TY - JOUR
T1 - Multipaths and rate stability
AU - Liu, Junjie
AU - Guerin, Roch
N1 - Publisher Copyright:
© 2016 IEEE.
PY - 2016
Y1 - 2016
N2 - Multipath solutions have been shown to help improve throughput, reliability and/or load balancing. This paper seeks to understand if and when they benefit rate stability. Rate stability is important to many real-time, interactive applications, e.g., streaming video, but whether multipath solutions can help is unclear. Of relevance is the time-scale at which bandwidth changes are detected and acted upon to rebalance transmissions across paths. Consider two boundary cases: instantaneous detection and rate re-allocation, and a static rate assignment based on long-term path statistics. When transmissions can be instantaneously rebalanced across paths based on real-time link rate information, a multipath solution trivially improves rate stability (it all but eliminates rate variations). In contrast, when rate allocations are static, we find that multipath cannot improve upon the best single-path solution when buffers are large. When buffers are small (and coding is used to overcome losses), a multipath solution can, however, be beneficial even under a static rate allocation. The paper provides insight into when and how multipath solutions can help improve rate stability.
AB - Multipath solutions have been shown to help improve throughput, reliability and/or load balancing. This paper seeks to understand if and when they benefit rate stability. Rate stability is important to many real-time, interactive applications, e.g., streaming video, but whether multipath solutions can help is unclear. Of relevance is the time-scale at which bandwidth changes are detected and acted upon to rebalance transmissions across paths. Consider two boundary cases: instantaneous detection and rate re-allocation, and a static rate assignment based on long-term path statistics. When transmissions can be instantaneously rebalanced across paths based on real-time link rate information, a multipath solution trivially improves rate stability (it all but eliminates rate variations). In contrast, when rate allocations are static, we find that multipath cannot improve upon the best single-path solution when buffers are large. When buffers are small (and coding is used to overcome losses), a multipath solution can, however, be beneficial even under a static rate allocation. The paper provides insight into when and how multipath solutions can help improve rate stability.
UR - https://www.scopus.com/pages/publications/85015420347
U2 - 10.1109/GLOCOM.2016.7842013
DO - 10.1109/GLOCOM.2016.7842013
M3 - Conference article
AN - SCOPUS:85015420347
SN - 2334-0983
JO - Proceedings - IEEE Global Communications Conference, GLOBECOM
JF - Proceedings - IEEE Global Communications Conference, GLOBECOM
M1 - 7842013
T2 - 59th IEEE Global Communications Conference, GLOBECOM 2016
Y2 - 4 December 2016 through 8 December 2016
ER -