90 likes | 304 Views
Span. Goals. Minimize energy consumption Wireless interface is largest power drain* Maximize OFF time Minimize end-to-end delay No centralized controller Capacity should not be diminished Generalizable to different link layers Interoperable with routing system. Span Design.
E N D
Goals • Minimize energy consumption • Wireless interface is largest power drain* • Maximize OFF time • Minimize end-to-end delay • No centralized controller • Capacity should not be diminished • Generalizable to different link layers • Interoperable with routing system
Span Design • Randomized,distributed algorithm • Nodes decide when to • Sleep • Receive packets (idle) • Forward packets (coordinate) • Forms backbone connected dominating set of coordinators • Geographic routing • Periodic HELLO messages • Node’s status, coordinators, neighbors • Operates under the routing layer, and above the MAC and physical layers * • Affects routing process • Utilizes link-layer power saving features of MAC
Coordinator announcement • Algorithm: • If: 2 neighbors can’t reach each other • Then: • calculate randomized backoff delay • Delay* • If no one else has become a coordinator, become coordinator • Sensitivity: • Number of nodes that could become connected • Current energy level • Round trip delay over wireless link
Coordinator withdrawal • If: • all pairs of neighbors are inter-reachable • OR have been a coordinator for a long time* • Then: • Mark self as “tentative” • Delay* • Withdraw if another node announces • Sensitivity: • Energy level • Round trip delay
Integration with 802.11 • Power Saving Mode • Beacons synchronize nodes • Traffic can be advertised during ATIM window • Span’s Enhancements: • No advertisements between coordinators • Packets routed via non-coordinator nodes during the Advertised Traffic Window *
Results: Capacity Preservation • Slight increase in latency over 802.11 • Large decrease in loss over 802.11 PSM* • ATIM window in 802.11 PSM not big enough • 802.11 without PSM has lower loss • Span uses a smaller ATIM window • Span decreases utilization due to backoff • Span reduces number of voids in geographic routing • Coordinators are unlikely to occur at edge of void
Results: Coordinator Election • Compared to hexagonal grid layout • Span elects more nodes than this: • Variation of node density • Energy constraints forces rotation
Results: Energy Consumption • Span saves considerable energy over 802.11 PSM • 802.11 PSM loses efficiency due to broadcasts in geographic forwarding • Energy savings increase sublinearly with the node density* • Node lifetime increases by about 2x