HIGH AVAILABILITY MICROCLUSTERING SERVER USING KUBERNETES K3S AND HAPROXY ON RASPBERRY PI - Dalam bentuk buku karya ilmiah

ZUFAR ASYRAF AL - HAMID

Informasi Umum

Kode

25.05.649

Klasifikasi

000 - General Works

Jenis

Karya Ilmiah - Thesis (S2) - Reference

Subjek

Server

Dilihat

63 kali

Informasi Lainnya

Abstraksi

<br /> High Availability (HA) is a crucial requirement in distributed systems, especially for edge computing environments with limited hardware resources. This research proposes a microclustering server architecture using Raspberry Pi devices, Kubernetes K3s, and HAProxy to provide a cost effective and resilient HA solution. The system consists of two master and two worker nodes configured with a Virtual IP (VIP) failover mechanism and agent based health checks for adaptive load balancing. Two load balancing algorithms, Round Robin and Least Connection, were evaluated under three traffic distribution scenarios: constant, incremental, and imbalanced. Experimental results show that Round Robin consistently achieved higher throughput and lower response time under constant and incremental loads, while Least Connection significantly outperformed in imbalanced scenarios due to its adaptive connection routing. Statistical tests using T-test and Wilcoxon signed rank test confirmed the significance of these differences, particularly for throughput under imbalanced load (p < 0.01). The system demonstrated high reliability, with an average Mean Time Between Failures (MTBF) of 478.27 seconds and Mean Time To Recovery (MTTR) of 1.73 seconds, resulting in 99.64% system availability. These findings validate the feasibility of deploying high availability microclusters using lightweight orchestration tools and lowcost hardware, offering practical insights for resilient edge computing infrastructures.