Analisis Kinerja Load Balancing Menggunakan Dragonfly Algorithm (DA) pada Cloud Environment

Camilla, Adiba Zalfa (2025) Analisis Kinerja Load Balancing Menggunakan Dragonfly Algorithm (DA) pada Cloud Environment. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5027211060-Undegraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (3MB) | Request a copy

Abstract

Perkembangan teknologi cloud computing yang pesat telah meningkatkan kebutuhan akan strategi load balancing yang efisien untuk memastikan distribusi beban kerja yang optimal di seluruh sumber daya komputasi. Load balancing menjadi aspek krusial untuk dapat meningkatkan kinerja aplikasi bebasis cloud dan menjaga performa serta responsivitas sistem secara keseluruhan. Untuk menjawab tantangan optimasi ini, penelitian ini berfokus pada Dragonfly Algorithm (DA), sebuah algoritma metaheuristik modern yang dipilih karena mekanisme uniknya dalam menyeimbangkan fase eksplorasi dan eksploitasi. Kinerja DA dianalisis dalam dua varian, yaitu Levy Flight dan Brownian Motion. Implementasi dilakukan pada simulator CloudSim menggunakan Java serta EclipseIDE dan juga pada real environment menggunakan Node.js dan Apache JMeter, serta dibandingkan dengan algoritma Particle Swarm Optimization (PSO) dan Genetic Algorithm (GA). Evaluasi dilakukan berdasarkan beberapa parameter, yaitu start time, wait time, finish time, execution time, makespan, dan imbalance degree untuk mengukur efektivitas load balancing. Hasil penelitian menunjukkan bahwa DA-Brownian merupakan algoritma paling unggul dalam konteks pemerataan beban kerja (imbalance degree) dan mencatat runtime yang lebih cepat dibanding DA-Levy pada simulator, dengan performa yang tetap kompetitif pada real environment. Secara keseluruhan, penelitian ini menyimpulkan bahwa DA-Brownian merupakan pilihan yang lebih tepat untuk meningkatkan keseimbangan beban antar virtual machine, karena mampu menghasilkan imbalance degree serendah 0,083 pada skenario beban kerja nyata (dataset SDSC) dengan waktu eksekusi simulasi yang hampir tiga kali lebih efisien dibandingkan DA-Levy. Sementara itu, DA-Levy tetap menunjukkan performa kompetitif pada beberapa parameter, namun memiliki runtime yang lebih tinggi dalam lingkungan simulasi Java, meskipun perbedaan tersebut tidak signifkan di real environment.
====================================================================================================================================
The rapid development of cloud computing technology has increased the need for efficient load balancing strategies to ensure optimal workload distribution across all computing resources. Load balancing is a crucial aspect for improving the performance of cloud-based applications and maintaining overall system performance and responsiveness. To address this optimization challenge, this study focuses on the Dragonfly Algorithm (DA), a modern metaheuristic algorithm chosen for its unique mechanism in balancing the exploration and exploitation phases. The performance of DA is analyzed in two variants, namely Levy Flight and Brownian Motion. The implementation was carried out in the CloudSim simulator (using Java and Eclipse IDE) as well as in a real environment using Node.js and Apache JMeter, and was compared with the Particle Swarm Optimization (PSO) and Genetic Algorithm (GA). Evaluation was conducted based on several parameters: makespan, start time, wait time, finish time, execution time, and imbalance degree. The results show that the DA-Brownian variant is the most superior algorithm in the context of workload distribution, capable of producing an imbalance degree as low as 0.083 with a simulation execution time that is nearly three times more efficient compared to DA-Levy. Meanwhile, the PSO and GA algorithms showed advantages in several speed-related parameters in the real environment, due to the higher computational overhead of DA. Overall, this study concludes that DA-Brownian is the most superior choice for the primary goal of balancing the workload among virtual machines. On the other hand, PSO and GA can be alternatives if the main priority is instantaneous response speed in a system with significant network latency.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	CloudSim, Dragonfly Algorithm (DA), EclipseIDE, JMeter, Komputasi Awan, Penyeimbangan Beban, Cloud computing, CloudSim, Dragonfly Algorithm (DA), EclipseIDE, JMeter, Load balancing
Subjects:	T Technology > T Technology (General) T Technology > T Technology (General) > T57.84 Heuristic algorithms. T Technology > T Technology (General) > T58.64 Information resources management
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Information Technology > 59201-(S1) Undergraduate Thesis
Depositing User:	Adiba Zalfa Camilla
Date Deposited:	23 Jul 2025 01:58
Last Modified:	23 Jul 2025 01:58
URI:	http://repository.its.ac.id/id/eprint/120662

Actions (login required)

View Item