DESYNC SelfOrganizing Desynchronization and TDMA on Wireless Sensor

DESYNC: Self-Organizing Desynchronization and TDMA on Wireless Sensor Networks Julius Degesys, Ian Rose, Ankit Patel and Radhika Nagpal IPSN 20073648 황재호

Contents � Introduction � Desynchronization � Desync-TDMA � Experiment � Conclusion Algorithm

Introduction Synchronization: Sometimes desirable to perform tasks at the same time [Werner-Allen et al - Sen. Sys ‘ 05, Maróti et al - Sen. Sys ’ 04, etc. ] Perfect Desynchronization: Other times, want even distribution • No two are happening at the same time • Perfect round-robin schedule 3

Introduction - framework Single node periodically “fires” by broadcasting a message A typical starting configuration (random) A system in desynchronization ? How do we get from a random start to desynchronization?

Desynchronization Algorithm(1) DESYNC Algorithm: 1. Record firing times of phase neighbors 2. Compute the average 3. When back neighbor fires, jump towards the average Jump corresponds to a mote advancing or delaying its timer

Desynchronization Algorithm(2)

Desynchronization Algorithm(3) Remove 1 Add 3

Desynchronization Algorithm(4) Completely distributed • No central authority • No time synchronization Simple! • 10 s of lines of code • Constant memory Self-healing • Robust to disturbances • Addition/Removal of nodes

Desync-TDMA(1) Time Division Multiple Access (TDMA) 1 Protocol for broadcast channel sharing, where nodes divide time into equal slots, and each node “owns” a slot 3 2 Nice Properties • Collision-free message transmission • Fair allocation of bandwidth • High bandwidth coverage in high load 5 4 1 2 3 4 5 … …

Desync-TDMA(2) DESYNC-TDMA Algorithm: • Define slots at midpoints of firings from the round before Slot Properties • Non-overlapping • Full bandwidth coverage • Well-defined (regardless of state) • • Can always send Fairness over time

Desync-TDMA(3) Traditional TDMA DESYNC-TDMA Centralized • Negotiate slot schedules • Time synchronization required Decentralized Not Adaptive • Limited addition/removal of nodes • Can waste bandwidth Self-adapting Complicated • Separate phases, maintain neighbor identities, etc. Simple

Experiment(1) Metrics 1. Throughput 2. Message loss All-to-all network Eavesdropping Base Period = 1 sec Comparisons 1. TDMA (Theoretical) 2. CSMA (Default Tiny. OS) 3. Hybrid (Z-MAC) Traffic Model • Nodes always try to send data whenever possible

Experiment(2) 1 35 30 25 20 15 10 5 0 System is not yet desynchronized …but throughput is still high and near-constant 0. 8 0. 6 0. 4 0. 2 Normalized Throughput Average Desync Error (ms) 40 0 Measured capacity: 62. 8 kbps

Experiment(3) In a network of 8 nodes, one is removed, and 45 seconds later, 3 are added -1 +3 The “distance” from desynchronization may spike But total throughput is only slightly affected (~10% = O(1/n))

Experiment(4) Setup: 20 -node all-to-all network, 10 sending Throughput (kbps) Msg Loss (%) Fixed TDMA Tiny. OS CSMA Z-MAC DESYNCTDMA 31. 4 46. 1 41. 0 57. 9 0 57. 1 32. 4 0. 2

Conclusion & Future Work Contributions � Self-organizing TDMA schedule � Desynchronization ◦ Can generalize (e. g. ADCs, traffic intersections) ◦ Round-robin scheduling or periodic tasks Future Work � Multi-hop TDMA ◦ Simulations give evidence for convergence ◦ Doesn’t solve hidden terminal problem