| Author Name | Affiliation | Postcode | | Swathi M | Department of Computer Science and Engineering, Anil Neerukonda Institute of Technology & Sciences, Vishakpatnam 530048, India | 530048 | | Krishna Sameera V | Department of Computer Science and Engineering, Anil Neerukonda Institute of Technology & Sciences, Vishakpatnam 530048, India | 530048 | | Satya Prakash K V S | Department of Information Technology, Gayatri Vidya Parishad College of Engineering, Vishakpatnam 530048, India | 530048 | | Kiran Kumar B* | School of Computing, SRM Institute of Science and Technology, Tiruchirappalli 621105, India | 621105 | | Prathyusha Dogga | Department of Mechanical Engineering, Vignan’s Institute of Information Technology, Visakhapatnam 530049, India | | | Mamatha Vayelapelli | Department of Computer Science and Engineering, ARKA Jain University, Jamshedpur 832108, India | |
|
| Abstract: |
| The multiplicity of the clouds improves the availability and elasticity; However, the load balancing is more complex when the resources are not balanced, the loads are not constant, and cross-providers failures occur. The present work suggests the initial dynamical load balancing in real-time and multi-cloud with the aim to optimize the resources of the system, reduce the latency (tail included), and enhance the fault tolerance. This paper presents an account of a hybrid algorithm which is a combination of Genetic Algorithm (GA) and Ant Colony Optimization (ACO). Although GA does not only consider the global search through candidate task-resource mappings, ACO concentrates on local search, using the pheromone search. This publication defines explicit parameter tuning and workflow to be used in evaluation. The hybrid GA-ACO was introduced and was compared to the older ones (Round-Robin, Least-Connections), adaptive methods and ML-based methods in the context of various multi-cloud cases (steady, spike, node-fail, regional-outage). The assessment was done using the real deployment-related KPIs: load distribution, mean/p99 latency, resource utilization, fault tolerance. The hybrid GA-ACO improved its load imbalance by 36%, mean response time by 48%, resource utilization by 37% and fault tolerance by 36%. It is also possible to maintain 150 ms requested binding per control window and less than 5 ms overhead of the request-path. The metrics rely on parameters and accuracy of metrics. Even though our hybrid solution will reduce imbalance and latency and increase utilization and fault tolerance, the usability and fault tolerance will be a priority in future efforts through extensive validation with real workloads. The proposed hybrid solution is a combination of GA and ACO and represents global and local search using parameter optimization and the outcome of a thorough consideration, offering a smooth, adaptive, and efficient way of load balancing between various clouds. |
| Key words: load balancing genetic algorithm ACO multi-cloud hybrid approach |
| DOI：10.11916/j.issn.1005-9113.2025059 |
| Clc Number:TP301.6 |
| Fund: |
