
Checkpointing in Workflow Optimization 121
engine may declare that the service has failed. In this case, without additional
knowledge about the node on which the service started N
ij
, the engine would
need to start a workflow rescheduling process as if the service was unavailable.
However, the service might still be running but may actually be slowed down
by additional load on node N
ij
. This scenario is shown in Figure 4.15.
FIGURE 4.15: Starting an alternative service after a timeout from an ex-
pected service termination time.
It is possible to optimize such a scenario by getting more knowledge about
the environment in which the service runs. In [57] ...