21.01.2014 Views

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

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

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

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

espectively. It is clearly that ACO requires much more CPU time than GA. The<br />

CPU time for ACO follows a steady increasing trend with the growing <strong>of</strong> number <strong>of</strong><br />

activities. GA also follows an increasing trend but in a much slower way. In an<br />

overall sense, the CPU time for both GA <strong>and</strong> ACO are all trivial compared with the<br />

execution time <strong>of</strong> scientific workflow activities which normally take minutes or<br />

hours.<br />

Figure 8.11 CPU Time<br />

To summarise, based on the results shown in Figure 8.8 <strong>and</strong> Figure 8.9, both GA<br />

<strong>and</strong> ACO can effectively optimise the total makespan <strong>and</strong> total cost <strong>of</strong> the entire<br />

integrated Task-Resource list. Furthermore, based on the results shown in Figure<br />

8.10(c) <strong>and</strong> Figure 8.10(d), we can claim that our two stage workflow local<br />

rescheduling strategy is capable <strong>of</strong> compensating the violated workflow segment as<br />

well as optimising the entire integrated Task-Resource list on both total makespan<br />

<strong>and</strong> total cost. Therefore, the requirements for rescheduling strategies for h<strong>and</strong>ling<br />

temporal violations as analysed in Section 8.1.2 are successfully met.<br />

Strategy Ranking based on User Preference<br />

The comprehensive evaluation results have been demonstrated above for both<br />

GA <strong>and</strong> ACO based rescheduling strategy to illustrate the effectiveness <strong>of</strong> our<br />

general two-stage workflow local rescheduling strategy. However, as for which<br />

strategy should be employed in the scientific workflow system to h<strong>and</strong>le temporal<br />

violations, a ranking function is required to make a decision based on user<br />

preference. The four basic measurements (including the optimisation ratio on total<br />

153

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!