Resource Management using Multi-Scale Attention Convolutional Neural Networks in Containerized Cross-cloud Multi-cloud Environment

Abstract

Cloud computing gives businesses rapid and easy access to a pool of dispersed, virtualized resources, likes virtual machines (VM) and containers, making them more competitive in the marketplace. They could also manage their business more skilfully. Even while the cloud makes it simple to install and manage business processes, it may be difficult to manage the highly variable resource requirements and guarantee the seamless execution of business activities in containerized multi-cloud scenarios. Hence, elastic resource provisioning is required to meet the demands of cloud providers, end users, manage over- and under-provisioning issues, and take QoS constraints, service level agreements (SLA) into account. In this manuscript, Resource management using Multi-Scale Attention Convolutional Neural Networks in containerized Cross-cloud multi-cloud environment (REMT-MCNN-CCMCE) is proposed for effective execution of business processes (BPs) in containerized multi cloud environment with guaranteed quality of service. The data for proposed REMT-MCNN-CCMCE method collected from GWA-T-12 bitbrains. Therefore, it is necessary to pre-process the workloads before considering them for more processing. Here workloads are filtered for abnormal, noisy interruptions with support of Regularized bias-aware ensemble Kalman filter (r-EnKF). This pre-processed data transferred to Self-Adaptative Multi-Kernel Clustering (SAMKC) for used to enable effectual scheduling of cloud workloads. In this work resource management is completed by using the Multi-scale Attention Convolutional Neural Networks (MACNN). This MACNN helps effective execution of BP's in containerized multi cloud environment with guaranteed QoS. The Sheep Flock Optimization Algorithm (SFOA) is utilized to suitable containers selection for dynamic cloud workloads. It enhances the performance of MACNN by optimizing its parameters, thereby increasing the effective execution of BP's in containerized multi cloud environment with guaranteed QoS. The proposed REMT-MCNN-CCMCE framework has been executed in Container Cloudsim platform with assessed utilizing performances likes SLA violation rate, CPU utilization, Response time, Execution cost, Energy consumption, Make‑span, Throughput. The REMT-MCNN-CCMCE method achieves 31.89%, 25.45% and 19.32% lower SLA violation rate, 32.12%, 23.49% and 30.94% higher CPU utilization and 26.87%, 34.65% and 23.94% lower response time when compared with the existing method’s such as Multi-agent QoS-aware autonomic resource provisioning framework for elastic BPM in containerized multi-cloud environment (MAQ-ARP-EBPM-CMCE), Multi-cloud service provision depend on decision tree with two-layer Restricted Monte Carlo Tree Search (MCS-DTLR-MCTS) and Orchestration in Cloud-to-Things compute continuum: taxonomy, survey, future directions (OCT-CC-TSFD) respectively.