Database Reference
In-Depth Information
abstraction levels, supporting information systems, interfaces, etc., these are outside the
scope of this chapter.
In choosing the respective terminology, we aspired consistency with the standards of
the Workfl ow Management Coalition (Fischer, 2001).
Workfl ow Management
When taking the perspective of an organization, it immediately becomes clear that there
are many factors that determine—or at least infl uence—how activities should be defi ned. One
can think of regulations and considerations that emerge from human resource management,
ergonomics, quality management, social sciences, accountancy, and various other fi elds.
One specifi c view is that of workfl ow management (WfM), which produces relatively
pragmatic directions on activity defi nition, aimed at process automation. Sharp and McDer-
mott (2001), for example, allow on each of the three to fi ve hierarchic levels of a workfl ow
process a number of fi ve to seven activities. In other words, no matter what the operations
are, they should be fi tted in somewhere.
Van der Aalst and Van Hee (2002) note that to prevent problems in supporting processes
by WfM systems it is necessary to only regard as an activity a logical unit of work (LUW).
This means that the so-called ACID properties known from transaction processing apply:
Atomicity, Consistency, Isolation, and Durability (Harder & Reuter, 1983). The authors
also state their decomposition criterion for distinguishing a LUW: There must exist unity
of time, place, and operation.
The unity of time, place, and operation criterion often does not act as an imperative,
but rather as a fi rst step for activity decomposition, according to Van der Aalst and Van Hee
(2002). They give the following additional decomposition criteria:
•
The recognizability of an activity by the members of the organization that must perform
it is important, with a clear function and objective.
•
Sensible interim states: All resulting interim states in the process caused by activity
decomposition should be sensible.
•
Acceptable “commit work” for each of the process activities: Violations of the ACID
properties should be acceptable, especially with respect to possible rollback and the
split up of tasks.
Van der Aalst and Van Hee (2002) admit that in practice it is not easy to address the
ACID properties, because of the properties and limitations of current workfl ow systems. A
partial solution to this problem is given by Grefen et al. (2001), who distinguish high-level
(long-living) and low-level (relatively short-living) processes. The latter are subprocesses of
the former, but both have different requirements. The low-level requires strict execution of
the ACID properties. The high-level needs relaxation of the atomicity and isolation require-
ments. Their WIDE model for WfMS's supports this view (Grefen et al., 1999).
In conclusion, in the WfM fi eld heuristics and rules of thumb are used to identify op-
erations that more or less naturally belong together. Characteristically, decisions on splitting
up or combining activities are context-sensitive. All the given rules provide a considerable
degree of design freedom.
