High Availability Service dengan Multiple Master pada Kubernetes Cluster Menggunakan Virtualisasi Kernel Based Virtual Machine (KVM)

Abstract

Servers have an important role in managing and serving requests from clients. So that high

availability is needed to handle risks in server use. Cluster technology is needed, so that the server load is

shared evenly. However, to run clustering, it requires the application of a method, namely the HAProxy load

balancer which works by evenly distributing the traffic load to the clustered physical servers. By implementing

Kernel-Based Virtual Machine (KVM) virtualization technology. In KVM virtualization there is a technology

used, namely Kubernetes. From the results of this final project, the results of the High Availability testing

carried out are known that the server performance with scenario 5 access experiments 20-500 users with stress

testing. Parameters measured are throughput and monitoring system usage of CPU, memory according to user

access load. Which is the highest value of Througput on parameter 500 clients worth 1330.6/ sec. The presence

of 2 servers provides performance test results. In standby, the lowest CPU usage and memory usage occurs on

the main server, which is 0.07% and 24% because there are no nodes running on top of the virtual machine. In

a busy state, the lowest CPU usage occurs on the main server is 1.06% due to the division of workload on

virtual machines. The lowest memory usage occurs on the main server server is 24% because there are no nodes

running on top of the virtual machine.