Dimensions of Dynamism

• Locus of control: external or internal?
• Locus of control: centralized or decentralized?
  • May be different at design & impl time
  • May not be a decision made by architect?
  • David: considering the most decentralized dynamic architecture... Kramer & Magee's Sieve of Eratosthenes prime number generating system
    • A filter generates positive integers in sequence starting at 2; creates a 2 multiple filter; it and following filters creates a new filter when a number is not divisible by this filter.
    • localized -- both creation and model is very localized
    • unbounded -- new filters can be continuously created
• Field of view: complete or partial?
• What may change vs. how does it change?
• How can we abstract away from these decisions at the modeling level?
• George: abstract data type and dynamic architecture--how same or different?
  • operations on ADT similar to operations on structure in dynamic arch
  • Jonathan: normally specify stack in terms of input/output, not usually in terms of the structure
• Structure vs Behavioral dynamism
  • Challenge: can we combine the two, e.g., what Wright and Rapide did?
    • Wright has limit on size, number of CSP channels static
    • Darwin, "pointers" problem
  • Jung Soo's take on behavior
    • Start with operational behavior of elements, no dynamism yet, and first be able to check conformance of behavior
    • Then extend to being able to support dynamism of behavior, such as element creation

What about brainstorming on goals?

• Extend from the list of current limitations?
• ?

Major Open Issues:

• Type dynamism
  • completely unexplored
• Modeling emergent systems
  • Does architecture have benefits here?
  • e.g., Kramer/Magee prime generator
• Modeling richer structural dynamism
  • "Not" on limitations
• Modeling structural & behavioral dynamism together
  • Jonathan hasn't seen an example of hierarchy and dynamism being considered together
  • e.g., In a plug-in architecture, modeling both the internal behavior of plug-in and the top-level system coordinating the plug-in
• Dynamism and abstraction
  • Hierachy, non-determinism
  • Multiple views
• How to transition to implementation and/or check conformance
  • Maybe more than one impl/more than one refinement
• Richer analysis
  • Incremental progress towards goal/do no harm
    • e.g., (David) Rainbow, which isn't based on notion of correctness, but of improving an attribute and being close to goal
  • Concurrency -- what changes interfere with each other or with overall system functionality & quality attributes
  • Stability (from the control world) -- can you guarantee that a fix to one problem doesn't destabilize another, whose fix breaks the original problem
    • notion of objective function
• How to separate concerns in dynamism, a serious engineering issues
  • e.g., Rainbow, in terms of adaptation, expertise about system will reside in different people; ideally you want to compose the change units without conflicts, how?
• Field study of dynamism applications today
  • What kinds of dynamism are needed?
  • What could be checked?

Reading Groups in Spring '05

• From SHS, retracing the idea from "universal architecture" for dynamic systems, one where a component has control interfaces (M/A) in addition to steady-state interfaces (I/O)
• Four papers:
  • Dynamic Wright
  • Wermelinger
  • ArchWare
  • Kramer/Magee
  • CHAM?
  • Graph grammars
  • Composite web services -- formalizations, e.g., state-machine based (IBM's BPL & WSDL)
  • Autonomic computing
• Develop Wiki web as part of course
  • c2 wiki