Multi-agent QoS-aware autonomic resource provisioning framework for elastic BPM in containerized multi-cloud environment

Cloud computing enables businesses to improve their market competitiveness, enabling instant and easy access to a pool of virtualized and distributed resources such as virtual machines (VM) and containers for executing their business operations efficiently. Though the cloud enables the deployment and management of business processes (BPs), it is challenging to deal with the enormous fluctuating resource demands and ensure smooth execution of business operations in containerized multi-cloud. Therefore, there is a need to ensure elastic provisioning of resources to tackle the over and under-provisioning problems and satisfy the objectives of cloud providers and end-users considering the quality of service (QoS) and service level agreement (SLA) constraints. In this article, an efficient multi-agent autonomic resource provisioning framework is proposed to ensure the effective execution of BPs in a containerized multi-cloud environment with guaranteed QoS. To improve the performance and ensure elastic resource provisioning, autonomic computing is utilized to monitor the resource usage and predict the future resource demands, then resources are scaled based on demand. Initially, the required resources for executing the incoming workloads are identified by clustering the workloads into CPU and I/O intensive, and the local agent achieves this with the help of an initialization algorithm and K-means clustering. Then, the analysis phase predicts the workload demand using the proposed enhanced deep stacked auto-encoder (EDSAE), further, the containers are scaled based on the prediction outcomes, finally, the multi-objective termite colony optimization (MOTCO) algorithm is used by the global agent to find suitable containers for executing the clustered workloads. The proposed framework has been implemented in the Container Cloudsim platform and evaluated using the business workload traces. The overall simulation results proved the effectiveness of the proposed approach compared to other approaches in terms of SLA violation rate, CPU utilization, response time, execution cost, energy consumption, make-span, and throughput. © 2022, The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature.