Graphics Programs Reference
In-Depth Information
CPU
terminals
Figure 3.15: Queueing network model of a timesharing system
p
1
T
1
p
2
T
2
Figure 3.16: Timed PN model of a timesharing system
of users, shown in Fig.
3.15.
It can be translated into a timed PN system
as shown in Fig.
3.16,
where the infinite-server station that models the ter-
minal think time is represented by timed transition T
1
with infinite-server
semantics, and the single-server station that models the CPU activity is
represented by transition T
2
with single-server semantics.
This same example also allows us to observe that service policies (like FIFO
or LIFO) cannot be directly embedded into the enabling/firing mechanism of
a single transition because tokens cannot be individually recognized. These
policies must be described with the help of specific subnets that actually
implement the selection algorithm of the specific queueing discipline.
The infinite-server and multiple-server semantics are just an example of the
possibility for temporal specifications to depend on the marking of the PN
system. With these semantics, only the places that determine the transition
enabling degree influence the transition timing. More complex relationships
can be envisioned, but an extensive exploitation of marking dependencies
may introduce problems in the model specification and in its dynamic be-
haviour. We shall return to this point in a later chapter
Search WWH ::
Custom Search