Machine-to-machine (M2M) traffic demands an increasing volume in 4G LTE networks. M2M traffic is characterized by a large number of very short messages whereas legacy 4G networks are designed for human users with larger data portions to be transported. Sending each M2M message separately over the network therefore results in an inefficient utilization of the network. In this project, methods of aggregating M2M messages are investigated so that the network utilization is improved while still keeping quality-of-service (QoS) requirements such as the delay.
Modern communication devices often have more than one interface, e.g. Ethernet, WLAN and LTE. Legacy TCP connections can only use one interface at at given time, whereas Multipath TCP allows to use multiple interfaces both at the server as well as the client side and thus deploying different paths for the same connection, transparent for the application. In this project, two major issues are being addressed. On the one hand, Multipath TCP connections should behave fairly against legacy connections, which in the first place requires comprehensive formal fairness definitions which then can be mapped to Multipath TCP's congestion control. Besides the investigations with simulations and physical testbeds, an analytical model describing different fairness methods is also developed. The second major issue of the project is the scheduling required to deliver data over the different MPTCP sub-connections in an optimum way so that QoS metrics such as delay or throughput can be optimised. Since TCP is an end-to-end protocol, this scheduler is located in the communication endpoints and has to take its decisions based on end-to-end QoS measurements, however as an option, assistance by the network, e.g. informing the endpoint how much load should be applied to which interface, may also be considered.