PCM Development/Palladio Concall/Minutes 20200406

Attendees

many open / in progress issues
reason for slow progress?
- unsure if issue should just fixed or looking for other solution (SIMUCOM-71)
- use second suggested option (execute transformation separately and check results, maybe fix it later)
handling of upcoming tasks while working
- pull them into the sprint (ignore warning about changing scope of sprint)

left over from last concall
difficult
- namespace URIs are scattered through multiple meta models
- we have to change versions in namespace URIs and provide scripts
steps
- identify affected projects
- think about how to carry out renaming
- perform refactoring
TODO Stephan: add a JIRA task about renaming and add it to maintenance category

Specify remote API for EDP2
status: add client for API to Palladio
next term: Sebastian proposed topic for structured storage of EDP2 in other formats (afro, parket, ...)

Handling of org.palladiosimulator.cost
- referenced from some ATs (see AT-7)
- only available in SVN but not on Github
Bundle looks interesting, might be interesting for various use cases
TODO Stephan: Migrate it to Github (Addon) and integrate it into aggregated update site

background: StoEx Xtext grammar
- contains keywords for bytesize, etc. that do not belong there
- new StoEx parser cannot create CharacterizedVariable (leads to problems e.g. in type inferror)
Replace NamedReferences (string only) by cross references to abstract model element that PCM can implement (e.g. by implementing the interface by Parameter)
Provide extension mechanism for scoping to include PCM elements
Replace usage of PCM::CharacterisedVariable with StoEx::Variable
Is essential for more flexible characteristics
Plan: do this in the next month
TODO Stephan: create issues and add it to the StoEx category

Issue at hand: Execution Result Metrics of different simulation runs cannot be compared since metric ids differ
Backgroud: Execution Result
- Result of usage scenario execution
- Textual Base Metric which specifies all possible literals, measurement contains one literal
- One success literal; one literal for each possible failure
  - Failure in InternalAction per Assenbly
  - ResourceFailure per Resource per ResourceContainer
  - Failure of Linking Resource
- Consequently, Execution Result Metric depends on PCM instance.
- Metric is created at the beginning of a simulation anew.
Possible solutions:
1. Won't fix. Execution result metrics cannot be compared by design.
  - drawback: not possible to draw to measurements in the same visualization, but there might be no real benefit
  - pie charts do not support two metrics anyway
2. Precalculate execution result metric (--> MeasurementsUI) and validate compatibility in simulation
3. Adapt to generic metric (Success, Failure / SoftwareFailure, HardwareFailure, NetworkFailure)
decision: won't fix