220 likes | 317 Views
Olaf Christ, Thomas C. Schmidt, Matthias Wählisch christ_o@informatik.haw-hamburg.de {t.schmidt, waehlisch@ieee.org} HAW Hamburg & link-lab. Towards Seamless Handovers in SSM Source Mobility – An Evaluation of the Tree Morphing Protocol. Agenda. Mobile SSM Sources: What is the problem?
E N D
Olaf Christ, Thomas C. Schmidt, Matthias Wählisch christ_o@informatik.haw-hamburg.de {t.schmidt, waehlisch@ieee.org} HAW Hamburg & link-lab Towards Seamless Handovers in SSM Source Mobility – An Evaluation of the Tree Morphing Protocol
Agenda • Mobile SSM Sources: What is the problem? • Tree Morphing: Routing for mobile SSM Sources • Design of the Tree Morphing Protocol • Simulation & Evaluation • Conclusion & Outlook
Source Specific Multicast • Listeners subscribe to source-specific (S,G) channels • Typically used for real-time applications • WebTV / IPTV • VoIP / VCoIP • Collaborative applications • Massive Multiplayer Games • Immediate shortest path trees • Routing simplified (in contrast to ASM) • Easy to deploy, domain transparent
Problem: Mobile SSM Sources • Real-time constraints (50 – 100 ms) • SSM was designed for known, fixed sources • On source handover, the delivery tree rooting at the source invalidates • Address duality: Source filtering in routers and receivers • Logical Group Identifier: Home Address • Topological Tree Locator: Care-of Address • Decoupling: Source cannot Control Receiver Initiated Updates • May loose receivers on handover
Solutions • Statically Rooted Distribution Trees • Handover compliant to Mobile IPv6 • Packets are tunneled via the Home Agent • Additional undesired latencies • Single Source of Failure • Reconstruction of Distribution Trees • Separate multicast control tree with information about source address changes or • Bicasting data into an old and a new tree via anchor points (APs) • Tree Modification Schemes • Attempt to re-use established states
Multicast Forwarding States: Change of Distribution Trees under Mobility 75 – 95 % Coincidence for a mobility ‘step-size’ of 5 and100 Receivers
Tree Morphing:Routing for mobile SSM sources • Preserve previous trees: • Keep contact subsequent to handover • Idea: Morph previous into next tree: • Elongate root (modify RPF-Check) • Send packets to previous root of delivery tree • Discover shortcuts, but re-use common parts of trees • Dismiss unneeded branches • A new SPT is generated • Need to change routing • Extend (CoA,G) states to (CoA,G,HoA)
Design of the Tree Morphing Protocol • State update - necessary information • Group context (HoA, G) • Tree topology (nCoA, G) • “Piggy-backing” of update information • Eliminates additional update packets • Minimum extension to existing mobility messages • Re-use of existing headers (see next slides) • Security and robustness of updates
Tree Morphing: State Update Message • Combination of a Binding Update with CGA headers, a Router Alert Option and a Routing Header • Routing Header directs packets from nDR to pDR (source routing) • Router Alert Option instructs routers, to further inspect the packet (RFC 2711) • CGA authenticates these updates
Benefits of Tree Morphing Protocol • Signaling of updates by combining existing IPv6 headers • Router Alert Option is slight addition to existing Mobile IPv6 Binding Update • Packet processing is well-defined and already well tested • Inserting the update message into the data stream does not introduce additional packets
Simulation • OMNeT++ • IPv6Suite
First Step: Test Topologies Net 1 Net 2
Delay Stretch Net 1 Net 2
Second Step: Real-world TopologiesSCAN + Lucent (1.540 Core Routers)
Conclusion & Outlook • Benefits of Tree Morphing • Algorithm enables smooth source handover with state re-use • Protocol signaling realized as compact combination with Mobile IPv6 headers • Evaluation: Full protocol implementation on OMNeT++ • Test topologies reveal strengths and weaknesses • Real-world topologies smoothly mix effects • Packet loss too high • Current work and outlook • Protocol improvement: decouple signaling, eleminate source routing • Heals performance deficits (loss in particular) • Optimized versions for Fast MIPv6 & Multihoming
Do you have any questions? Thank you very much for your attention!