The effects of distribution algorithms and congestion avoidance techniques on application response times in the IMS architecture

The IP Multimedia Subsystem is a generic architectural approach to offering convergence for IP multimedia services. Components of the IMS architecture select a server from a pool of servers that will satisfy a particular request. In the signaling plane, the I-CSCF will choose the S-CSCF; in the data plane, the S-CSCF will choose the Application Server; the above servers will be assigned to the user for the entire session. It is important for the quality of experience offered to the user to have the best S-CSCF selected for signaling and the best Application Servers for the data plane.;Optimization of end-user QOS is treated herein as a function of load distribution across multiple servers. Quantitative comparisons have been made of the effects of load distribution algorithms on application response times. In depth analyses of the most common distribution algorithms have been performed for both stateless and stateful session treatments. A customized simulation platform has been built to study the effects of the various load distribution algorithms on application response times.;In addition, a mathematical model was built to validate that the simulation results were accurate; the results confirmed the validity of the simulation platform. A new adaptive algorithm is developed that provides the shortest application response time by selecting one of the existing algorithms previously discussed; this selection is triggered based on traffic load, network topology, session length and the distribution of the delays.;The effect of the congestion avoidance algorithms that affect the application's response time is also analyzed when the network is congested. A new adaptive algorithm is developed that minimizes the application response and the rejection rate by selecting one of the existing algorithms previously discussed; selection is triggered based on load limits, call gapping interval, traffic load, network topology, session length and the distribution of the delays.;Finally, a simple protocol is designed to convey the delay information to the I-CSCF from the S-CSCF and P-CSCF. It includes both pushes from the P-CSCFs/S-CSCFs and periodical and/or on-demand pulls from I-CSCF. SIP extra-headers are used to convey the delay information.