1. Fault Tolerant Routing in Tri-Sector Wireless Cellular Mesh Networks Yasir Drabu and Hassan Peyravi Kent State University Kent, OH - 44240

2. Kent State University PDCS - 06 2 Agenda Introduction  Wireless Network Overview Problem Definition Proposed Solutions  Shortest Path Routing  Fault Tolerant Routing Conclusion

3. Kent State University PDCS - 06 3 Intro - Wireless Network Architectures Point to MultipointMultipoint to MultipointPoint to Point Dedicated links Currently Most Common Our Focus TopologyReliabilityAdaptabilityScalabilityRouting Complexity P to PHighLowNone P to MLow Moderate M to MHigh

4. Kent State University PDCS - 06 4 Wireless Mesh Networks (WMN) Structured, energy rich wireless multi-hop networks:  Wireless Client Mobile, no routing, limited power  Wireless Router Low mobility, routing and power rich  Wireless Gateway Access to wired network. Solves – “Last Mile Connectivity” INTERNET

5. Kent State University PDCS - 06 5 Problem Definition How do you route packets in a Wireless Mesh/Multi-Hop network? Given:  Faulty wireless – multi path fading, selective fading, noise etc.  Shortest path may not be the best alternative.  Multiple hops/ multiple channel – radio limitations, channel allocation problem etc.

6. Kent State University PDCS - 06 6 Background Many wireless routing algorithms: Pro-active (DBF), reactive (DSR, TORA) and hybrid (ZRP)  None are very fault tolerant and very focused on energy poor applications Few provide fault tolerance  Agarwal 2004 (Stony Brook research lab) – build routing using spanning trees then re-associate to different root when link fails. Slow, high message complexity, order of seconds. However not much work done on using topological properties of wireless network

7. Kent State University PDCS - 06 7 Proposed Honey Comb WMN Model No wired backbone for each node Place wireless elements on the edge instead of the center in a typical network Uses more nodes with lower power for better coverage and higher throughput Modified to Honey Comb Typical Cellular Network Proposed Honeycomb Model

8. Kent State University PDCS - 06 8 Honey Comb Network Comparison Cellular Network:  Central Base Station  Omni-directional antenna Advantages + Established Technologies + Fewer Base Stations Limitations – High power consumption – Limited coverage – Lower bandwidth – No Fault tolerance – Expensive to deploy and maintain due to wired back bone infrastructure. Honeycomb Network:  BS as the edge  Directional antennas Advantages + Lower power per node + Better coverage + Higher throughput + Fault tolerance + Wireless interconnect, cheaper to deploy when wired infrastructure is factored in Limitations – More complex hardware – More nodes for same area

9. Kent State University PDCS - 06 9 Proposed Tri-Sector Node Model N 120 0 Wireless Router Four Radios  Three directional antennas for communication with other routers  One omni-directional for wireless clients The directional antenna can be on the same channel as they are spatially multiplexed.  Using different channels on different lobes will add to the complexity of the problem. Omni-directional antenna is on a separate channel to minimize interference.

10. Kent State University PDCS - 06 10 Earlier routing in Honeycomb Network Honeycomb routing was introduced in [Stojmenovic:97] Issues:  Uses (x, y, z) co- ordinates to route.  No consideration for link failure. Src: Stojmenovic:97

11. Kent State University PDCS - 06 11 Honeycomb Brick Representation Two dimensional representation of honeycomb Each node can be represented by a co-ordinate (x,y) They have 25% smaller degree than regular grid meshes. stretch Isomorphic pruned 2D square mesh

12. Kent State University PDCS - 06 12 Shortest Path Routing Algorithm

13. Kent State University PDCS - 06 13 Fault Tolerance In Brick Networks Link faults common in wireless networks How do we handle a fault in a mesh network?  Localized Temporal Routing Temporal Routing Based on  Final direction of packet  Position of fault  Number of faults

14. Kent State University PDCS - 06 14 Fault Tolerant Routing Algorithm Fault detection: Physical layer or the Medium Access Layer detects the fault. Fault avoidance: Once a fault has been detected, the algorithm goes into recovery mode. Exploited topological properties to define alternate path.

15. Kent State University PDCS - 06 15 Fault Routing – Single and Multiple failures

16. Kent State University PDCS - 06 16 Limitations Fault tolerance is a trade-off between delay and deliverability.  More hops introduce delay. Model needs ground up deployment Topological Rigidity  Cannot be deployed on all terrains

17. Kent State University PDCS - 06 17 Conclusions Contributions:  Modeled fault tolerant network topology  Efficient addressing scheme  Shortest path routing algorithm  Developed fault tolerant routing which can handle multiple faults. Future Work:  Gateway Placement  Resource allocation (channel assignment)

18. Questions?

19. Kent State University PDCS - 06 19 Wireless Multi-hop Networks Type of multi-hop networks (Application Level Classification):  Ad hoc Limited power, high mobility, relatively small. Primary application – file sharing and collaboration.  Sensor Networks Very low power, low bandwidth, large networks. Primary application – Data accusation and sensing.  Wireless Mesh Networks (WMN) Power rich, structured, high throughput Primary application – access network to end users.

20. Kent State University PDCS - 06 20 Routing Challenges in WMN Time varying link behavior Shortest Path not always the best route Using spanning trees do not exploit the natural robustness of a WMN. Exploit alternate routes to make WMN fault tolerant How to achieve load balancing.  How to maintain alternate routes?  How to choose one route over the other? On what basis/metrics?