Supporting SIP Port Mapping and RTP Affinity Constraints in Container Orchestration Environments

The use of container-oriented orchestration platforms such as Kubernetes, is becoming more popular due to their ability to hide from developers the deployment details of the application and the capability of these platforms to support autoscaling and failure recovery. Most container orchestration environments have focused on supporting the HTTP protocol and are challenging to use in protocols such as SIP and RTP. When SIP/RTP services are deployed in the Kubernetes environment, failures related to port mapping and affinity occur. These failures are reflected as packets sent to the non-existent ports in either the overlay or underlay network in Kubernetes. This paper details these failures and presents solutions for resolving the SIP/RTP port mapping and the RTP affinity constraint failures. The SIP/RTP port mapping failure is solved by correctly mapping the internal SIP/RTP ports to those exposed externally by Kubernetes while the RTP affinity constraint solution leverages the concept of the creation of unique endpoints and labels for each RTP service port opened for a SIP call. The paper also offers some insight into the challenges faced in replicating the SIP constraint affinity failure. These solutions have been implemented in a Kubernetes aware SIP Proxy (SIP-K8S-Proxy) that supports port mapping and managing of the RTP affinity as well. The overhead associated with implementing this solution has been computed to be about 2 seconds in the establishment of a SIP call.