Multiple Adaptive-Resource-Allocation Real-Time Supervisor (MARS) for Elastic IIoT Hybrid Cloud Services

In this paper, a Multiple Adaptive-resource-allocation Real-time Supervisor (MARS) scheme for hybrid cloud-assisted Industrial Internet of Things (IIoT) is proposed to support reliable cloud services even under rapidly changing service demands that occur in massive IIoT networks. Virtual Machines (VMs) in both private cloud and public clouds can be elastically and accurately allocated through the proposed MARS scheme, which uses Karush-Kuhn-Tucker (KKT) optimization applied to the VM Continuous-Time Markov Chain (CTMC) scheme. Because the MARS scheme can immediately determine the optimal number of VMs based on the hybrid cloud situation, a significant improvement in the elasticity performance can be obtained. Compared to using the CTMC scheme, the results show that the MARS scheme can improve the response time up to 19.3
$\sim$
73% (based on the activation rate) and the elasticity by 26.7%, and reduce the cost by 1.2%.