TY - GEN
T1 - STMS
T2 - 2018 USENIX Annual Technical Conference, USENIX ATC 2018
AU - Shi, Hang
AU - Hu, Yuming
AU - Cui, Yong
AU - Dai, Minglong
AU - Zheng, Kai
AU - Wang, Xin
AU - Wang, Fanzhao
N1 - Publisher Copyright:
© Proceedings of the 2018 USENIX Annual Technical Conference, USENIX ATC 2018. All rights reserved.
PY - 2018
Y1 - 2018
N2 - Using multiple interfaces on mobile devices to get high throughput is promising to improve the user experience. However, Multipath TCP (MPTCP), the de-facto standardized solution, suffers when different paths have heterogeneous quality. This problem is especially severe when the difference is the path latency. Our experimental results show that it causes the burst sending of packets from the fast path, which requires the in-network buffer to be big to achieve the full benefit of the bandwidth aggregation. In addition, it also requires bigger host buffer to fully utilize the fast path. To solve these problems, we propose and implement a new scheduler, which pre-allocates packets to send over the fast path for in-order arrival. Instead of relying on the estimation of network path condition, our scheduler dynamically adapts the MPTCP-level send window based on the packets acknowledged. Our evaluation shows that our scheduler can improve the throughput by 30% when the in-network buffer is limited, 15% when the host buffer is limited.
AB - Using multiple interfaces on mobile devices to get high throughput is promising to improve the user experience. However, Multipath TCP (MPTCP), the de-facto standardized solution, suffers when different paths have heterogeneous quality. This problem is especially severe when the difference is the path latency. Our experimental results show that it causes the burst sending of packets from the fast path, which requires the in-network buffer to be big to achieve the full benefit of the bandwidth aggregation. In addition, it also requires bigger host buffer to fully utilize the fast path. To solve these problems, we propose and implement a new scheduler, which pre-allocates packets to send over the fast path for in-order arrival. Instead of relying on the estimation of network path condition, our scheduler dynamically adapts the MPTCP-level send window based on the packets acknowledged. Our evaluation shows that our scheduler can improve the throughput by 30% when the in-network buffer is limited, 15% when the host buffer is limited.
UR - https://www.scopus.com/pages/publications/85077458192
M3 - Conference contribution
AN - SCOPUS:85077458192
T3 - Proceedings of the 2018 USENIX Annual Technical Conference, USENIX ATC 2018
SP - 719
EP - 730
BT - Proceedings of the 2018 USENIX Annual Technical Conference, USENIX ATC 2018
PB - USENIX Association
Y2 - 11 July 2018 through 13 July 2018
ER -