Improving QoS Support in Mobile Ad Hoc Networks Agenda • Motivations • Proposed Framework • Packet-level FEC • Multipath Routing • Simulation Results • Conclusions
Motivations • Challenges for improving QoS in MANET • Network congestion, buffer overflow Same as we met in wired networks, but bandwidth is much lower • Radio channel characteristics Multipath propagation, path loss, interference … • Frequent topology reconfigurations Constant rerouting & packet dropping due to link/path failures
Motivations • QoS provisioning in MANET requires QoS-based routing protocol, Medium Access Control (MAC) protocol, and resource reservation protocol to work together. • This work focuses on improving QoS performance at the network layer, addressing packet losses due to link and path failures resulting from node mobility. • Improve packet delivery ratio • Improve end-to-end delay and jitter • Maintain low control overhead • Reduce bursty packet losses
Proposed Framework • Multi-path routing protocol (MPR) • Spatial reutilization of wireless channel • Improve packet delivery ratio, end-to-end delay and jitter, routing overhead • Reduce burstiness of packet losses • Packet-level Forward Error Correction scheme • Reduce average packet loss rate • Avoid retransmission
Observations • Packet-level FEC can reduce packet loss, avoid retransmission and associated delay. • But, no significant gains can be attained by adopting packet-level FEC over single path routing in MANET. • Packet loss tends to be bursty due to frequent path failures • A packet interleaving scheme is needed. • Delay -- packets need to be buffered for interleaving before being sent out. • Memory requirements. • Multi-path routing can also act as a packet interleaver.
A 3x4 Packet Interleaver • Input sequence: [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12] • Output sequence: [1, 4, 7, 10, 2, 5, 8, 11, 3, 6, 9, 12] • No impact on packet loss rate, but effectively reduces the average burst length, converting bursty losses to random losses.
Multipath Routing Scheme • Dynamic Source Routing (DSR) is chosen as the basis protocol for MPR implementation. • Major difference between MPR and DSR: • Route Discovery: • Target node replies indiscriminately to all incoming route requests carrying node-disjoint routes. • Intermediate nodes no longer reply to route requests. • Route Maintenance: • New route discovery initiated only after all active routes broke. • Packet Distribution: • Round robin packet distribution over multiple routes. • Up to 3 node-disjoint paths are concurrently in use.
Simulation Model • All modifications were integrated directly into Qualnet. • Random waypoint mobility model. 50 mobile nodes randomly placed in a terrain of dimension (1500, 1500). • CBR traffic, 5 sessions, from 10, 11, 12, 13, 14 to 25, 26, 27, 28, 29, respectively. Each with 500 data packets of size 512 bytes. • IEEE 802.11 MAC with RTS/CTS. • Metrics: • Packet delivery ratio, end-to-end delay and jitter, average routing overhead, burst length of packet loss.
Conclusions • MPR significantly outperforms SPR in all criteria. • SPR-FEC performs worse than SPR, due to inherent packet loss correlation in MANET. • MPR reduces most consecutive packet losses to single packet losses, desirable by real-time video/audio applications. • MPR-FEC further improves packet delivery ratio, but at the cost of higher delay, jitter and control overhead (compared to MPR). • Higher FEC redundancy may not always be good.
The End Questions? Thanks!