The Institute for Genomic Research (TIGR) has many process piplelines that need to be created, executed, and monitored on an on-going basis. Each pipeline may include multiple discrete process that can be executed either sequentially or in parallel. To reduce manual intervention, and streamline the process flow, TIGR's Annotation software team has designed a system called Workflow that can be used to build, run, and monitor such process pipelines or workflows.
The goal of the Workflow system is to fully automate the process of executing a pipeline. The fully functional version of the Workflow system is expected to provide the ability to do the following:
This section here describes the Work Flow System development plan, possible release schedules and project milestones.
|Iteration 1||January 2003||
|Initial beta release to support serial and parallel processors|
|Iteration 2||March 2003||
|Support GUI for browsing and creating workflow|
|Iteration 3||May 2006||September 2006||
|Compression of file, template editor|
|Iteration 4||August 2006||January 2007||
|Remote dispatcher, workflow checker, control workflow, local grid submissions|
Iteration 1: This iteration is basically intended to come up with a working prototype with basic functionality. The features to be implemented in this iteration are 2, 3.1, 3.2, 4, 7.2, and 8.1 from the above specifications.
Iteration 2: GUI for browsing workflows. Features 1.2, 1.4, 1.5.
Iteration 3: GUI for creating and editing templates 1.1
Iteration 4: Reset and restart commands (4.1, 4.2), provide status on command line (7.1)
Implementation:The execution of a pipeline is driven by a CommandSet. Each CommandSet can contain multiple Commands or CommandSets. For a detailed CommandSet definition, refer to the CommandSet Schema or dtd.
Anup Mahurkar, Dan Sommer and Sreenath Nampally.
Last Updated: January 16th 2007