Simulation vs experimental-testbeds_ioanna_tsalouchidou
0 likes280 views
This document compares wireless mesh network simulations to experimental testbeds. It finds that while simulations are convenient, they provide inaccurate modeling of real world effects like antenna diversity and path loss. At the physical layer, testbeds performed better outdoors compared to indoors while simulations did not capture this effect. At higher layers, simulations did not accurately model the impact of transmission rate or number of hops on performance. Overall, the document concludes that testbeds are still needed to validate simulation results due to limitations in channel modeling in current simulators.
Simulation vs experimental-testbeds_ioanna_tsalouchidou
- 1. Comparing Simulation Tools and Experimental Testbeds for Wireless Mesh Networks Written by: Kefeng Tan, Daniel Wu, An (Jack) Chan, Prasant Mohapatra Presented by: Ioanna Tsalouchidou
- 2. Presentation Outline: • Simulations Vs. Experimental Testbeds, discrepancies and more • Methodology • Point to Point measurements • PHY-layer: Beyond inaccurate channel modeling • MAC-layer: Flow level unfairness due to interference • IP-layer: Route stability • Conclusions
- 3. Experimental Testbeds • Real world scenarios • Affected by the testing environment • Conditions highly random and uncontrollable • High costs • Hardware limitations
- 4. Simulations • Network scenarios easily constructed and modified • Data can be easily collected • Modeling of large scale network topologies • Do not require special hardware and labor resources • Low cost comparing to testbeds • Not as trustworthy as testbed experiments • PHY layer inadequate modeling
- 5. Methodology • Simulations o NS-2 o QualNet • Experimental Testbeds: o single antenna embedded system (Soekris) o Laptops with modern Wi-Fi antennas o QuRiNet
- 6. Network scenarios • Metrics oGoodput o Delay, Delay jitter o Probability distribution • Parameters o PHY layer: path loss and antenna diversity o MAC layer: traffic load, bit rate, number of hops, flow fairness, interference o Routing layer: route length, persistence, stability and diversity of routes
- 7. Point to Point measurement • Send packets from one laptop to an other • TCP and UDP protocols • Different bitrates, distances and traffic loads o Indoor and outdoor o Transmission distance of 40 and 80 meters o Packet delivery 5 and 10 Mbps o Bitrate: auto and 54Mbps
- 8. Point to Point measurement • Simulations do not affect much • UDP goodputs are higher by the environment indoor than outdoor • In experiments laptops have • Qualnet always has the better performance indoor than highest results outdoor
- 9. PHY: Beyond Inaccurate channel modeling • Antenna Diversity o Missing from simulators o Difference of signal when antennas spaced one wavelength apart • Configuring PHY Channel o Path loss determines how fast a signal fades NS-2 path loss global parameter QualNet does not have this setting o Can be configured to fit real world scenarios
- 10. PHY: Beyond Inaccurate channel modeling • QualNet constant curves • NS-2 goodput changes as path loss increases • Laptop-outdoor outperforms laptop-indoor • Soekris doesn't show consistent results
- 11. MAC: Impact of transmission data rate • At low rates all results increase linearly • NS-2 results match laptop-outdoor • The drop of laptop indoor is not captured
- 12. MAC: Impact of multihop • Goodput and delay vary with the number of hops • For small number of hops NS-2, QualNet and testbed agree
- 13. Routing: Route Stability • Network settings oThe dominating factor for routing stability is the channel quality and stability. • Route prevalence and persistence o Prevalence: occurrence of a route over that of the dominant route o Persistence: the duration(sec) that a route lasts before it changes • Spatial prevalence and spatial distribution o Spatial prevalence: number of cumulative occurrences of one node over the number of all records
- 14. Routing: Route Stability • (a) Dominating route occurs much more often in QuRiNet • (c) NS-2: the most spatial diversity with approximately 17% of routes covering 90% of nodes,then QuRiNet (20%) and finally QualNet (60% spatially concentrated) • Routes in simulations are less stable and persistent
- 15. Conclusions ● NS-2 and QualNet simulations. ● Testbets: Soekris, laptops, QuRiNet . ● PHY, MAC and IP layers. ● Antenna diversity, path loss, multihop, transmission rate, routing stability. ● Discrepancies due to inaccurate channel modeling, antenna diversity. ● Simulators: path loss is well modeled but routes less stable and persistent.