TIME IN PETRI NETS - Modelling with Generalised Stochastic Petri Nets - page 57

p active 1 p active 2

T request 1 T request 2

p requesting 1 p requesting 2

p idle

t start 1 t start 2

p accessing 1 p accessing 2

T end 1 T end 2

Figure 3.5: Timed version of the PN system in Fig. 1.5.

while accessing their private memories. The temporal characterization of
these two transitions is based on either measurements or assumptions about
the duration of such operations. Transitions T end 1 and T end 2 model the ac-
tivities performed by the processors while accessing the shared memory. The
temporal characterization of these two transitions is again derived by mea-
surements or assumptions about the durations of shared memory accesses.
Transitions t start 1 and t start 2 model the acquisition of the interconnection
network and of the shared memory; the choice of using immediate transitions
here amounts to neglecting the delays inherent in such operations.
As a second example of use of immediate transitions, consider now the fol-
lowing modelling problem: when a given event occurs, all the tokens that
are present in place p a must be removed and an equal number of tokens
must be generated in place p b . Informally, we may say that all the tokens
contained in p a are moved (or migrate) to place p b . This is a typical problem
that arises when several items are waiting for a synchronization before being
allowed to proceed to an execution phase. Implementing this operation in
one step is not possible, due to the fact that arc weights are assumed to
be constant in the nets that we conside r 1 . The same result can however be

1 Some authors extended the formalism by introducing marking-dependent arc weights,