Methodologies for Design of Service-Based Systems - Intentional Perspectives on Information Systems Engineering - page 309

to be potentially invoked in a process at run time are part of the process design
phase.
The process description is annotated with local and global quality of service con-
straints for the process. In this way, at run time, concrete services can be selected and
invoked using QoS optimization techniques, and process self-healing mechanisms
can be activated in case of failure.
In this framework, the adaptation may depend also on the context of execu-
tion. The context may determine the way in which component services are selected,
changing the quality needs for a business process. For instance, the importance of
different quality of service dimensions can vary according to the service consumer
profile, or also to the context in which the process consumer is operating.
Another direction for designing adaptive processes in the PAWS framework is
to extend the process description with variants, thus allowing to model different
process fragments which are alternatively selected according to the process and its
context of execution. An example is given in Fig. 5, where two different process
fragments are defined for different QoS levels.
While service design in PAWS is focused on providing adaptivity mechanisms,
it does not provide a way to define different strategies to achieve the process goals
through adaptation and to select among strategies at run time, as envisioned in [ 3] .
As seen above, all these approaches have the problem that there is a weak and
informal link between requirements expressed in the initial design phase and the
following process construction and service specification phases.
The intentional process presented in [ 10] has instead the advantage of linking
goals to processes and then to service design and construction. From the intention
maps, compositions of executable services are derived, as shown in Fig. 6.
On the right part of Fig. 6, at the bottom level executable services are represented.
The top level represents an agent with controls the achievement of the process goal,
while in intermediate levels agents are created to control compositions of services.
As mentioned above, the Maps of [ 10] do not represent fixed process schemas
defined in all details, but allow describing alternative paths and variants. This flexi-
bility can be exploited in operationalizing intentional services into software services.

Conf1: High QoS Conf2: Low QoS

Show
video

Show text

Low
QoS

High
QoS

Interactive
map

static
map

Fig. 5 Context-aware processes [7]