Triquetrum/Kepler

This page is a work in progress.

The goal of this page is to illustrate how to use Kepler Scientific Workflow System Actors in Triquetrum.

Note that the Kepler Scientific Workflow System is not the same as the Kepler release of Eclipse. In this page, the term Kepler is used to mean the Kepler Scientific Workflow System.

Kepler

The Kepler Scientific Workflow System has actors that we would like to use in Triquetrum. In this case, we would like to use Kepler's Distributed Data-Parallel (DDP) actors.

For a brief overview of DDP, see:

• Ilkay Altintas, Daniel Crawl, Jianwu Wang. A Distributed Data-Parallel Execution Framework in the Kepler Scientific Workflow System. October 16, 2015. Presented at the Eleventh Biennial Ptolemy Miniconference, Berkeley.

The biggest issue with the integration is that Kepler's classes are not set up for OSGi. Triquetrum/Extending Triquetrum discusses how to set up the packages.

Kepler Installation

1. See Kepler and Eclipse for how to download and configure Kepler. To get the DDP work, we use the biokepler configuration, so:

  mkdir kepler
  cd kepler
  svn co https://code.kepler-project.org/code/kepler/trunk/modules/build-area
  cd build-area/
  ant change-to -Dsuite=biokepler  
  ant clean-cache
  ant eclipse

1. Import the Kepler projects into the same workspace as to where Triquetrum is set up
2. Create a Kepler Run configuration an run it:

• project: biokepler
• main class: org.kepler.Kepler

Kepler Fails to start: ProvKAREntryHandler$Factory

If, during startup, we get this message:

java.lang.ClassNotFoundException: org.kepler.kar.handlers.ProvKAREntryHandler$Factory
  at java.net.URLClassLoader.findClass(URLClassLoader.java:381)
  at java.lang.ClassLoader.loadClass(ClassLoader.java:424)

The solution is to select use the biokepler project, not the kepler project.

Run the DDP demo

Kepler will start up. Close the splash window and under Components, double click on DDP and then wordcount-execution-choice.xml. Below is a screen shot of the initial Kepler window:

The wordcount demo has multiple ways of counting words. The default way is to use Unix commands.

Below is a run that counts words using Unix commands:

To use the other ways of counting words, double click on the Word Count actor and change the Choice parameter.

Our first goal is to get Kepler running and count words using Unix commands.

What Kepler classes are necessary?

The next step is to determine what Kepler classes are necessary.

The demo primarily uses org.kepler.ddp.actor.ExecutionChoice, which is a composite actor that contains submodels that implement the different ways the actor can be executed.

The -verbose option of the java command will report what classes are loaded.

Unfortunately, it seems that adding -verbose to the run configuration in Eclipse does not report the Kepler classes that are loaded, so we use the command line

To run the model without the UI from the command line

cd kepler/build_area
ant run -Dworkflow=../ddp/workflows/demos/wordcount-execution-choice.xml

Edit kepler/build-area/src/org/kepler/build/Run.java and insert the following into the runSuite() method:

 java.createJvmarg().setLine("-verbose");

Compile the build system:

 ant -f kepler-tasks.xml

Run the model and save the output to a file:

 ant run -v -Dworkflow=../ddp/workflows/demos/wordcount-execution-choice.xml >& /tmp/run.txt

To see how many org.kepler classes are loaded (the sed command converts any inner classes that have $ in their name to the parent class, so org.kepler.util.sql.Table$IndexType becomes org.kepler.util.sql.Table)

grep 'Loaded' /tmp/run.txt | grep org.kepler | awk '{print $3}' | sed 's/\$.*\//' | sort | uniq | wc -l

The above returns 195 classes. Not all of these classes will be necessary at runtime when running the model, but the important thing here is to make a guess at what classes are necessary.

Creating a new Kepler Bundle

See Creating an actor bundle.

Resources

• Bug 84: Test out integrating Kepler code into Triquetrum: Use DDP/Hadoop as driver.
• Triquetrum/Extending Triquetrum