Experiment 4: EMSC - Githubissues

ai-se / RIOT

IEEE Cloud18: RIOT: a Novel Stochastic Method for Rapidly Configuring Cloud-Based Workflows

MIT License

1 stars 0 forks source link

Experiment 4: EMSC #8

Closed ginfung closed 7 years ago

ginfung commented 7 years ago

EMSC is a state-of-the-art meta-heuristic algorithm in solving workflow tradeoff balancing (cost & makespan).

What happens right now?

According to Zhu'16 paper, I also need to add some special initial candidates into 0th generation population.
Runtime .... terrible.... >20 hours in large models. [still running in hpc]

Some results inspiral_30 inspiral_50 inspiral_100

Next steps

Add special seed into 0th population
Reduce the precision in CloudSim (otherwise, execution time is unacceptable)
Compare to another state-of-the-art heuristic algorithm (MOHEFT)

ginfung commented 7 years ago

MOEA/D is not applicable due to the problem coding in EMSC

ginfung commented 7 years ago

cybershake_30 cybershake_50 cybershake_100 cybershake_1000

ginfung commented 7 years ago

epigenomics_24 epigenomics_46 epigenomics_100 epigenomics_997

ginfung commented 7 years ago

inspiral_30 inspiral_50 inspiral_100 inspiral_1000

ginfung commented 7 years ago

montage_25 montage_50 montage_100 montage_1000

ginfung commented 7 years ago

sipht_30 sipht_60 sipht_100 sipht_1000

ginfung commented 7 years ago

Results moved to #9