AUTONOMOUS QOS-AWARE RESOURCE PROVISIONING IN DISTRIBUTED CLOUD ENVIRONMENTS: A UNIFIED PERSPECTIVE FROM ROBOTICS, EDGE, AND BPM DOMAINS

Abstract

Distributed computing has been a long existing technology that has allowed computer collaboration in terms of handling complex tasks with very high efficiency. The trend in distributed computing has witnessed some revolution in terms of the domains that use it as well as the methodologies in its implementation. Recently, some among the range of domains that utilizes distributed computing include robotics, mobile edge networks and business process management (BPM). Each domain may have a different use case but they commonly leverage on the increasingly growing intelligence in terms of resource provisioning. Some of the challenges systems using distributed computing have attempted to solve includes dealing with dynamic workloads while at the same time meeting certain constraints like Quality of Service (QoS) and Service Level Agreement (SLA). Traditionally robotics, mobile edge networks and BPM as distinct domains have differing use cases and cannot be autonomously utilized by systems that may require the use of all three domains. This research work focuses on synthesizing the potentials of the three key research areas, namely multi-agent cloud robotics, heuristic and learning-based edge computing, and BPM with the aim of obtaining an efficient resource allocation in a dynamic workload environment. The proposed framework combines the MAPE-K loop, EDSAE and MOTCO techniques through enhanced K-Means clustering. Comprehensive experiment demonstrates that after a 24-hour simulation using Extended Container CloudSim and BPM workload traces revealed that the proposed framework outperformed both static provisioning and reactive auto-scaling strategies. The proposed framework shows significant reduction in response time, energy consumption...