Evaluating the Effect of Path Diversity over QoS and QoE in a High Speed Indoor Mesh Backbone Sandip Chakraborty 12, Sukumar Nandi Department of Computer Science and Engineering Indian Institute of Technology Guwahati, Guwahati 781039 INDIA 08 January, 2014 1 This work is supported by TATA Consultancy Services (TCS), INDIA through TCS Research Fellowship program 2 Supported by COMSNETS 2014 Travel Grant Sandip Chakraborty (CSE, IITG) COMSNETS 2014 08 January, 2014 1 / 23
Preface: QoS and QoE in the Network Quality of Service (QoS): Performance guarantee from network perspective, - Throughput (available bandwidth) - End-to-end delay - Jitter (variation in per packet delay) Quality of Experience (QoE): Performance for user s perspective, - Mean Opinion Score (MOS) for voice traffic - Peak Signal to Noise Ratio (PSNR) for video traffic - Structural Similarity Index Measurement (SSIM) for video traffic How path diversity affect QoS and QoE in a multi-hop IEEE 802.11n mesh network in an indoor scenario? Sandip Chakraborty (CSE, IITG) COMSNETS 2014 08 January, 2014 2 / 23
Diversity in a Mesh Network Path Diversity: Multiple paths from a source to a destination Traffic Diversity: Diversity in Path Quality: Path quality significantly varies with respect to time. Data Rate Diversity: Varies from 2 Mbps to 600 Mbps with IEEE 802.11n Sandip Chakraborty (CSE, IITG) COMSNETS 2014 08 January, 2014 3 / 23
Forwarding in a Mesh Network (IEEE 802.11s) Hybrid Wireless Mesh Protocol (Combination of proactive and reactive routing) Routing metric: Airtime Link Metric [ C = O ca +O p + B ] t 1 r 1 e f Where, O ca and O p are the constants, named as the channel access overhead and the protocol overhead, B t is the test frame size. The input parameters r and e f are the bit rate in Mbps and the frame error rate for the test frame size B t. Sandip Chakraborty (CSE, IITG) COMSNETS 2014 08 January, 2014 4 / 23
Testbed Environment N3 N 2 N 1 50 m G RS Lab 2 RS Lab 1 N4 N9 N10 N 8 Security Lab N5 N 6 N 7 Sandip Chakraborty (CSE, IITG) COMSNETS 2014 08 January, 2014 5 / 23
Testbed Setup Router: RaLink RT-3352 RoC: 2T2R MAC/BBP/PA/RF, 400MHz MIPS24KEc CPU, 64MB of SDRAM and 32MB of Flash IEEE 802.11n: 300 Mbps, channel bonding Open80211s: http://www.open80211s.org Linux Kernel 2.8.54 TCP (FTP) and UDP (TFTP) using iperf (http://iperf.sourceforge.net/) Tx Power 16dBm, Rx Sensitivity 0 dbm (45-55 mt in indoor) Sandip Chakraborty (CSE, IITG) COMSNETS 2014 08 January, 2014 6 / 23
Effect of Diversity over Airtime Link Metric SINR (dbm) -70-80 -90-100 -110-120 2 2.4 2.8 3.2 3.6 4 4.4 Load 60 50 40 30 20 10 0 2 2.4 2.8 3.2 3.6 4 4.4 ALM (ms) 1.2 1 0.8 0.6 0.4 2 2.4 2.8 3.2 3.6 4 4.4 Time (Hrs) Sandip Chakraborty (CSE, IITG) COMSNETS 2014 08 January, 2014 7 / 23
Selective Greedy Forwarding (SGF) 3 Selection of a set of potential forwarders (Proactive approach) Selection of the next hop from the set of potential forwarders (Greedy approach) - Effect of the variation in link information over the path information S 3 Chakraborty, S.; Chakraborty, S.; Nandi, S., Beyond conventional routing protocols: Opportunistic path selection for IEEE 802.11s mesh networks, in proc. of IEEE 24th International Symposium on Personal Indoor and Mobile Radio Communications (PIMRC), pp. 3224-3228, 8-11 Sept. 2013 Sandip Chakraborty (CSE, IITG) COMSNETS 2014 08 January, 2014 8 / 23
Route Flapping for QoS/QoE High Route Flapping: Network inconsistency Low Route Flapping: Network fails to adopt with channel variation Sandip Chakraborty (CSE, IITG) COMSNETS 2014 08 January, 2014 9 / 23
SGF: Route Flapping SINR Variation (dbm) 30 25 20 15 10 5 0 1 2 3 4 5 6 7 8 9 10 11 Route Flapping Low Load 60 50 40 30 20 10 0 Reactive HWMP Proactive HWMP SGF 0 1 2 3 4 5 6 7 8 9 10 11 Route Flapping High Load 60 50 40 30 20 10 0 Reactive HWMP Proactive HWMP SGF 0 1 2 3 4 5 6 7 8 9 10 11 Router Number Sandip Chakraborty (CSE, IITG) COMSNETS 2014 08 January, 2014 10 / 23
Forwarder Percentage (δ) Percentage of neighbors selected as the set of potential forwarders. Route Flapping Low Load Route Flapping High Load 45 40 35 30 25 20 15 10 5 0 45 40 35 30 25 20 15 10 5 0 δ = 60% δ = 40% δ = 20% 0 1 2 3 4 5 6 7 8 9 10 11 δ = 60% δ = 40% δ = 20% 0 1 2 3 4 5 6 7 8 9 10 11 Router Number Sandip Chakraborty (CSE, IITG) COMSNETS 2014 08 January, 2014 11 / 23
QoS Metrics: MAC Throughput Average Per User MAC Throughput (Mbps) 8 7 6 5 4 3 2 1 HWMP: Proactive HWMP: Reactive SGF: δ=30% SGF: δ=40% SGF: δ=50% 0 10 20 30 40 50 60 Average Number of Users Sandip Chakraborty (CSE, IITG) COMSNETS 2014 08 January, 2014 12 / 23
QoS Metrics: Forwarding Delay Average End-to-End Delay (ms) 1100 1000 900 800 700 600 500 400 300 200 HWMP: Proactive HWMP: Reactive SGF: δ=30% SGF: δ=40% SGF: δ=50% 100 10 20 30 40 50 60 Average Number of Users Sandip Chakraborty (CSE, IITG) COMSNETS 2014 08 January, 2014 13 / 23
QoS Metrics: Average Jitter Average Jitter (ms) 32 30 28 26 24 22 20 HWMP: Proactive HWMP: Reactive SGF: δ=30% SGF: δ=40% SGF: δ=50% 18 16 10 20 30 40 50 60 Average Number of Users Sandip Chakraborty (CSE, IITG) COMSNETS 2014 08 January, 2014 14 / 23
QoE Metrics: MOS (Voice) 4.4 4.2 4 HWMP: Proactive HWMP: Reactive SGF: δ=30% SGF: δ=40% SGF: δ=50% 3.8 MOS 3.6 3.4 3.2 3 2.8 2.6 10 20 30 40 50 60 Average Number of Users Sandip Chakraborty (CSE, IITG) COMSNETS 2014 08 January, 2014 15 / 23
QoE Metrics: PSNR (Video) 35 30 HWMP: Proactive HWMP: Reactive SGF: δ=30% SGF: δ=40% SGF: δ=50% 25 PSNR (db) 20 15 10 5 10 20 30 40 50 60 Average Number of Users Sandip Chakraborty (CSE, IITG) COMSNETS 2014 08 January, 2014 16 / 23
QoE Metrics: SSIM (Video) 1 0.95 0.9 HWMP: Proactive HWMP: Reactive SGF: δ=30% SGF: δ=40% SGF: δ=50% SSIM 0.85 0.8 0.75 0.7 0.65 10 20 30 40 50 60 Average Number of Users Sandip Chakraborty (CSE, IITG) COMSNETS 2014 08 January, 2014 17 / 23
Adopting δ with Traffic Load Variation At the beginning of every DTIM interval, mesh routers compute traffic load in terms of number of associated users (Q r (t)). δ min 2, δ max 0.5N r Model the network as a birth-death process; pi 1 i δ +1 min δ min δ i 1 δ i δ max pi i 1 Sandip Chakraborty (CSE, IITG) COMSNETS 2014 08 January, 2014 18 / 23
Average Forwarding Delay Average End-to-End Delay (ms) 1100 1000 900 800 700 600 500 400 300 200 HWMP: Proactive HWMP: Reactive SGF + Adaptive SPF 100 10 20 30 40 50 60 Average Number of Users Sandip Chakraborty (CSE, IITG) COMSNETS 2014 08 January, 2014 19 / 23
Average Jitter 32 30 HWMP: Proactive HWMP: Reactive SGF + Adaptive SPF 28 Average Jitter (ms) 26 24 22 20 18 16 10 20 30 40 50 60 Average Number of Users Sandip Chakraborty (CSE, IITG) COMSNETS 2014 08 January, 2014 20 / 23
Voice: MOS 4.4 4.2 4 HWMP: Proactive HWMP: Reactive SGF + Adaptive SPF 3.8 MOS 3.6 3.4 3.2 3 2.8 2.6 10 20 30 40 50 60 Average Number of Users Sandip Chakraborty (CSE, IITG) COMSNETS 2014 08 January, 2014 21 / 23
Channel Fluctuation vs MOS SINR Variation (dbm) 0-10 -20-30 -40 5.2 5.6 6 6.4 6.8 7.2 7.6 8 Load 80 70 60 50 40 30 20 10 5.2 5.6 6 6.4 6.8 7.2 7.6 8 MOS 7.2 6 4.8 3.6 2.4 HWMP Proactive HWMP Reactive SGF 5.2 5.6 6 6.4 6.8 7.2 7.6 8 Time (Hrs) Sandip Chakraborty (CSE, IITG) COMSNETS 2014 08 January, 2014 22 / 23
Conclusion Explores the effects of path diversity over QoS and QoE in a high speed mesh network Evaluated Selective Greedy Forwarding : A new routing paradigm through testbed results, - Channel fluctuation and traffic load affects the performance of SGF - Use large δ at low traffic load, and small δ at high traffic load, to avoid both high and low route flapping - Adopt forwarding paths based on network conditions Sandip Chakraborty (CSE, IITG) COMSNETS 2014 08 January, 2014 23 / 23
Conclusion Explores the effects of path diversity over QoS and QoE in a high speed mesh network Thank You Evaluated Selective Greedy Forwarding : A new routing paradigm through testbed results, - Channel fluctuation and traffic load affects the performance of SGF - Use large δ at low traffic load, and small δ at high traffic load, to avoid both high and low route flapping - Adopt forwarding paths based on network conditions Sandip Chakraborty (CSE, IITG) COMSNETS 2014 08 January, 2014 23 / 23