Tetris: Near-optimal Scheduling for Multi-path Deadline-aware Transport Protocol

Abstract

Deadline-aware Transport Protocol (DTP) is a new QUIC-based transmission protocol that provides deliver-before-deadline service. Single-path DTP is not conducive to flow fairness and does not make full use of bandwidth. Compared with single-path DTP, the decision space and the solving difficulty of multipath DTP (MPDTP) scheduling are larger. In this paper, we propose a near-optimal scheduling algorithm Tetris for MPDTP. Tetris is a block scheduler based on stream characteristics at the transmission layer. We have verified its feasibility on the simulator under the deployed heterogeneous path of different network environment. The results show that our scheduling algorithm allows data blocks to be delivered before the delivery time as much as possible. The transmission completion time has been increased by 19.53% on average, and the transmission delay of all blocks have been reduced by 11.27%.