Abstract:
"
Cloud computing can be defined as a technology that provides services to external
users over the internet by utilizing the central remote servers. It has become a very
important and popular concept among many computer users mostly due to the
unlimited storage capacity and high performance. Scheduling the workflow tasks of
cloud computing has become a prominent issue and has been a very popular topic in
the research area mainly due to the impact it has on the performance of cloud
computing. Even though there are many researchers conducted in the area of workflow
scheduling, there is no proper solution built for workflow scheduling by thoroughly
investigating and considering the main objectives such as budget constraint, energy
constraint, makespan, load balancing, etc. The developed RL-based scheduling
algorithm will provide a near optimal solution for the scheduling problem by
considering the makespan and cost QoS requirements. The algorithm is developed
using machine learning and reinforcement learning. The algorithm contains a
prediction model to predict the execution time and cost of tasks which is used by the
reward function to calculate the reward for the selected action. The algorithm is tested
and evaluated with several domain experts and it is compared with other existing
algorithms. The results obtained through evaluation and benchmarking show that the
developed algorithm can give near optimal solutions for the workflow scheduling
problem in cloud computing."
