Zum Inhalt springen
PCM Development/Palladio Concall/Minutes 20161212
Attendees
- Steffen Becker (TUC)
- Christian Vögele (Fortiss)
- Markus Frank (TUC)
- Michael Langhammer (KIT)
- Misha Strittmatter (KIT)
- Sebastian Krach (FZI)
- Stephan Seifermann (FZI)
Stateful/Stateless in Palladio
- Christian speaks for Felix
- Several PCM extensions developed / planned by fortiss
- Recurring issue: stateful/stateless
- Examples
- Memory extension by state that decides about garbage collection (by Felix)
- Need state for several extensions (simulation works, analytical solver most probably not)
- Arguments against stateful plugins?
- Steffen: old discussion (publications of Lucia, stateful PCM branch)
- Steffen: main problem are solvers because do not converge anymore (state space explosion)
- Steffen: often probabilistic modeling is a workaround for state
- Steffen: use cases for stateful PCM might exist (e.g. embedded systems) but do not require properties that cannot be calculated with state anymore
- Misha: State in language or in simulation/solver?
- Christian: Example state-based GC triggering
- Steffen: Garbage collection can be modelled with probabilistic modeling
- Christian: Paper available?
- Steffen: Watch processing rate, if 10% GC time, ...
- Christian: So, the major issue is introducing new parameters in language?
- Steffen: Yes, because wrong usage of PCM.
- Christian: Passive resources have state. Do they work with LQN solvers?
- Steffen: Yes, is treated as a special construct.
- Christian: Built an extension that introduces guards for Actions introduced (if action A did not run yet, action B must not run)
- Steffen: Steady-state assumption does not hold anymore, not proovable. If not important for use case, it is ok.
- Steffen: Would use two separate runs for the two situations.
- Christian: Will talk with Felix if probabilistic modeling could address issues.
- Misha: Might be a topic for the next concall again
JIRA Documentation Type
Editor Value Editing in Ressource Environment
- Resource environment holds several values
- linking resource: throughput, latency
- are random values
- good enough if editable via edit box or should StoEx editor be available?
- latency could be useful, but is processing rate really a random variable?
- Steffen: Garbage collection or background processes can be modelled with this feature
- Misha: Do simulators support this?
- Steffen: Yes, Simulizar and SimuCom at least.
- Misha: OK, will add the StoEx dialogs
Updated QuAL Report
- Only one author and branded with CloudScale
- Steffen: No reason against converting the document. Would be appreciated.
EMFProfiles update site
- Update site was down, Christian repaired it and pointed to old build
- Steffen: We have own EMFProfiles that works and deploys nightly builds
- Steffen: Issue with GIT-based builds fixed for Sirius editors but other projects might have to be adjusted as the Sirius editors
- Update site still has to be checked
- Michael: Nightly drop from last week worked but not tested deeply
Meta-Model Generator
- Steffen: generation of meta-model code works again, so ready for meta-model release
- Simulizar and SimuCom build against nightly build of PCM, has to be changed back to release when releasing