Enhanced Socket API for MPTCP - Controlling Sub-flow Priority

TL;DR

This paper introduces new socket APIs and modifications to MPTCP to control sub-flow priorities and scheduling, enhancing its usability for UAVs and similar applications by enabling application-specific control and persistent sub-flow management.

Contribution

The paper presents novel socket APIs and a sub-flow scheduler for MPTCP, allowing application-level control and persistent sub-flow priorities, especially useful for UAVs.

Findings

APIs enable fine-grained control over MPTCP sub-flows.

Proposed scheduler segregates data based on QoS requirements.

Validated through Mininet and Raspberry Pi testbed experiments.

Abstract

Multipath TCP (MPTCP) can exploit multiple available interfaces at the end devices by establishing concurrent multiple connections between source and destination. MPTCP is a drop-in replacement for TCP and this makes it an attractive choice for various applications. In recent times, MPTCP is finding its way into newer devices such as robots and Unmanned Aerial Vehicles (UAVs). However, its usability is often restricted due to unavailability of suitable socket APIs to control its behaviour at the application layer. In this paper, we have introduced several socket APIs to control the sub-flow properties of MPTCP at the application layer. We have proposed a modification in MPTCP kernel data-structure to make the sub-flow priority persistent across sub-flow failures. We have also presented Primary Path only Scheduler (PPoS), a novel sub-flow scheduler, for UAVs and similar applications/devices where it is necessary to segregate data on different links based upon type of data or Quality of Service (QoS) requirements. We have also introduced the socket APIs for providing the fine grained control over the behaviour of PPoS for particular application(s) rather than changing the behaviour system wide. The scheduler and the socket APIs are extensively tested in Mininet based emulation environment as well as on real Raspberry Pi based testbed.