SCTP SmartSwitch

WP1 targeted smart session management in 3GPP core IP networks, focusing on a more efficient startup on the alternate path after a switchover. A more efficient startup may also improve vertical handover performance in wireless access networks. In order to establish what gains are achievable through an improved startup scheme, a theoretical model was developed that shows how the message arrival time on the switchover target path depends on slow start. The model was validated in a series of experiments in the Emulab network testbed. On the basis of the model, it was found that slow start can, in cases with long round-trip times and relatively intense traffic, have a significant impact on the message transfer times during a switchover. An increase of the initial congestion window was therefore studied as a method to shorten the delay caused by slow start. Experiments conducted in a network testbed suggested that the transfer times of messages during a switchover were significantly shortened by an increased initial congestion window, and this with only marginal effects on competing traffic. Still, our experiments also indicated that increasing the initial congestion window would not completely solve the problem. The concurrent use of multiple paths, i.e., CMT-SCTP, was investigated as a further possibility to decrease the delay caused by slow start for signaling applications. The initial results indicate that at least for paths with similar delay and packet-loss characteristics, CMT-SCTP could indeed decrease the delay. A follow-up project has been defined to further study the use of CMT- SCTP.

WP2 targeted smart session management in wireless access networks. It aimed to design a session layer that treats handover/failover in an integrated fashion and provides smart and quick reactions to various mobility events. As a first step, an SCTP-based mobility management framework was designed and implemented in Ubuntu Linux. A key component of the mobility framework is the so-called Mobility Manager. The Mobility Manager reacts to various network interface events, e.g., interface failures and addition/removal of interfaces, and enables a swift failover between available network interfaces. To demonstrate the ability of the mobility management framework to provide for seamless failover, a modified OpenSSH client-server application was built. In a series of experiments, we tunneled HTTP audio and video inside an SSH connection and showed the ability of the tunnel to survive network interface failures. The mobility management framework was next extended with a session layer to form a complete session management framework. The session layer preserves communication by utilizing SCTP's multi-homing and dynamic address reconfiguration features and suspends and resumes communication when a mobile host experiences network disconnections of arbitrary length. Furthermore, it offers mobility-on-demand, by enabling an application to actively request suspension and resumption of communication. The session layer has been validated through a
range of tests establishing the basic functionality as well as the performance impact of various
protocol parameters.

In WP3, the outcomes of WP1 and WP2 were merged into a complete prototype of a smart session management framework running in Android. The work in WP3 thus involved porting the mobility architecture as well as the session layer to this platform. The different parts of the framework were evaluated in a series of real-world experiments involving both file transfers and video streaming over 3G and WLAN. As a result of the evaluations, a number of optimizations were introduced to both the basic mobility management architecture and the session layer. Vertical handovers from WLAN to 3G that occurred as a result of WLAN connectivity loss suffered from poor performance due to the warm-up required on the 3G interface. In order to resolve this issue, the framework was complemented with a lightweight vertical handover scheme. The vertical handover scheme makes handover decisions on the basis of the received signal strength indicator (RSSI). The performed real-world experiments suggest that the RSSI-based handover scheme provides for seamless handover between WiFi and 3G networks for soft real-time applications, such as live streaming video, at walking speed and, significantly improves handover performance also at faster speeds. The implementation of how the session layer resumes functionality after a network initiated disconnections was also optimized.

Publications within the project
● J. Eklund, K-J Grinnemo, and Anna Brunstrom. In Proceedings Second International Workshop on Protocols and Applications with Multi-homing Support (PAMS). Fukuoka, Japan. March 2012.
● P. Söderman, K-J Grinnemo, G. Cheimonidis, Y. Ismailov, and Y. Ismailov. "An SCTP-based Mobility Management Framework for Smartphones and Tablets". In Proceedings Second International Workshop on Protocols and Applications with Multi-homing Support (PAMS). Fukuoka, Japan. March 2012.
● J. Eklund, K-J Grinnemo, Anna Brunstrom, G. Cheimonidis, and Y. Ismailov. "Impact of Slow Start on SCTP Handover Performance". In Proceedings First Workshop on Flexibility in Broadband Wireless Access Network (FlexBWAN). Maui, Hawaii, U.S. July 2011.
● J. Eklund, K-J Grinnemo, Anna Brunstrom, G. Cheimonidis, and Y. Ismailov. "Impact of Slow Start on Real-Time Traffic during a SCTP Handover". In Proceedings Seventh Swedish National Computer Networking Workshop (SNCNW). Linköping, Sweden. June 2011.
● J. Eklund, K-J Grinnemo, and A. Brunstrom. "Theoretical Analysis of an Ideal Startup Scheme in Multihomed SCTP". In Proceedings Network Services and Applications - Engineering, Control and Management (EUNICE). Trondheim, Norway, June 2010.