1. Evaluating the Cost of Frequency Diversity in Communication and Routing Overview Jorge Ortiz* ♦ David Culler* Causes of Loss  Pairs of nodes sharing a vertex may not share a common communication channel  2 Transmissions necessary for either multiple frequency transmission of multihop routing  How much do we need frequency diversity?  Goal: pin down view of the network under various levels of external interference  Examine the various factors that cause loss  Observe and quantify how routing diversity can improve communication efficiency and robustness  802.11 may not interfere as much as we think  Find a good channel and stick to it  Establish channel and connectivity comparison metrics Methodology  Peak hours see the highest average noise floor in the mid- afternoon between 4pm-4:15pm  Most channels are free most of the time throughout the entire network  On channel 17 in the worst time interval, only 21% of the samples are greater than -77 dBm (the default clear-to-send threshold CC2420 Radio)  Plenty of opportunities for transmission by 802.15.4 motes Network-Wide Noise Observations Temporal Dynamics  How do graphs change over time?  What’s the distance between the best and worst network topologies?  To what extent can we simply rely on route diversity, rather than frequency diversity, to improve communication efficiency?  Do the quality of links necessarily get better? Current Results and Status  Sufficient connectivity even during peak 802.11-traffic hours on worst channels  802.15.4 channels that do not overlap with 802.11 channels experience little change between peak traffic hours and quiet hours  802.11 channels are static and therefore the best 802.15.4 channels do not change  Network topology (route diversity) may make the need for multiple channels irrelevant  How do non-office environments look? Would link and connectivity behavior be similar?  More sampling and experimentation  What does this mean for MAC and Routing protocol design? Future Work RSSI (dBm) Peak Time Quiet Time Link Quality Comparison  Interested in real networks: use traces  Noise Monitoring: Wi-Spy Spectrum Analyzer and passive RSSI mote monitoring  Connectivity: Round-robin 100-packet broadcasts with 20 ms inter-packet transmission interval on each of 16 channels  51-56 MicaZ motes  163,200 transmissions, 1,781,231 receptions  10 Access Points in Computer Science Building’s 4 th floor A B C 1 2  Internal Interference  Collisions from cross traffic within nodes in the same network  External Interference  802.11 traffic and devices that do not play nice with 802.15.4  Microwave oven, cordless phones, physical obstructions  Radio irregularities  Narrow-band fading  Choose optimal routes in topology graph according to a route selection criteria Channel Grade (%)Duty Cycle 1181.780562.052778 1277.1253.958333 1375.897224.719444 1479.672232.977778 1582.138891.827778 1681.369452.230556 1778.711113.705555 1876.472224.627778 1980.641662.425 2083.419451.380556 2180.555562.411111 2278.530563.419445 2379.044443.405555 2481.416662 2583.972221.027778 2683.594441.172222 *{jortiz, culler}@cs.berkeley.edu; EECS UC Berkeley