Analysis of Load Balancing Least Connection and Shortest Expected Delay Algorithm for Web Server Using Kube-Proxy on Kubernetes

ABSTRAKPenelitian ini membandingkan dan menganalisis kinerja algoritma load balancing shortest expected delay dan Least Connection untuk Web Server menggunakan Kube-Proxy di Kubernetes. Peningkatan jumlah pengguna dan perubahan jumlah node pekerja yang mempengaruhi kinerja system menjadi focus penelitian ini. Hasil penelitian menunjukkan bahwa pada parameter elapsed time, shortest expected delay dengan memiliki waktu yaitu 216.295 ms. Dalam waktu pemrosesan server, shortest expected delay lebih baik dengan menghasilkan 214.257ms. Throughput saat menggunakan algoritma shortest expected delay lebih besar, rata-rata 560.256 paket/detik. Algoritma Least Connection lebih baik dengan memiliki efisiensi 35,24% dalam hal penggunaan CPU, dibandingkan shortest expected delay. Meningkatkan klaster dari dua node pekerja menjadi empat node pekerja menghasilkan pengurangan waktu pemrosesan server yang signifikan, yang berarti penyeimbangan beban menggunakan 4 node pekerja lebih efektif.Kata kunci: komputasi awan, load balancing, Kubernetes, Web Server. ABSTRACTThis research compares and analyzes the performance of the load balancing algorithm Shortest Expected Delay and Least Connection for web server using Kube-Proxy on Kubernetes. The increasing number of users and changes in the number of worker nodes that affect system performance are the focus of this research. The results show that in elapsed time parameter, shortest expected delay has the time, which is 216.295ms. In server processing time, the shortest expected delay is better by producing 214.257ms. Throughput when using the shortest expected delay algorithm is greater, an average of 560,256 packets/second. The Least Connection algorithm is better with an efficiency of 35.24% in terms of CPU usage, compared to the shortest expected delay. Upgrading the cluster from two worker nodes to four worker nodes resulted in a significant reduction in server processing time, which meant more effective load balancing using 4 worker nodes.Keywords: cloud computing, load balancing, Kubernetes, Web Server