workflow addenda
The concrete swineherd runnables each have eponymous stage names.
-
CpRunner
-
ScpRunner
-
PigRunner
-
HadoopRunner
-
HadoopStreamingRunner
-
WukongRubyRunner
-
FlumeOgConfigRunner
Any simple Rake task should work as a swineherd flow
- task
- desc
- namespace
- ... (flesh out)
Can dispatch to another stage dynamically -- for instance, the FilenameRule stage handles the familiar "build any .o file from its eponymous .c file" rule.
(?mystery: how to handle resolution order? Part of the unsolved specificity question-?)
Does not guarantee parallel execution -- just allows it
- many have an
updated_atdate, ornilif it doesn't exist.
mystery: what about things without a meaningful date?
http://www.workflowpatterns.com/documentation/documents/phd_bartek.pdf
-
Activity represents an atomic piece of work
- trigger conditions: any, all, custom
-
Work Node -- once executed, not again until reset
-
Merge Node
-
Split Node
-
Route Node
-
Abort Node
-
Reset Arc
-
Parallel Split
-
Synhronization
-
XOR
-
Simple Merge
- resource: an entity that is capable of doing work
- durable or consumable in nature
- resource may have a schedule and history associated with them
http://www.workflowpatterns.com/patterns/
In process-aware information systems various perspectives can be distinguished.
- The control-flow perspective captures aspects related to control-flow dependencies between various tasks (e.g. parallelism, choice, synchronization etc). ...
- The data perspective deals with the passing of information , scoping of variables, etc
- The resource perspective deals with resource to task allocation, delegation, etc.
- Finally, the patterns for the exception handling perspective deal with the various causes of exceptions and the various actions that need to be taken as a result of exceptions occurring.