TY - GEN
T1 - The Behavior and Performance of TCP Congestion Control Algorithms under Multi-path Routing
AU - Sutawijava, Bayu
AU - Basuki, Achmad
N1 - Publisher Copyright:
© 2019 IEEE.
PY - 2019/9
Y1 - 2019/9
N2 - Multiple paths with equal-cost are commonly found on the Internet, which will trigger the multi-path routing to balance the traffic across those paths. On the other hand, TCP as the reliable transport protocol will treat those multiple paths of equal-cost similar to other single paths. This paper studies the behavior of several TCP Congestion Control Algorithms (e.g. Reno, CUBIC, BBR) and the impact of multi-path routing across multiple paths of equal-cost. The comparison analysis focuses on the link delay, loss rate, packet reordering. Based on the experiment results, we found TCP packet reordering occurs when multi-path routing is employed, but on the overall does not significantly degrade the measured throughput. On the link delay variations, the average RTT estimation on BBR is 20% lower than the average RTT estimation on Reno and CUBIC. While in loss rate variations from 1 to 10%, the average measured throughput on BBR is steadily achieve near the available bandwidth. On the contrary, Reno and CUBIC experience dropped of the average measured throughput up to 86% lower.
AB - Multiple paths with equal-cost are commonly found on the Internet, which will trigger the multi-path routing to balance the traffic across those paths. On the other hand, TCP as the reliable transport protocol will treat those multiple paths of equal-cost similar to other single paths. This paper studies the behavior of several TCP Congestion Control Algorithms (e.g. Reno, CUBIC, BBR) and the impact of multi-path routing across multiple paths of equal-cost. The comparison analysis focuses on the link delay, loss rate, packet reordering. Based on the experiment results, we found TCP packet reordering occurs when multi-path routing is employed, but on the overall does not significantly degrade the measured throughput. On the link delay variations, the average RTT estimation on BBR is 20% lower than the average RTT estimation on Reno and CUBIC. While in loss rate variations from 1 to 10%, the average measured throughput on BBR is steadily achieve near the available bandwidth. On the contrary, Reno and CUBIC experience dropped of the average measured throughput up to 86% lower.
KW - multi-path routing
KW - TCP performance
KW - traffic load balancing
UR - https://www.scopus.com/pages/publications/85080134326
U2 - 10.1109/SIET48054.2019.8986012
DO - 10.1109/SIET48054.2019.8986012
M3 - Conference contribution
AN - SCOPUS:85080134326
T3 - Proceedings of 2019 4th International Conference on Sustainable Information Engineering and Technology, SIET 2019
SP - 169
EP - 174
BT - Proceedings of 2019 4th International Conference on Sustainable Information Engineering and Technology, SIET 2019
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 4th International Conference on Sustainable Information Engineering and Technology, SIET 2019
Y2 - 28 September 2019 through 30 September 2019
ER -