Information Technology Reference
In-Depth Information
setting strategy which consists of a win-win negotiation process between customers
and service providers for the overall deadline, and an automatic propagation process
for local deadlines, has been proposed.
The state-of-the-art solution for workflow temporal constraint management is pro-
posed in [24]. In most cases, after the setting of temporal constraints at workflow
build-time stage, temporal constraints as a type of QoS requirements cannot be
changed. However, in this paper, the authors suggested that local temporal constraints
can be dynamically relaxed or strengthened according to the runtime workflow execu-
tion states. Accordingly, strategies for probability time deficit/redundancy propaga-
tion process for activities on both critical and non-critical paths have been proposed.
2.2.2 Research Issue #2: Temporal Checkpoint Selection
Temporal checkpoints are the decision points for further actions such as temporal
verification and temporal violation handling in the temporal verification framework.
Therefore, checkpoint selection plays a critical role in the whole temporal verification
framework since the number of selected checkpoints basically determines the possible
number of times for temporal verification and temporal violation handling. Since it is
normally too expensive to conduct temporal verification at every activity point but
insufficient to conduct temporal verification only at a few pre-defined activity points,
the target of a checkpoint selection strategy (CSS) is to select only essential activity
points where potential temporal violations can be detected. In such a case, the cost of
temporal verification and possible temporal violation handling can be minimized.
In recent years, many checkpoint selection strategies, from intuitive rule based to
sophisticated model based, have been proposed. The work in [17] takes every
workflow activity as a checkpoint. The work in [35] selects the start activity as a
checkpoint and adds a new checkpoint after each decision activity is executed. It also
mentions a type of static activity point which is defined by users at the build-time
stage. The work in [9] selects an activity as a checkpoint if its execution time exceeds
the maximum duration while the work in [10] selects an activity as a checkpoint if its
execution time exceeds the mean duration. The first checkpoint selection strategy
which satisfies the property of necessity and sufficiency (which is regarded as the
benchmark for temporal checkpoint selection) is proposed in [6] where minimum time
redundancies for SC (Strong Consistency) and WC (Weak Consistency) are defined.
Here, necessity means that only those activity points where intermediate temporal
violations take place are selected and sufficiency means that there are no any omitted
activity points. For example, an activity point is selected as a WC checkpoint if and
only if its execution time is larger than the sum of its mean duration and its minimum
WC time redundancy. The comparison result shows that with the measurement of
necessity and sufficiency, the one based on minimum time redundancy has outper-
formed all the other checkpoint selection strategies.
The state-of-the-art checkpoint selection strategy is proposed in [7] where the au-
thors investigate the temporal dependency between different upper bound temporal
constraints. The results have shown that at any specific checkpoint, for those temporal
constraints which cover the checkpoint and have the same type of temporal depend-
ency, such as SC (Strong Consistency) temporal dependency, temporal verification is
Search WWH ::




Custom Search