Multi-objective Optimization for Resource Efficiency in Clustered Architectures

Authors: Kalesha Khan Pattan

DOI: https://doi.org/10.37082/IJIRMPS.v9.i4.232768

Short DOI: https://doi.org/g98f47

Country: Malaysia

Full-text Research PDF File:   View   |   Download

Abstract: Efficient utilization of computational resources has become a critical challenge in clustered architectures, particularly with the rapid expansion of heterogeneous workloads in modern distributed environments. Traditional resource management policies often focus on a single objective, such as minimizing latency or maximizing throughput, which leads to suboptimal trade-offs between performance, energy consumption, and cost. This paper presents a multi-objective optimization framework designed to enhance resource efficiency across diverse cluster configurations. The proposed approach simultaneously considers multiple, often conflicting, objectives—throughput maximization, latency minimization, and energy efficiency—to achieve a balanced and sustainable operational state. The framework integrates adaptive decision-making based on evolutionary and heuristic optimization techniques to dynamically allocate compute, memory, and network resources. It employs a Pareto-based optimization strategy to generate a set of non-dominated solutions, allowing system administrators to select operating points that best meet specific workload and business requirements. Throughput increased by an average of 30%, while latency and energy consumption were reduced by approximately 25–30%. These improvements were consistent across all cluster sizes, highlighting the robustness of the optimization strategy. The results validate that multi-objective optimization not only enhances overall resource utilization but also improves system adaptability to fluctuating workloads and changing infrastructure conditions. Furthermore, the framework supports modular integration with existing orchestration tools such as Kubernetes and Spark, enabling practical deployment in production-grade systems. In addition to quantitative results, this study discusses the impact of parameter tuning, convergence characteristics of the optimization algorithms, and the implications of Pareto front diversity on decision quality. The findings indicate that incorporating multi-objective intelligence into cluster management enables sustainable performance gains without additional hardware investment. The proposed framework provides a scalable, energy-aware, and cost-effective solution that can serve as a foundation for intelligent resource orchestration in future cloud-native and edge-computing ecosystems.

Keywords: Optimization, Clusters, Resources, Efficiency, Throughput, Latency, Scalability, Performance, Scheduling, Allocation, Energy, Cost, Evolutionary, Pareto, Adaptability.

Paper Id: 232768

Published On: 2021-08-06

Published In: Volume 9, Issue 4, July-August 2021

Share this