolbox Job Manager Model Datatypes Reg. ...
4. Job Handlers Job Manager Job Handlers Job Handlers <<thread>>
5. Splitting by process galaxy.util. pastescript.serve Application Stack galaxy.webapps.galaxy. GalaxyWebApplication galaxy.util. pastescript.serve Application Stack Job Handlers Job Manager Job Handlers Job Handlers
6. Views Model Controller Web UI API CLI JS Lib UI Architecture
   Python Lib (BioBlend)
7. Code organization and standards http://wiki.galaxyproject.org/Develop/BestPractices
8. (let’s look at the code)
9. These are best practices.
   “foolish consistency is the hobgoblin of little minds”
   Standards
   Galaxy mostly follows PEP-8, readability is the ultimate goal
   Avoid “from module import *”
   Comment lines should be under 79 characters, code lines can be up to 200 characters if it improves readability
   Docstrings need to be reStructured Text (RST) and Sphinx markup compatible, documentation is automatically generated
   Whitespace: whatever is most readable, both for blank lines and space around operators
   What about Javascript?
10. Galaxy data storage
11. Metadata stored in relational database
    (PostgreSQL Please!)
    ! Dataset files stored on filesystem
    in objectstore
12. Data Abstraction
    Disk Disk S3 iRODS Galaxy Object Store
13. Data Abstraction
    >>> fh = open( dataset.file_path, 'w' ) >>> fh.write( ‘foo’ ) >>> fh.close() >>> fh = open( dataset.file_path, ‘r’ ) >>> fh.read() >>> update_from_file( dataset, file_name=‘foo.txt’ ) >>> get_data( dataset ) >>> get_data( dataset, start=42, count=4096 )
14. Data Abstraction
    Distributed Object Store FS FS FS FS Galaxy Distributed Object Store Distribution by weight Zero weight
15. Data Abstraction
    Benefits • Grow beyond original capacity • Avoid migrating data offline • Tier storage • Let your users bring their own storage • Use resources w/o a shared filesystem (with iRODS) • Remove IO bottlenecks
16. Job scheduling and running
17. Job Control job_conf.xml Torque DRMAA SSH Globus SSH Job Handler Torque (CLI) SGE (CLI) Condor
18. Extension: adding a new job runner
19. Running as actual user
    ! User and group specific storage (?)
20. Galaxy workflow system
21. Workflow representation: data flow graph
    ! Scheduling like any other job
    ! New: fix and rerun from failure point (workflow automatically paused)
    ! Future: background scheduling, decision points during scheduling
22. Galaxy API
23. Galaxy API
    Technologies • Representational State Transfer (REST) • Sessionless operations via HTTP • JavaScript Object Notation (JSON) • Uses a key (rather than user/password)
24. Galaxy API
    Core Interfaces
    • Library permissions • Forms • Server configuration (view) • Sample tracking requests and samples • Manage users, roles, and quotas • Execute tools and workflows
25. Galaxy API
    Core Interfaces
    • Histories and History Datasets • Libraries and Library Datasets • Tools • Workflows
26. API Interactions with Galaxy Galaxy Galaxy’s REST API $GALAXY_HOME/scripts/api python lib: blend JavaScript lib
27. Bare Bones Galaxy API >  create (POST) >  display (GET) >  update (PUT) >  delete (DELETE)
28. Making the Calls Wrapper methods exist (in /scripts/api/) to make the API calls easier: ./{action}.py-<api-key>-http://<ip>/api/{module}/[unit]-[args]-- action:-create-|-display-|-update-|-delete- api_key:-obtained-from-the-UI- module:-datasets-|-forms-|-histories-|-libraries-|-permissions-|- quotas-|-requests-|-roles-|-samples-|-tools-|-users-|- visualizations-|-workflows-- unit:-dataset_id-/-history_id-/-library_id-/-…- args:-name-/-keyHvalue-pair-/-…-
29. http://galaxy-central.readthedocs.org
30. http://galaxy-central.readthedocs.org
31. http://galaxy-central.readthedocs.org
32. Extending Galaxy
33. Tools & Datatypes
    ! Hopefully everyone understands this by now?
    ! There is another session,
    “Introduction to Tool and Data Source Configuration”
34. Visualizations
    ! (See talk from Carl Eberhard)
35. Visualization and visual analytics
36. Architecture 36 Datasets ... Tools ... Galaxy Web browser Galaxy HTML UI
37. Architecture 37 Datasets ... Tools ... Data Providers Galaxy Web browser Galaxy HTML UI
38. Architecture 38 Datasets ... Tools ... Data Providers Galaxy ... Web browser Data Managers Visualizations d3.js (SVG) HTML5: Canvas, CSS Galaxy HTML UI
39. Technical  Highlights Rough  pluggable  support  for  generic  JavaScript  visualizations  +  data  providers   ! Data  providers  =  fast,  random  access  to  data  in  Python,  JS   ! API:  run  tools,  run  them  on  data  subsets   ! Backb0ne  +  HTML5  objects  for  Web-­‐based  genomic  visualizations   ✦ e.g.  data  managers,  linear  and  circular  views   ! JS  binding  to  Galaxy  API  (blendJS?)   ✦ visualizations,  tools,  datasets   ✦ custom  Galaxy  UIs
41. Eggs and dependency packaging
42. New controllers and webapps
43. Q&A