1. Smart Networks Project University of California, Berkeley DARPA NMS PI Meeting Miami, Jan 21-23, 2004

2. Group Members Faculty Jean Walrand Pravin Varaiya Venkat Anantharam David Tse Industry David Jaffe (Cisco) Staff Bill Hodge Students Eric Chi Antonios Dimakis Rajarshi Gupta Linhai He Zhanfeng Jia John Musacchio Wilson So Teresa Tung

3. Outline Modeling interference in MANET Using models for QoS strategies Clustering Admission Control Routing New multi-channel MAC Conclusions and Future Work

4. Why Interference is critical In wired networks, all links may be used simultaneously In MANET, neighboring links interfere Interference Range (Ix) > Transmission Range (Tx)

5. Outline Modeling interference in MANET Using models for QoS strategies Clustering Admission Control Routing New multi-channel MAC Conclusions and Future Work

6. Interference may be modeled as a Conflict Graph ‘Cliques’ in a Conflict Graph Clique = Complete Subgraph Only one vertex in a clique may be active at once Capacity closely related to cliques Conflict Graph and Cliques Cliques: ABC, BCEF, CDF

7. Theoretical Result Unfortunately, capacity constraints based on cliques are not sufficient Graph theory result: Flows that satisfy scaled clique constraints have a realizable schedule Scaling factor: Clique constraints suggest a rate of 0.5 per link But only 0.4 per link is achievable “Graph Imperfection I”, S. Gerke and C. McDiarmid, Journal of Combinatorial Theory, Series B, vol. 83 (2001), pp. 58-78.

8. Complete Distributed Mechanism Local link state exchange: position, flow Distributedly compute all cliques Recompute upon topology change Requested flow (rate + path) checked by all nodes in neighborhood of path Check allocated and requested flows against clique constraints scaled by 0.46 Admit flows if satisfied

9. OPNET Simulation Model

10. Received vs Sent Rates -- 3 Flows -- 4 Flows -- 5 Flows Clique Predicted Limit – 3 Flows Clique Predicted Limit – 4 Flows Clique Predicted Limit – 5 Flows All flows have the same sending rate X-axis: average rate of sent traffic Y-axis: average rate of received traffic Vertical lines show theoretical capacity limits predicted by clique constraints

11. Outline Modeling interference in MANET Using models for QoS strategies Clustering Admission Control Routing New multi-channel MAC Conclusions and Future Work

12. Routing using Clustering with Interference Considerations Routing without clustering does not seem to scale Consider the effects of interference in clustering Minimize cross cluster interference Decompose intracluster route computation

13. Clustering

14. Routing Source Dest

15. Decomposition of Clique Constraints Intracluster routing strategies OSPF Integer Linear Programming Decomposing to per-cluster computation Decomposed clique constraints by cluster Comparison against network-wide clique constraints Simulations show that decomposed constraints result in reasonable network performance See poster for detailed results

16. Outline Modeling interference in MANET Using models for QoS strategies Clustering Admission Control Routing New multi-channel MAC Conclusions and Future Work

17. New Multi-Channel MAC For Infrastructure and ad hoc wireless networks with many channels and a high node density Propose A new protocol to increase network throughput by allowing parallel packet transfers on different channels Key Distinction Parallel contention on all channels Per-packet dynamic channel selection Initial Simulation Results Seems stable under high load in various simulations Reasonable delay statistics

18. Conclusions and Future Work Conclusions Modeled interference constraints in MANET Routing and Clustering considering interference Simulations validate theoretical models Novel multi-channel MAC to augment throughput Future Work Distributed QoS routing algorithm for a general MANET Measurements to refine clique constraints Incorporate timing and mobility considerations Handle multiple channels and classes of service