Xiao Liu PhD Thesis.pdf - Faculty of Information and Communication ...

fine-grained temporal constraints (the overrun of expected individual activity
durations) may often take place. Therefore, temporal consistency states need to be
kept under constant monitoring to detect and handle potential violations in a timely
fashion. For example, during the second step of data pre-processing, delays may
occur in the activity of De-dispersion which needs to process terabytes of data and
consumes more than 13 hours computation time. Here, for example, we assume that
the De-dispersion activity takes 14.5 hours (a delay of 90 minutes, i.e. around 10%
over the mean duration), then given U ( SW1
) of 15.25 hours, there would only be 45
minutes left for the Accelerate activity which normally needs 1.5 hours. Another
example is that during the third step of pulsar seeking, we assume that the overall
duration for the FFT Seek, Get Candidates and Eliminate Candidates activities is
108 minutes (a delay of 18 minutes, i.e. around 20% over the mean duration). In
such a case, given U ( SW2)
being 5.75 hours, there will probably be a 3 minutes
delay if the subsequent activity completes on time. In both examples, time delays
occurred and potential temporal violations may take place.
Therefore, based on the above two examples, we can see that monitoring of
temporal consistency is very important for the detection of temporal violations.
Effective and efficient strategies are required to monitor temporal consistency of
scientific cloud workflow execution and detect potential violations as early as
possible before they become real major overruns.
3) Handling temporal violations. If temporal violations like the two examples
mentioned above are detected, temporal violation handling strategies are normally
required. However, it can be seen that the temporal violations in the two examples
are very different. In the first example with U ( SW1
) , even if we expect the
Accelerate activity can be finished 10 minutes less than its mean duration (around
10% less than the mean duration), i.e. finished in 80 minutes, there would still be a
35 minute time deficit. Therefore, under such a situation, at that stage, some
temporal violation handling strategies should be executed to decrease the duration of
the Accelerate activity to at most 45 minutes by, for instance, rescheduling the Taskto-Resource
assignment or recruiting additional resources in order to maintain
temporal correctness. Cleary, in such a case, a potential temporal violation is
detected where temporal violation handling is necessary. As in the second example
with U ( SW2)
, the 3 minutes time deficit is a small fraction compared with the mean
duration of 4 hours for the Fold to XML activity. Actually, there is a probability that
it can be automatically compensated for since it only requires the Fold to XML
activity to be finished 1.25% shorter than its mean duration. Therefore, in such a
situation, though a temporal violation is detected, temporal violation handling may
6