Information Technology Reference
Clearly, since human intervention is impossible for such a long-term process, soft-
ware mechanism is required.
Based on the above motivating example, we can see that to ensure the on-time
completion of a large-scale scientific cloud workflow is a challenging issue which
requires a systematic solution. Temporal QoS can be measured by either the on-time
completion rate or temporal violation rate. Clearly, the higher the on-time completion
rate (or the lower the temporal violation rate), the better the temporal QoS is. In addi-
tion, there are two general requirements for scientific cloud workflow temporal verifi-
cation which are automation and cost-effectiveness .
Automation : To speed up the execution, scientific workflow applications are designed
to be as automatic as possible, so should be the temporal verification strategies. Be-
sides, given such a large-scale workflow instances, human intervention is impossible
for efficient and effective monitoring and control of workflow execution. We must
rely on automatic software solutions.
Cost-effectiveness . In a cloud computing environment, every task for the workflow
temporal verification incurs some cost. Here, the cost is generally referred to both
monetary costs and time overheads. According to our observation, the total cost for
temporal verification can grow dramatically with the increase of the workflow size.
Therefore, we need to seriously consider cost effectiveness in every strategy design.
Clearly, given similar temporal QoS (e.g. on-time completion rate), the smaller the
total cost for temporal verification, the better the cost effectiveness is.
Basic Research Issues and State-of-the-Art Solutions
Based on the philosophy of “divide and conquer”, a generic workflow QoS frame-
work is proposed in  to provide a lifecycle QoS assurance in cloud workflow
systems. Based on that, we can define a generic temporal verification framework
which consists of four components including temporal constraint setting, temporal
checkpoint selection, temporal verification and temporal violation handling. Clearly,
how to design and implement these four components are the basic research issues.
Here, we will introduce each of the components with their basic requirements and
their representative and state-of-the-art solutions.
2.2.1 Research Issue #1: Temporal Constraints Setting
The first component of the framework is temporal constraint setting which deals with
the negotiation of overall deadlines (as a part of the workflow specification) between
customers and service providers, and the assignment of local temporal constraints to
the local workflow segments or workflow activities for monitoring purpose.
Temporal constraints mainly include three types, viz. upper bound, lower bound and
fixed-time . Among them, an upper bound constraint is a relative time value so
that the durations between two activities must be less than or equal to it. Upper bound
constraints are the most general type of temporal constraints where the others can be
transformed to its special cases. Therefore, upper bound constraints are the most
widely used type of temporal constraints in workflow temporal verification.