Container-Based VoIP Failover Design on Kubernetes Overlay Networks

This paper studied how to utilize a popular container scheduling orchestrator - Kubernetes (K8s) - in today's cloud computing scenario to support the Voice over IP (VoIP) applications. In VoIP, two major protocols - Session Initiation Protocol (SIP) and Real-time Transport Protocol (RTP) - usually require network traffic to obey the following pattern: the packets from both end devices must be directed to the same server during a call. However, in a K8s network environment, the automatic LoadBalancer mechanism only ensures that traffic is directed to a healthy Pod but does not consider the need for both sides of the session to be handled by the same server. Therefore, after VoIP is introduced into the container, some subtle design must be proposed to integrate SIP and RTP into the existing K8s network architecture to handle two-way communication and failover so that the above problem can be properly addressed. In this paper, we use the K8s Label selector, K8s leader-election, and a few open-source software to solve the problem of failover, and measure the pressure of failover on K8s and the underlying operating system.