Controlling Overload in SIP Proxies: An Adaptive Window Based Approach Using No Explicit Feedback

The performance of SIP servers is largely degraded during overload conditions due to the built in message re-transmission mechanism of SIP. In this paper we propose a distributed and end-to-end adaptive window based overload control algorithm, by which upstream SIP servers control the amount of calls that are forwarded to a downstream SIP server in an attempt to prevent it from being overloaded. Our scheme uses the call establishment delay as a measure of the amount of load on the downstream server and unlike other methods does not rely on any explicit feedback from the downstream server. As a result the proposed algorithm imposes no additional complexity on the downstream server. Moreover, it requires no additional processing to be done at the downstream server, which is overloaded, making it a very robust approach. Using simulations we show that our proposed method achieves higher throughput than a commonly used overload control algorithm and is also fair among different upstream servers. It is also shown that compared with approaches using explicit feedback, the performance of our scheme is less sensitive to link delays. Furthermore, the effect of server queue size is also considered and a simple method for finding an upper bound on the queue size of the downstream server is given.