Latest revision as of 21:32, 3 February 2021

Abstract

To effectively manage the traffic distribution inside a network, traffic splitting is needed for load sharing over a set of equal-cost-multi-paths (ECMPs). In this paper, a new traffic splitting algorithm, called Table-based Hashing with Reassignments (THR), is proposed. Based on the load sharing statistics collected, THR selectively reassigns some active flows from the over-utilized paths to under-utilized paths. The reassignment process takes place in such a way that the packet out-of-order problem is minimized. As compared with the existing traffic splitting algorithms, THR provides close-to-optimal load balancing performance, less than 2% of packets arrived out-of- order, and a very small end-to-end packet delay performance. Although additional traffic monitoring function is needed by THR, we show that the extra complexity incurred is marginal.

Original document

The different versions of the original document can be found in:

• http://hub.hku.hk/bitstream/10722/46457/1/90910.pdf under the license https://creativecommons.org/licenses/by-nc-nd

• http://hub.hku.hk/bitstream/10722/46457/1/90910.pdf under the license https://creativecommons.org/licenses/by-nc-nd

• http://xplorestaging.ieee.org/ielx5/9179/29119/01312691.pdf?arnumber=1312691,

http://dx.doi.org/10.1109/icc.2004.1312691 under the license cc-by-nc-nd

• https://api.elsevier.com/content/article/PII:S1389128605000411?httpAccept=text/xml,

https://api.elsevier.com/content/article/PII:S1389128605000411?httpAccept=text/plain,

http://dx.doi.org/10.1016/j.comnet.2005.01.011 under the license cc-by-nc-nd

• https://dblp.uni-trier.de/db/conf/icc/icc2004.html#ChimY04a,

https://www.sciencedirect.com/science/article/pii/S1389128605000411,

http://hub.hku.hk/handle/10722/155286,

http://hub.hku.hk/bitstream/10722/46457/1/90910.pdf,

https://dl.acm.org/citation.cfm?id=1648781,

https://doi.org/10.1016/j.comnet.2005.01.011,

https://core.ac.uk/display/109449674,

https://academic.microsoft.com/#/detail/2137186894

under the license https://www.elsevier.com/tdm/userlicense/1.0/

DOIS: 10.1016/j.comnet.2005.01.011 10.1109/icc.2004.1312691