Abstract

Multipath transmission on overlay network is proved to effectively improve transmission quality for Internet applications requiring stringent QoS guarantees, e.g., real-time conversational applications. The orchestration of overlay nodes is the basis of overlay multipath transmission, which fundamentally impacts the management, control and transmission quality of overlay network. In this paper, we introduce SDN (Software Defined Networking) paradigm to orchestrate overlay nodes for optimal global management and intelligent forwarding control and we name the overlay multipath framework Software Defined Service Overlay Network (SDSON). To achieve efficient and reliable orchestration of nodes in SDSON, we focus on two major challenges, 1) the match between overlay and underlay networks for reducing the transmission quality degradation from the uncertainty of overlay paths, 2) the enhancement of scalability for eliminating control plane bottleneck of SDSON centralized structure. Correspondingly, we propose a scalable overlay orchestration structure. 1) coordinate-based global topology view of overlay nodes is employed to characterize and constrain relationships between overlay nodes. Coordinates are generated by using MDS-MAP to convert measured delays. And a balanced-binary-tree measure method is designed to balance measure loads. 2) The two-level distributed control plane is designed by decoupling control and management functions in control plane and together a novel collaborative control mechanism. The first level focuses on global management and the second concentrates on conducting global control with the collaborative control mechanism among distributed controllers. We use modified ISODATA with density-based initialization improvement to partition global topology view into distributed sub-domains, which are governed by controllers composing the second level control plane. Evaluations on OMNeT++ demonstrate that coordinate-based view matches well with practical underlay networks, sub-domains partition quality outperforms classical algorithms, and the scalable control planes with collaborative control mechanism alleviate control pressure and improve control performance efficiently.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call