Ad Hoc Networking with AODV

1. Ad Hoc Networking with AODV Charles E. Perkins Nokia Research Center Mountain View, CA USA http://people.nokia.net/charliep charliep@iprg.nokia.com

2. Outline of Presentation • Ad Hoc Networks in general • AODV in particular • Recent results from manet • Internet Gateways for ad hoc networks • Address autoconfiguration

3. peer-to-peer multihop dynamic zero-administration low power autonomous autoconfigured Ad Hoc Network characteristics But, most of these have exceptions! Idea: let (?almost?) every node be a router

4. Commercial (or not!) Opportunities • Conferencing • Home networking • Range extension for cellular base stations • Emergency services • Ambulance • Police • Hospitals • Embedded computing applications • Ubiquitous computers with short-range interactions • Automotive/PC interaction • Enable computing where subnets do not exist • Jungle telemetry

5. Technical/Market/Political hurdles • Scalability (memory search time, bandwidth, processing) • Power budget vs. latency • Protocol deployment, incompatible standards • Why should one node “waste power'' to help a neighbor ? • Wireless data rates • Obsoletes the client/server model...  breaks a lot of protocols • User education, acculturation • Antenna inconvenience (not anymore, really) • Higher bit-error-rate (BER) • Additional security exposure • Non-ubiquitous coverage

6. On-Demand Routing Protocols • Eliminate route table updates for routes that are not used • Fewer control packets: •  better scalability, reduced congestion, better robustness •  reduced processing requirement • Even more localization for topology changes if distance vector • Also can be made to work for (partial) link state • or, better, hybridized distance vector and link state • Downsides: • Latency • Route Discovery broadcasts • ICMP Unreachable only after Route Discovery attempt

7. Mobile Ad Hoc Networking (manet)/ AODV • AODV: on-demand, and distance vector • Route caching & timeout offers improvement over others • Proved “correct” • Interoperability testing, and (soon?!) Experimental RFC status • AODV uses network-wide RREQ, unicast RREP along reverse path to source of the request. • DSR uses similar route discovery, maintains source routes • OLSR and TBRPF are link state, proactive protocols • Active discussion about Internet Gateways • Address Autoconfiguration • Reducing retransmissions for system-wide flooding

8. AODV Features • Reactive routing protocol; route discovery cycle for route finding • Route repairs and TTL restrictions reduce network-wide flooding • Maintenance of active routes • Loop freedom achieved through sequence numbers • No overhead on data packets • Scalability shown to 10,000 nodes • performance suffers • Integrated multicast protocol (MAODV) specified • multiple next hops • group leader maintains sequence number • QoS extension specified (undergoing revision) • AODV for IPv6 is specified, built, and works

9. AODV Unicast Route Discovery Initiation Route Request (RREQ) broadcast flood Destination Source

10. AODV Unicast Route Discovery Completion Route Reply (RREP) propagation Destination Source

11. Internet Gateways for Ad Hoc Networks Entry node • Our model: do not inject per-host routes into Internet • Good start: ad hoc nodes use gateway as default router • but it could be multiple hops away • plus, the ad hoc nodes need to know its IP address • router solicitation/advertisement “work”, with changes • Gateway should be “protocol-agnostic” (for any manet protocol) • Gateway needs a host route for each manet node Gateway

12. Gateway Address Autoconfiguration • Node discovers Internet-routable prefix from Internet Gateway, if any • Otherwise, use canonical site-local address • Required: some variety of Duplicate Address Detection (DAD) • For connected networks, RREQ/RREP does the job • tricky part: what is the source address? • have specified AREQ and AREP for “general” case (should work with protocols other than AODV) • The hard part: dealing with network merge or healing

13. Ad Hoc Networking Research • MobiHoc (ACM SIGMOBILE) (plus quite a few others!) • Third conference held in June – 150 papers submitted • Active research areas (a few among many!) • Inherent capacity bounds? • Better Routing • Automotive (“parallel” one-dimensional networks?) • Backbones, Clustering • Power control • Simulations seem quite untrusted • AODVng • Gray zones (interference range vs. signal range; HELLO nonworking) • QoS/Diffserv/”no free lunch” • Security (!!) • Implementer’s mailing list • Multipath • AODVjr.

14. Summary and Conclusions • Ad Hoc Networking is well-established as a viable research area • Infrastructureless operation has many applications • On-demand protocols offer many advantages • AODV makes use of advantages from both Distance-Vector and On-demand • AODV has good chances for standardization • Ad hoc networks can be glued to the Internet and then provide wireless extension domains • Address autoconfiguration techniques have been adapted

15. DSR (Dave Johnson, CMU) WINGs (JJ Garcia/UCSC) ROAM(JJ Garcia/UCSC) WAMIS (Gerla/UCLA) ODMRP (Gerla et.al/UCLA) TRAVLR (Kleinrock/UCLA) Tora/IMEP (Park, Corson/UMD) Link Quality (Rohit Dube/UMD) LAR (Texas A&M) TBRPF (SRI) OLSR (Inria: Clausen./Jacquet) DSDV (Dest. Sequence #'s) AODV (refinement of DSDV) AOMDV (Multipath – Das/Marina) Hierarchical (Akyildiz/Georgia Tech) GPSR (Karp/Harvard) CBRP (Singapore) Terminodes (EPFL) MMWN (Steenstrup/BBN) ABR (C.K. Toh) STAR (JJ Garcia/UCSC) ZRP (Zygmunt Haas/Cornell) Fisheye/Hierarchical (UCLA) CEDAR (Urbana-Champaign) Various Ad Hoc Routing Projects

16. Distance Vector Routing Protocols • Route table has (metric, next hop) – i.e., (distance, vector). • Other metrics (e.g., time) may be more useful in many cases • “Distributed Bellman-Ford” algorithms • Can be made loop-free • Easy to program • Low memory and processor utilization • Localized update operations (important for ad hoc) • Susceptible to counting-to-infinity problem • Previous solutions (poison reverse/split-horizon) must be undone