Adaptive Cloud Bundle Provisioning and Multi-Workflow Scheduling Via Coalition Reinforcement Learning

Xiaogang Wang,Jian Cao,Rajkumar Buyya
DOI: https://doi.org/10.1109/tc.2022.3191733
IF: 3.183
2022-01-01
IEEE Transactions on Computers
Abstract:The efficient cloud resource provisioning for the execution of complex workflow applications has always been one of the important research issues. Most of the existing approaches focus on the resource provisioning of single-type virtual machine (VM) instances for the single or multiple workflows, while few consider the situation of provisioning multi-type VM instances simultaneously. As a result, the executing performance of complex workflows degrades. Different from the existing work, this paper proposes an adaptive cloud bundle provisioning and multi-workflow scheduling model to dynamically perform both the horizontal and vertical cloud resource scaling on multi-type VM instances for the execution of complex workflows. Among the model, a depth-first-search coalition reinforcement learning (DFSCRL) provisioning policy is presented to realize the resource scaling, which integrates the physical machine (PM) coalition formation with the Q-learning algorithm, then dynamically generates an optimal multi-type VM instance bundle from the PM coalition, and finally provisions these instances to the concurrent execution of multiple workflows. The theoretical proofs and various experiments with the multifaceted metrics demonstrate that the performance of the proposed algorithms is superior to that of the state-of-the-art relevant policies.
What problem does this paper attempt to address?