TY - JOUR
T1 - Efficient and Consistent Flow Update for Software Defined Networks
AU - Wu, Kun Ru
AU - Liang, Jia Ming
AU - Lee, Sheng Chieh
AU - Tseng, Yu-Chee
PY - 2018/3/1
Y1 - 2018/3/1
N2 - Software defined network (SDN) provides flexible and scalable routing by separating control plane and data plane. With centralized control, SDN has been widely used in traffic engineering, link failure recovery, and load balancing. This work considers the flow update problem, where a set of flows need to be migrated or rearranged due to change of network status. During flow update, efficiency and consistency are two main challenges. Efficiency refers to how fast these updates are completed, while consistency refers to prevention of blackholes, loops, and network congestions during updates. This paper proposes a scheme that maintains all these properties. It works in four phases. The first phase partitions flows into shorter routing segments to increase update parallelism. The second phase generates a global dependency graph of these segments to be updated. The third phase conducts actual updates and then adjusts dependency graphs accordingly. The last phase deals with deadlocks, if any, and then loops back to phase three if necessary. Through simulations, we validate that our scheme not only ensures freedom of blackholes, loops, congestions, and deadlocks during flow updates, but is also faster than existing schemes.
AB - Software defined network (SDN) provides flexible and scalable routing by separating control plane and data plane. With centralized control, SDN has been widely used in traffic engineering, link failure recovery, and load balancing. This work considers the flow update problem, where a set of flows need to be migrated or rearranged due to change of network status. During flow update, efficiency and consistency are two main challenges. Efficiency refers to how fast these updates are completed, while consistency refers to prevention of blackholes, loops, and network congestions during updates. This paper proposes a scheme that maintains all these properties. It works in four phases. The first phase partitions flows into shorter routing segments to increase update parallelism. The second phase generates a global dependency graph of these segments to be updated. The third phase conducts actual updates and then adjusts dependency graphs accordingly. The last phase deals with deadlocks, if any, and then loops back to phase three if necessary. Through simulations, we validate that our scheme not only ensures freedom of blackholes, loops, congestions, and deadlocks during flow updates, but is also faster than existing schemes.
KW - Computer network
KW - OpenFlow
KW - protocol
KW - routing
KW - SDN
KW - switching
UR - http://www.scopus.com/inward/record.url?scp=85043450279&partnerID=8YFLogxK
U2 - 10.1109/JSAC.2018.2815458
DO - 10.1109/JSAC.2018.2815458
M3 - Article
AN - SCOPUS:85043450279
VL - 36
SP - 411
EP - 421
JO - IEEE Journal on Selected Areas in Communications
JF - IEEE Journal on Selected Areas in Communications
SN - 0733-8716
IS - 3
ER -