Information Technology Reference
In-Depth Information
needed once only to verify all of them. Therefore, many times of unnecessary tempo-
ral verification are avoided.
2.2.3 Research Issue #3: Temporal Verification
As one of the major dimensions of software verification whose goal is to assure that
software fully satisfies all the expected requirements, temporal verification is to as-
sure that workflow applications satisfy temporal constraints. Specifically, according
to a temporal consistency model, temporal verification is to check whether the current
temporal consistency state of the runtime workflow application satisfies its build-time
specification. Temporal verification is usually conducted after a checkpoint is se-
lected. The original task of temporal verification has two steps. The first step is to
check the current temporal consistency state so as to determine whether a specific
type of temporal violation has occurred. The second step (after a temporal violation is
detected) is to calculate the time deficits (the time delays at the current checkpoint
given different temporal constraints) and other runtime information to facilitate the
temporal violation handling (i.e. a time deficit compensation process) of temporal
violations.
Traditionally, there are only binary states of consistency or inconsistency defined
by the temporal consistency model. However, as stated in [5], it argues that the con-
ventional consistency condition is too restrictive and covers several different states
which should be handled differently for the purpose of cost effectiveness. Therefore,
it divides conventional inconsistency into Weak Consistency (WC), Weak Inconsis-
tency (WI) and Strong Inconsistency (SI) and treats them accordingly. However,
multiple-state based temporal consistency model cannot support quantitative meas-
urement of temporal consistency states and lacks the ability to support statistical
analysis. Therefore in [22], a probability based build-time temporal consistency
model is presented to facilitate the setting of temporal constraints. Different from
previous discrete temporal consistency model such as binary-state or multiple-state,
the probability based temporal consistency model is a type of continuous-state model
where the temporal consistency state is measured using the probability confidence for
on-time completion such as 90% or 80%. The runtime version of the probability
based temporal consistency model is proposed in [33] to facilitate runtime temporal
checkpoint selection and verification.
Due to the significant closeness between temporal checkpoint selection and tempo-
ral verification, they are often referred as temporal consistency monitoring in many
studies [25, 26]. Therefore, the state-of-the-art temporal verification strategy can be
regarded as the combined strategy of the temporal dependency based checkpoint se-
lection strategy and the temporal verification strategy with the probability based tem-
poral consistency model, which is proposed in [28].
2.2.4 Research Issue #4: Temporal Violation Handling
If temporal violations are detected after temporal verification, temporal violation
handling strategies will be requested. The work in [52] introduces five types of work-
flow exceptions where temporal violations can be classified into deadline expiry.
However, it should be noted that what we detected and handled temporal violations
