Symanzik, Juergen and Baker, Albert L. (1997) Timed Data Flow Diagrams. Technical Report TR96-23, Department of Computer Science, Iowa State University.
Timed Data Flow Diagrams
Juergen Symanzik and Albert L. Baker
Data Flow Diagrams (DFD's) are widely used in industry to express requirements
specifications. However, as used in practice, there has been no precise
semantics for DFD's, let alone an incorporation of a model of time.
In this paper, we augment the Formalized Data Flow Diagrams (FDFD's)
defined in (Leavens et al., 1996) by adding a deterministic (or stochastic)
time behavior for the consumption of values from in--flows to processes
and the production of values to the out--flows from processes.
We call our new FDFD model Timed (or Stochastic) Data Flow Diagrams
(TDFD's or SDFD's). We identify two factors in determining how time can
affect the choice of how an FDFD can change state. The first factor has
to do with when the decision is made as to which state transition
will be next occur. The two possibilities are a ``Preselection Policy''
and a ``Race Policy''. The other timing factor is the past history
of an FDFD execution. We identify three alternatives:
``Resampling'', ``Work Age Memory'', and ``Enabling Age Memory''.
Certain combinations of these alternatives allow us to model systems where
components are competing for limited resources. Other combinations allow us
to model systems where components work concurrently. Preemption can also
be modeled using these alternatives. Major results for the quantitative and
qualitative analysis of TDFD's can be borrowed from the literature on
Timed Petri Nets.
Contact site administrator at: email@example.com