Mobile Group Communication
Mobility & Multicast Status
The Multicast Mobility Problem
Listener Mobility
ASM Source Mobility
SSM Source Mobility
Conclusions & Outlook
Where are we going Mobile today?
2 Prof. Dr. Thomas Schmidt http:/www.informatik.haw-hamburg.de/~schmidt
We need Multicast Services for …
Public Content Broadcasting
Voice and Video Conferencing
Collaborative Environments
Gaming
Self Organisation of Distributed Systems
…
All of this MOBILE!
Status: Where are we today?
o Internet Mobility
- Mobile IPv6 - RFC 3775 (June 2004, implemented)
o Real-Time Mobility
- FMIPv6, HMIPv6 – RFCs 4068, 4140 (August 2005, first implementations)
o Multicast Mobility
- Rough ideas in IETF … - Some proposals in IRTF
- First problem statement in Vancouver, Nov. 2005 o NGNs: DVB IP distribution services …
4 Prof. Dr. Thomas Schmidt http:/www.informatik.haw-hamburg.de/~schmidt
Multicast:
What is the Problem?
o There is Any Source Multicast:
- Listeners harvest streams from any source - Sparse & dense mode routing
- Complex to deploy, Interdomain problem o There is Source Specific Multicast:
- Listeners subscribe to known sources - Immediate shortest path trees
- Easy to deploy, domain transparent
Multicast Receiver Mobility
6 Prof. Dr. Thomas Schmidt http:/www.informatik.haw-hamburg.de/~schmidt
o Multicast Routing is ‘Mobile’ :
- On handover listeners may re-subscribe to multicast group (ASM & SSM)
- Routing will construct delivery branch
o Problem: Multicast routing not seamless, but slow - Branch construction up to 30 s
- Real-time communication requirements:
handover below 100 ms!
Multicast Mobility Approaches
o Remote Subscription
- Show all movement by local multicast subscription
o Bi-directional Tunneling
- Hide all movement by tunneling via Home Agent
o Agent Based
- Compromise: Intermediate agents shield Mobile - Approaches: Extend unicast acceleration schemes
M-FMIPv6, M-HMIPv6, …
Fast Multicast Protocol for MIPv6
draft-suh-mipshop-fmcast-mip6-00
Remote subscription with agent support
Mobile multicast reception only
Built on Fast Handovers for MIPv6 (FMIPv6)
Predictive handovers based on L2 : L3 map
Handled at access routers
Extends signalling of FMIPv6 by multicast address option
8 Prof. Dr. Thomas Schmidt http:/www.informatik.haw-hamburg.de/~schmidt
M-FMIPv6: Prediction
M-FMIPv6: Forwarding
10 Prof. Dr. Thomas Schmidt http:/www.informatik.haw-hamburg.de/~schmidt
M-FMIPv6: Handover
M-FMIPv6: Completion
12 Prof. Dr. Thomas Schmidt http:/www.informatik.haw-hamburg.de/~schmidt
Seamless Multicast Handover in a HMIPv6 Environment
draft-schmidt-waehlisch-mhmipv6-04
Agent based: MAP as Multicast agent
Mobile multicast reception and source
Built on Hierarchical MIPv6 (HMIPv6)
For mobile listeners:
Micro mobile handovers hidden by MAP
Reactive handovers between MAPs
Unicast (tunnel) forwarding MN : MAP
Change to remote subscription with MAP attendance
Extends signalling of HMIPv6 by multicast advertisement flag
M-HMIPv6: Multicast Listener
14 Prof. Dr. Thomas Schmidt http:/www.informatik.haw-hamburg.de/~schmidt
M-HMIPv6: MAP-Local Handover
M-HMIPv6: Inter-MAP Handover
16 Prof. Dr. Thomas Schmidt http:/www.informatik.haw-hamburg.de/~schmidt
M-HMIPv6: Completion
Multicast Source Mobility Problem
18 Prof. Dr. Thomas Schmidt http:/www.informatik.haw-hamburg.de/~schmidt
o Distribution tree (somehow) rooted at source:
- Collapses after movement
- Reconstruction very slow (up to minutes) o Address Duality Problem:
- Source address is logical & topological ID o Source Activation Problem:
- New source to be discovered by interdomains o Decoupling Problem:
- Source has no feedback from receivers
Multicast Source Mobility Approaches
o Remote Subscription
- Show all movement by local multicast subscription
o Bi-directional Tunneling
- Hide all movement by tunneling via Home Agent
o Agent Based
- Use current Care-of-Address for Routing
- Use Home-Addres in Destination Option at Receiver
- “Tree Walking”
M-HMIPv6: Mobile Multicast Source
Mobile multicast sender anchored at MAP:
Use Home Address Destination Option (CN must not verify BC on mcast)
On handover:
Initially tunnel through old MAP
Initiate new tree by ‘pure IP packets’
Handover on (protocol dependent) timeout
On rapid Movement:
stay with established MAP
20 Prof. Dr. Thomas Schmidt http:/www.informatik.haw-hamburg.de/~schmidt
M-HMIPv6: Multicast Source
M-HMIPv6: MAP-Local Handover
22 Prof. Dr. Thomas Schmidt http:/www.informatik.haw-hamburg.de/~schmidt
M-HMIPv6: Inter-MAP Handover (1)
M-HMIPv6: Inter-MAP Handover (2)
24 Prof. Dr. Thomas Schmidt http:/www.informatik.haw-hamburg.de/~schmidt
Multicast Source Mobility:
SSM Problems
o Receivers need to Subscribe to Source Address:
- HoA & current CoA needed at Receiver o Routers Maintain Source Specific States:
- HoA & current CoA needed at Routers o Decoupling: Source cannot Control Receiver
Updates:
- May loose receivers on handover
Multicast Source Mobility Approaches
26 Prof. Dr. Thomas Schmidt http:/www.informatik.haw-hamburg.de/~schmidt
o Remote Subscription
- Show all movement by local multicast subscription
o Bi-directional Tunneling
- Hide all movement by tunneling via Home Agent
o Agent Based
- Use current Care-of-Address for Routing
- Use Home-Addres in Destination Option at Receiver
- “Tree Walking”
Bi-Directional Tunneling
How good is BT?
Simulation: BT in ATT Network
28 Prof. Dr. Thomas Schmidt http:/www.informatik.haw-hamburg.de/~schmidt
Internet with 154 Core Nodes
Internet with 1.540 Core Nodes
30 Prof. Dr. Thomas Schmidt http:/www.informatik.haw-hamburg.de/~schmidt
Internet with 15.400 Core Nodes
What to do with Mobile SSM ?
Need to Change Routing:
Logical (HoA) and topological (CoA) identifier needed at client and routing layer
Need to keep contact to clients after handover :
Some mobility persistent entities?
Current Ideas:
Re-introduce shared trees
Re-use trees
What effect does mobility impose on SPTs?
32 Prof. Dr. Thomas Schmidt http:/www.informatik.haw-hamburg.de/~schmidt
Analysis of ‘Moving’ Distribution Trees
Multicast Distribution Trees subsequent under Mobility are highly correlated
Previous and Next Tree overlap from receivers downward
Coinciding subtrees: self-similar URT
Results in frequent re-use of Mcast Routers
Two characteristic measures
‘Step-Size’: pDR-to-nDR Distance
Tree evolvement: Number of Receivers
34 Prof. Dr. Thomas Schmidt http:/www.informatik.haw-hamburg.de/~schmidt
Simulation Study:
Tree Coincidence wrt. ‘Step-Size’
≈ 80 % Coincidence for 40 Receivers and
a mobility ‘step-size’ of 5
Simulation Study:
Tree Coincidence wrt. Tree Evolvement
> 80 % Coincidence for a mobility
‘step-size’ of 5 and
100 Receivers
Shared Tree Mobility
Use Rendezvous Point as Mobility Anchor:
Mobility-aware Rendezvous Point (MRP)
Need to Change Routing:
Extend (*,G) states to (HoA, G, MRP-ID)
Modify RPF-Check to identify CoA* sources with HoA
Operate interdomain handovers via MRP-ID
36 Prof. Dr. Thomas Schmidt http:/www.informatik.haw-hamburg.de/~schmidt
Source – Romdhani et al: Transparent Handover for Mobile Multicast Sources ,
In: Proceedings of the IEEE ICN'06, IEEE Press, April 2006
Mobility-aware Rendezvous Points
Source Specific Tree Modification
Need to Change Routing:
Extend (CoA,G) states to (CoA,HoA,G)
Need to Preserve Previous Trees:
Keep contact subsequent to handover
Idea: Morph Previous into Next Tree:
Elongate root (modify RPF Check)
Discover shortcuts
Dismiss unneeded branches
38 Prof. Dr. Thomas Schmidt http:/www.informatik.haw-hamburg.de/~schmidt
Source - Schmidt & Wählisch: Extending SSM to MIPv6 -- Problems Solutions and
Improvements, Computational Methods in Science and Technology, Vol. 11, No. 2,
pp. 147--152, Poznan:Scientific Publ. OWN --- Polish Academy of Sciences, Nov. 2005.
Tree Morphing
Root Elongation Phase
40 Prof. Dr. Thomas Schmidt http:/www.informatik.haw-hamburg.de/~schmidt
First Shortcut
42 Prof. Dr. Thomas Schmidt http:/www.informatik.haw-hamburg.de/~schmidt
Optimized Tree
Performance Snapshot
Conclusions & Outlook
44 Prof. Dr. Thomas Schmidt http:/www.informatik.haw-hamburg.de/~schmidt