Translated note from the discussion on this from the Kieker/Palladio Days.
Google translate would probably have been the preferred service for this . Will attach the German notes (thanks to Nils for taking them) to this ticket as a reference. Feel free to fix items.
This ticket could serve to collect related issues, referring to other more specific tickets.
- Default cases (automatic generation of dependency graphs)
- refer to graphical tool
- Also for monitoring (AOP)
- Start with useful default configuration
- Default case without further refinements should work already
- Refinements can be made later
- Double-tick on file should do s.th. useful
- Tool for measuring resource utilization (CPU and MemSwap Utilization)
- If more is needed, user need to configure
- Goal: fast but usable
- Currently many lookups, String comparisons
- Refine concepts:
- currently quite "fixed/inelastic"
- Ports should become first-class entities
- Reflection is expensive
- Low potential for parallelization
- EMF Meta model
- Extension of models?
- Only for serialization not "useful"
- Throw away or keep?
- String concatenation
- Implement concurrency/parallelism from the beginning; one thread per filter
- Repositories: Somebody developing repositories should not have too many possibilities to make mistakes
- Related: System Meta Model should become EMF Model
- Analysis (closed) should be independent of EMF
- Florian: Has 1:1 Ports only (Possibility: Inherit from 1:2 or 2:3 port filters); Ring buffer; Concurrency; event.driven; every second, a record is sent through the architecture; Have a look at the Disruptor framework!!
- Visualization of interconnection