Dynamic Scaling of Call Stateful SIP Clusters in the Cloud

One of the key assets of cloud computing comprises the ability to build dynamically scaling systems. This enables adjusting on-the-fly the amount of resources (represted by VMs, OS containers, etc.) allocated to a clustered service. In this paper, we investigate the application of dynamic scaling in telecommunication services, focusing in particular on call stateful SIP servers. As a first contribution, we present and evaluate two protocols to transparently migrate ongoing sessions between call stateful SIP servers. This enables to instantaneously shutdown stateful SIP servers in response to a scale down request, excluding the need to wait until these servers' ongoing calls have finished. Additionally, to preserve the stringent carrier grade requirements inherent to telecommunication services in general and SIP in particular, we propose a self-adaptive Kalman filter to implement limited look-ahead call load predictions and combine this with history-based Kalman predictions to forecast call load variations. We believe that both techniques are key enablers to reduce the OPEX of a cloudified SIP service and to increase the resource utilization ratio of a telco cloud provider in a safe manner ­ that is, without breaking the associated SLAs.