Investigating performance optimization through balanced task scheduling
heuristics and DVFS in hybrid fog-cloud computing platforms
Abstract
In the evolving landscape of fog-cloud hybrid computing, efficient task
scheduling plays a vital role in meeting real-time requirements while
optimizing resource utilization and energy consumption. This work
proposes balanced minimum response time (MRT) and balanced minimum
energy consumption (MEC) heuristics to address the challenges of task
allocation on hybrid computing platforms in a balanced manner. These
heuristics are employed to assess the efficacy of complementing cloud or
fog networks, supported by dynamic voltage and frequency scaling (DVFS),
leveraging deadline laxities without compromising user satisfaction.
Three hybrid computing sub-scenarios are introduced: cloud-oriented,
fog-oriented, and balanced-hybrid, based on the relative compute
capacity of fog and cloud networks. The results reveal that a cloud
network complementing a fog network is more beneficial for achieving
lower makespan, while for lower energy consumption, it is preferable for
fog network to complement the cloud network. The research emphasizes
the advantages of balanced heuristics under different computing
scenarios to optimize makespan, energy consumption, and workload
allocation, leading to overall computing cost reduction using realistic
workloads. Further, judiciously applying DVFS to low-capacity fog nodes
is more fruitful than powerful cloud nodes due to extravagant idle
energy consumption in the generated scheduling holes.