The Scalability Study of AODV

The Scalability Study of AODV Sung-Ju “S.J.” Lee sjlee@hpl.hp.com Joint work with Elizabeth Belding-Royer and Charles Perkins Introduction w w w First to model up to 10,000 nodes in wireless ad hoc networks using detailed simulation (GloMoSim) Evaluate the performance of the Ad hoc On-Demand Distance-Vector (AODV) algorithm, which is one of the leading reactive protocol What can be improved? – Route discovery flood – Route repair w Incorporated the following enhancement schemes into AODV: expanding ring search, query localization, and local recovery Related Work w Clustering and hierarchical routing – Scalable in wired networks – Periodic messaging required to maintain clusters – Longer paths w Proactive table-driven routing – – – – Prioritized tier connectivity information exchange algorithm Threshold distance vector routing algorithm Landmark routing Large table sizes w Hybrid protocols – Zone Routing Protocol (ZRP): routing to the nodes inside the zone is proactive while to the node outside the zone is reactive AODV Overview w w w w On-demand route creation Two dimensional routing metric: sequence number and hop distance The source floods RREQ to search a route A node can send RREP if – it is a destination, or – has a fresh enough route to the destination w When a node detects a route break in active sessions, sends RERR to upstream nodes and the source Route Discovery Flood w Expanding ring search – Search successively larger areas centered around the source node, until a node with a route to the destination is located – Initially, the source uses a small TTL value (e.g., 2) – If no route is found, the source increments the TTL value – When re-discovering a route after a link break, the last known hop count to the destination is used as the TTL w Query Localization – Flooding of the RREQ is restricted to some area that is based on the previously known route to the destination – Whenever a node that was not on the previous route receives the RREQ, it increments the counter – When a node that was on the route receives the RREQ, the counter is reset to zero – Once the counter exceeds the threshold, RREQ is dropped Route Repair w Local Repair – The upstream of the broken link attempts to repair the broken link instead of sending RERR – TTL is set to the last known hop count to the destination plus an increment value – If route recovery fails, the node sends a RERR to the source node Performance Evaluation w w Goal: Determine benefit of modifications Protocols P ro to co l C o m b in atio n AODV A O D V an d E xp an d in g R in g Search A O D V an d Q uery Lo caliza tion A O D V an d Local R ep air A O D V , E xp and ing R in g Search, and Lo cal R ep air A O D V , Q uery Localizatio n an d Local R ep air A b b re v ia tio n AODV A O D V -E R S A O D V -Q L A O D V -LR A O D V -E R S-LR A O D V -Q L-LR Simulation Environment # of nodes 50 100 500 1000 5000 10000 w w w R oom size (m * m ) 1000 * 1000 1500 * 1500 3500 * 3500 5000 * 5000 11500 * 11500 16000 * 16000 A vg. # of neighbors 7.32 7.46 7.33 7.69 7.22 7.50 MAC: IEEE 802.11 DCF 20 sources each sending 4 packets/sec Random waypoint mobility model – Min speed: 0 m/s; Max speed: 10 m/s – Pause time: 30 seconds Throughput w w Throughput degrades in larger networks because of longer path lengths Local repair improves performance Route Recovery Rate Path Length Number of Route Recovery Attempts Control Overhead w Local recovery schemes have less overhead than source recovery schemes Ratio of RREQs to Total Control Packets Number of All Packet Transmission per Data Delivery End-to-End Delay w w Source initiated recovery schemes have longer route re-establishment latency Expanding ring search yields longer delays Lessons Learned w w w w w Scaling routing protocols in ad hoc networks is inherently difficult due to the mobility of the nodes and the transience of network links Local repair proves to be effective in enhancing AODV performance in large networks Local repair works efficiently with expanding ring search and query localization Significant amount of work remains to reach the goal of ad hoc network scalability Paper: S.-J. Lee, E.M. Belding-Royer, and C.E. Perkins, “Scalability Study of the Ad Hoc On-Demand Distance Vector Routing Protocol,” to appear in International Journal of Network Management, 2002.