Volume: 19 Issue-01 (January-June) 2024

---

CLUSTERING BASED VM MIGRATION AND BIDIRECTIONAL RECURRENT NEURAL NETWORK (BIRNN) FOR EDGE-CLOUD COMPUTING ENVIRONMENTS

ABSTRACT

Utilization of Internet of Things (IoT) gadgets and the information generated by these gadgets has grown significantly. It utilizes cloud and mobile-edge resources with ease. Because of the limited resource capacities, IoT flexibility aspects, resource diversity, network ranking, and problematic nature, it is difficult to schedule application jobs effectively in these kinds of contexts. The issue cannot be effectively solved by current heuristics and RL (Reinforcement Learning)-methods since they are not generalizable or quickly adaptable. This research introduces an innovative way to effectively utilize multi-layer resources in stochastic situations through dynamic task scheduling in the Edge-Cloud systems. Firstly, Fuzzy Election Based Optimization Algorithm (FEBOA) has been introduced for minimizing the metrics such as ART (Average Response Time), Average Energy Consumption (AEC), SLAV (Service Level Agreements Violations), and Average Migration Time (AMT). Secondly, Adaptive Threshold Fuzzy C-Means (ATFCM) clustering has been introduced for VM load detection. Features of host from RMS (Resource Monitoring Service) are utilized through Bidirectional Recurrent Neural Network (BiRNN) method to detect the next scheduling decisions. For forward and reverse input TS (Task Scheduling), BiRNN takes advantage of data to regulate connections. Each RNN cell's input (task) and output (host allocated to task) dependencies were defined for the BiRNN classifier in the forward as well as the reverse direction. Lastly, a scheduler using Asynchronous-Advantage-Actor-Critic (A3C) is shown, which is capable of taking into account all relevant task and host parameters in order to make task scheduling that would improve efficiency. Power usage, timing of responses, SLA, and operational charges are used to quantify the outcomes.

FULL TEXT

PDF