A Performance Comparison of Load Balancer Algorithm on Virtualized Environment: Case Study on Round Robin, Least Connection, and Weighted Least Connection Using Machine Learning = Perbandingan performa algoritma load balancer pada lingkungan eksekusi virtual: studi kasus dengan round robin, least connection, dan weighted least connection menggunakan machine learning

Tedja, Richard David (2021) A Performance Comparison of Load Balancer Algorithm on Virtualized Environment: Case Study on Round Robin, Least Connection, and Weighted Least Connection Using Machine Learning = Perbandingan performa algoritma load balancer pada lingkungan eksekusi virtual: studi kasus dengan round robin, least connection, dan weighted least connection menggunakan machine learning. Bachelor thesis, Universitas Pelita Harapan.

	Text (Title) Title.pdf Restricted to Repository staff only Available under License Creative Commons Attribution Non-commercial Share Alike. Download (1MB)
Preview	Text (Abstract) Abstract.pdf Available under License Creative Commons Attribution Non-commercial Share Alike. Download (307kB) | Preview
Preview	Text (ToC) ToC.pdf Available under License Creative Commons Attribution Non-commercial Share Alike. Download (1MB) | Preview
Preview	Text (Chapter1) Chapter1.pdf Available under License Creative Commons Attribution Non-commercial Share Alike. Download (419kB) | Preview
	Text (Chapter2) Chapter2.pdf Restricted to Registered users only Available under License Creative Commons Attribution Non-commercial Share Alike. Download (1MB)
	Text (Chapter3) Chapter3.pdf Restricted to Registered users only Available under License Creative Commons Attribution Non-commercial Share Alike. Download (1MB)
	Text (Chapter4) Chapter4.pdf Restricted to Registered users only Available under License Creative Commons Attribution Non-commercial Share Alike. Download (12MB)
	Text (Chapter5) Chapter5.pdf Restricted to Registered users only Available under License Creative Commons Attribution Non-commercial Share Alike. Download (1MB)
	Text (Chapter6) Chapter6.pdf Restricted to Registered users only Available under License Creative Commons Attribution Non-commercial Share Alike. Download (329kB)
Preview	Text (Bibliography) Bibliography.pdf Available under License Creative Commons Attribution Non-commercial Share Alike. Download (398kB) | Preview

Abstract

Cloud computing is a computational model which utilizes concepts of virtualization, processing power, storage, connectivity, and sharing to provide resources elastically and with an on-demand basis. The need for proper workload balancing arise as there are significant increase on data traffic volume and user demand for computational resources during peak time, which may cause workload imbalance or server overloading. A proposed solution to this problem is to implement a load balancer system with a load balancing algorithm that has the best performance based on incoming workloads, which in turn will be able to allocate incoming workload evenly to available nodes, so that there are no overloaded nor underloaded nodes. This research compared the performances of Round Robin, Least Connection, and Weighted Least Connection algorithms. To simulate real server traffic conditions, various levels of workloads are applied onto the load balancer using Apache JMeter. The performance results of each algorithm are captured and analyzed using methods of data analytics to determine which algorithm has the best performance under various simulated workloads. Analysis on the experiment results is done using a machine learning model comprising three machine learning algorithms, which are AdaBoost with an average AUC of 0.839, kNN with an average AUC of 0.802, and Random Forest with an average AUC of 0.791. Predictions to determine which load balancing has the best performance are made by sampling 10 instances from the test dataset. AdaBoost predicted WLC is the best load balancing algorithm on 80% of sampled instances. Random Forest predicted WLC is the best load balancing algorithm on 100% of sampled instances. kNN predicted WLC is the best load balancing algorithm on 80% of sampled instances. Based on the timestamp, amount of threads received, and session receive rate sampled in the test dataset, the analysis found that WLC had the best performance compared to RR and LC. / Komputasi awan adalah sebuah model komputasi yang mempergunakan konsep virtualisasi, kemampuan pemrosesan, kapasitas penyimpanan, konektivitas,dan pembagian sumber daya untuk menyediakan sumber daya komputasi secara elastis dan sesuai permintaan. Kebutuhan akan pembagian beban kerja yang memadai timbul ketika terdapat penambahan signifikan terhadap volume lalu-lintas data dan permintaan pengguna untuk sumber daya komputasi pada saat jam-jam sibuk, yang dapat mengakibatkan beban kerja yang tidak seimbang maupun beban berlebih pada server. Solusi yang ditawarkan terhadap permasalahan tersebut adalah mengimplementasikan sebuah sistem penyeimbang beban kerja (load balancer system) yang terkonfigurasi dengan algoritma load balancing yang memiliki performa terbaik berdasarkan karakteristik beban kerja yang diterima, sehingga beban kerja dapat terbagi secara merata dan tidak terdapat server yang mendapatkan beban berlebih ataupun kekurangan beban. Penelitian ini bertujuan untuk melakukan perbandingan terhadap performa tiga algoritma load balancer, yaitu Round Robin, Least Connection, dan Weighted Least Connection. Untuk menirukan kondisi lalu-lintas server yang umum terjadi, beban kerja dengan karakteristik yang beragam disimulasikan dan dikirim kepada load balancer menggunakan Apache JMeter. Hasil performa dari setiap algoritma diolah dan dianalisa menggunakan metode data analitik untuk menentukan algoritma yang memiliki performa terbaik. Analisa hasil eksperimen dilakukan menggunakan model machine learning yang terdiri atas tiga algoritma machine learning, yaitu AdaBoost dengan rata-rata AUC 0.839, kNN dengan rata-rata AUC 0.802, dan Random Forest dengan ratarata AUC 0.791. Prediksi dilakukan dengan mengambil sampel sebanyak 10 data dari test dataset. AdaBoost memprediksi WLC adalah algoritma dengan performa terbaik pada 80% sampel. Random Forest memprediksi WLC adalah algoritma dengan performa terbaik pada 100% sampel. kNN memperdiksi WLC adalah algoritma terbaik pada 80% sampel. Dengan demikian dapat disimpulkan berdasarkan waktu, jumlah threads yang diterima, dan laju penerimaan threads dari sampel data yang berasal dari test dataset, algoritma WLC memiliki performa yang lebih baik jika dibandingkan dengan RR dan LC.

Actions (login required)

View Item