Fine-grained Forward Prediction based Dynamic Packet Scheduling Mechanism for multipath TCP in lossy networks

Nowadays, multi-interface terminals in heterogeneous network may access Internet through various access technologies, and further aggregate network resources from multiple paths as much as possible. Multipath TCP exploits multiple paths simultaneously by stripping data of a connection over multiple TCP flows, each of which is through a disjoint path. However, it encounters the problem caused by the great number of out-of-order packets at receiver due to dissimilar path characteristics, i.e. latency, bandwidth, packet loss rate, etc. The previous intelligent scheduling mechanisms to keep in-order delivery all ignored packet losses and became fragile in lossy networks. In this paper, we present Fine-grained Forward Prediction based Dynamic Packet Scheduling Mechanism(F²P-DPS) for multipath TCP. It utilizes the idea of TCP modeling to estimate the latency on the path under scheduling and the data amount sent on the other paths simultaneously, which takes packet loss rate into consideration, and then decides which packets to send on the under-scheduling path. From the simulation, we can see that our mechanism obviously improves throughput and reduces cache occupancy at receiver in lossy networks.