Intelligent Federated Cloud Scheduling with Transfer Learning and Hybrid Reinforcement-Meta Heuristic Optimization

Abstract: Cloud computing has transformed the sharing of
the resources and service delivery but effective scheduling of tasks across federated multi-clouds environment is a difficult task because of the heterogeneous nature of resources, dynamic nature of workloads, and SLA constraints. Current solutions tend to be either local or global at the cost of other. This paper presents a proposal of an Intelligent Federated Cloud Scheduling model that combines transfer learning based on predictive workload modeling and a hybrid reinforcement-meta-heuristic
optimization of dynamic task allocation. The framework was
tested on the Google Cluster Trace dataset, where it was
compared to five state-of-the-art algorithms, with a decrease in
makespan of 15 percent, energy usage reduced by 12 percent,
SLA violation rate of 2.5 percent and resource utilization
Praveen B M
Institute of Engineering and Technology
Srinivas University
Karnataka,574146,India
bm.praveen@yahoo.co.in
energy, task late completion and breach of SLA, which
directly affect the operational performance and cost
efficiency[2].
Classical
efficiency of 92 percent. The findings indicate the scalability,
strength and viability of the framework to process dynamic and
heterogeneous workloads. This paper presents the promise of
predictive AI with hybrid optimization on intelligent cloud
management, which provides a general understanding of
practical research on automated, energy-efficient, and SLA
compliant federated cloud scheduling.