Harmonic DiffServ: Scalable support of IP multicast with Qos heterogeneity in DiffServ backbone networks

  • Authors:

  • Sheau-Ru Tong;Chun-Cheng Chang

  • Affiliations:

  • Department of Management Information Systems, National Pingtung University of Science and Technology, No.1, Hseuh-Fu Rd., Nei-Pu Hsiang, Pingtung 91201, Taiwan, ROC;Department of Management Information Systems, National Pingtung University of Science and Technology, No.1, Hseuh-Fu Rd., Nei-Pu Hsiang, Pingtung 91201, Taiwan, ROC

  • Venue:
  • Computer Communications
  • Year:
  • 2006

Quantified Score

Hi-index 0.24

Visualization

Abstract

The Differentiated Service (DiffServ) has been considered as a scalable approach to provide class-based quality-of-service (QoS) control in Internet. However, the direct implant of DiffServ in IP multicast shows some limitations on supporting a multicast session exhibiting QoS heterogeneity. Also IP multicast by itself has to maintain a large amount of the session-based forwarding information and potentially causes another routing scalability problem in routers. In this paper, we therefore propose a novel adaptation framework, termed harmonic DiffServ, to solve the QoS heterogeneity and routing scalability problems at the same time. In principle, (1) multicast trees are clustered and aggregated into a given number of multicast sessions, and (2) within each multicast session, packets are further marked by a set of undefined DSCPs which lead packets to specific QoS treatments in each router in a way that the respective QoS requirements of multicast trees are realized properly. We propose a heuristic clustering scheme and based on it, two DSCP encoding schemes, namely, the fixed encoding (FE) and the dynamic encoding (DE). Extensive simulations based on different network configurations are conducted. The simulated result reveals that our schemes can effectively reduce the forwarding information and achieve excellent scalability in terms of the number of multicast sessions and DSCP in use. And DE always consumes less number of DSCP than FE at the cost of slightly more computation and control overheads.