11/23/2009 Outline Paper presentation � Ultra-Portable Devices � Introduction Paper: � MAC protocol design � Lower bound of energy performance Ultra-Low Duty Cycle MAC with Scheduled Channel � Protocol implementation Polling � Experimental evaluation Wei Ye, Fabio Silva and John Heidemann USC Information Science Institute � Summary and discussion Proceedings of the 4th international conference on embedded networked sensor systems, 2006, pages 321-324, ACM Presented by: Nafiseh Mazloum 2009-11-23 Paper Presentation - Ultra Portable Devices 1 2009-11-23 Paper Presentation - Ultra Portable Devices 2 Introduction(I) Introduction(II) � Major sources of energy waste � Scheduled contention � Idle listening => duty cycle � Overhead due to schedule maintenance � Collision � Listening during contention interval when nothing to send � Overhearing � Control overhead � Approaches to keep connectivity of network � Scheduled channel access like TDMA � Random channel access � Scheduled contention � Low-power listening 2009-11-23 Paper Presentation - Ultra Portable Devices 3 2009-11-23 Paper Presentation - Ultra Portable Devices 4 1
11/23/2009 Introduction(III) MAC protocol design(I) � Low-power listening � Main goals � Efficiency is at cost of transmitter � To reduce the duty cycle � Sensitive to traffic and network size � To adapt to variable traffic loads � Challenging to adapt LPL to new radios � Scheduled channel polling MAC (SCP-MAC) design � Synchronized channel polling � Adaptive channel polling 2009-11-23 Paper Presentation - Ultra Portable Devices 5 2009-11-23 Paper Presentation - Ultra Portable Devices 6 MAC protocol design(II) MAC protocol design(III) � Synchronized channel polling � Adaptive channel polling � short wake-up tone � Channel characteristics � Scheduled neighbor polling time � mix periodic and bursty traffic � unpredictable traffic mixes � Robust to varying traffic load � To detect bursty traffic � Penalty => maintaining schedule synchronization � To add dynamically high-freq. channel polling 2009-11-23 Paper Presentation - Ultra Portable Devices 7 2009-11-23 Paper Presentation - Ultra Portable Devices 8 2
11/23/2009 MAC protocol design(IV) Lower bound of energy performance(I) � Design optimization � Assumption � Two phase contention to reduce collision probability � Periodic traffic � Overhearing avoidance � Single hop with n+1 nodes � Broadcast traffic with interval T data � based on address destination in MAC header � optional RTS/CTS � Expected energy consumption by radio cost of maintaining + synchronization � t poll = radio transision from sleep to listen + sampling time to detect channel � ignor radio transition costs of other states 2009-11-23 Paper Presentation - Ultra Portable Devices 9 2009-11-23 Paper Presentation - Ultra Portable Devices 10 Lower bound of energy performance(II) Lower bound of energy performance(III) � Wake up tone duration � Analytical results � T sync is synchronization period � T sync is clk drift rate � trade off in determining T sync � Synchronization frequency and polling period � Best case: perfect piggybacking � Worst case: explicit synchronization � Synchronization can be quite rare (7xT data -16xT data ). � Piggybacking reduces energy by half when send data is very rare. � Energy usage in LPL increases with higher speed radio. 2009-11-23 Paper Presentation - Ultra Portable Devices 11 2009-11-23 Paper Presentation - Ultra Portable Devices 12 3
11/23/2009 Protocol implementation(I) Protocol implementation(II) � Software architecture � Efficient piggybacking of information � Tiny OS (and Mica2 motes and CC1000 radio) � Using dest. address for piggyback synchronization info. in broadcast packets � MAC functionality is broken to � physical layer (PHY) � basic CSMA layer � LPL layer � SCP layer � Port to IEEE.15.4 radio � Interaction with Tiny OS � To adapt SCP-MAC to run on IEEE.15.4 radio � CPU sleeps when radio is off � Implementation of timer with very low-jitter 2009-11-23 Paper Presentation - Ultra Portable Devices 13 2009-11-23 Paper Presentation - Ultra Portable Devices 14 Experimental evaluation(I) Experimental evaluation(II) � Optimal set up with periodic traffic � Performance with unanticipated traffic � 10 nodes in a single-hop network � Duty cycle of 3% and polling of every second � Each node generates 40B msg. periodically � Twenty 100B long broadcast message by each node � Msg. time interval is 50-300s � Msg. time interval is 50-300s � Optimal polling period from analytical model � 2009-11-23 Paper Presentation - Ultra Portable Devices 15 2009-11-23 Paper Presentation - Ultra Portable Devices 16 4
11/23/2009 Experimental evaluation(III) Summary � Performance in a multi-hop network � A new MAC protocol based on scheduled channel polling is proposed. � By optimally combining channel polling and scheduling, duty cycle � Varying traffic load reduces to 0.1% and lower. � Duty cycle of 0.3% and channel polling of 1 second � A lower bound on energy consumption is derived by analytical model. � 9-hop linear network with 10 nodes � SCP-MAC can handle bursty and varying traffic load. � Inter-packet interval is 0-10s (20 packet send, each 50B long) 2009-11-23 Paper Presentation - Ultra Portable Devices 17 2009-11-23 Paper Presentation - Ultra Portable Devices 18 5
Recommend
More recommend