Conclusions

• Some basic problems of TBM/SimEvents
  • timing of concurrent events
    • null server is the wrong solution!
  • implementation of alternative queueing policies
    • a new (or old) abstraction is needed
  • storing and retrieving of entities
    • again: a basic concept is missing
  • gathering and processing of statistical data
    • a few supplementing blocks would be handy
• Proposal: new Argesim benchmark with
  • modeling of jockeying queues
  • statistical data over several queues
  • additional complication: many queues
• SimEvents 5
  • complete redesign of the library
  • based on a unifying theoretical description (but not DEVS)
  • migration from SimEvents 4 difficult!
  • is it better? We'll see ...