Sybot: An Adaptive and Mobile Spectrum Survey System for WiFi Networks Kyu-Han Kim Deutsche Telekom R&D Lab USA Alexander W. Min and Kang G. Shin Real-Time Computing Lab, University of Michigan ACM MobiCom 2010 © Kyu-Han Kim Why Spectrum Site-Survey for WiFi? Coverage and capacity Interference or attack RF-based localization WiFi Spectrum Map Survey system for efficient and accurate monitoring 2 Outline Limitations and Challenges System Design Performance Evaluation Conclusion 3 Limitations and Challenges Exhaustive measurements Comprehensive results [Raniwala03] Easy to visualize and analyze data Labor-intensive operation Sensor-based measurements Continuous monitoring [Yin08] Can be inexpensive [Bahl05] Inflexible, due to its static location Accuracy and repeatability Efficiency and flexibility Adaptation and awareness 4 Outline Challenges and Limitations System Design Performance Evaluation Conclusion 5 Sybot: Spectrum Survey Robot Design Accuracy and repeatability Efficiency and flexibility Adaptation and awareness Periodic and aperiodic surveys Decomposition of a survey task Extraction of site-specific spectrum characteristics Controlling key survey parameters to meet requirements 6 Sybot Operations Periodic & on-demand Adaptive monitoring Grid-based spectrum map Build/control a profile Thour Tmin Diagnostic GUI Selective Mobility Controller Diagnostic MAP Spectrum Monitor Complete Selective Complete Filters driver Tday Scheduler App. layer Metric of Interests 1 m - i rssi ( j ) m j 1 - i 1 m ( rssi ( j ) i ) 2 m j 1 7 Complete Monitoring Complete Selective Diagnostic Comprehensive measurement co rri d or AP AP AP ii Measurement Unit grid point ro o m Good Cumulate n spectrum maps - Baseline spectrum map, Bi Selection of a grid size Bad 8 Complete Selective Monitoring Selective Diagnostic Cope with temporal variance Identify areas with correlation b(i) { j | for grid i and j, | (i) ( j ) | } ido r AP AP i co rr Measurement point ro o m Good Reference grid Bad Block b(i) 1 3 2 4 R, a set of reference grids b(1)={1,2} b(2)={1,2,4} b(3)={3,4} b(4)={2,3,4} Candidate R {1,2,3} {2,3} {1,3} {1,4} 9 Complete Diagnostic Monitoring Selective Diagnostic AP ido r APi co rr Measurement point ro o m Good Detect areas with deviation - diff (i) | i i | , diff (i) k Update area w/ suspicious grids Perform diagnostic movements Bad Diagnostic movements Suspicious reference grids g1 g0 10 Outline Challenges and Limitations System Design Performance Evaluation Conclusion 11 Performance Evaluation Prototype Wireless Router IEEE 802.11 Router (Linux) iRobot Create for automation Sensors iRobot Sybot Prototype Measurement and analysis Corridors and office rooms 4 weeks and >10,000 points WiFi Test-bed 12 Generating Repeatable Baseline Map Complete monitoring result Histogram of σ 87% of grids < 4 dBm 13 Reducing Space to Survey Complete monitoring result Selected reference grids Measurement space reduction > 50 % 14 Building a Profile for Efficiency vs. Accuracy Efficiency Profile Accuracy Profile 70% reduction b(i) { j | for grid i and j, | (i) ( j ) | } Construction of a trade-off profile per site 15 Effectiveness of Diagnostic Monitoring Complete monitoring result OBSTACLE Diagnostic monitoring result Measurement space reduction > 56 % 16 Conclusion Spectrum site-survey for WiFi networks is important for key network management and services. Key challenges and limitations in designing a spectrum survey system have been identified. Sybot is a novel spectrum survey system that adaptively uses three complementary monitoring techniques. A prototype and extensive measurement study show its feasibility and effectiveness (> 50% reduction). 17 Q&A Thank You Contact Information: kyu-han.kim@telekom.com 18