Chapter 6 Time Synchronization

1. Chapter 6Time Synchronization

2. Outline • 6.1. The Problems of Time Synchronization • 6.2. Protocols Based on Sender/Receiver Synchronization • Network Time Protocol (NTP) • Timing-sync Protocol for Sensor Networks (TPSN) • Flooding Time Synchronization Protocol (FTSP) • 6.2.4. Ratio-based time Synchronization Protocol (RSP) • 6.3. Protocols Based on Receiver/Receiver Synchronization • Reference Broadcast Synchronization (RBS) • Hierarchy Referencing Time Synchronization (HRTS) • 6.4. Summary Jang Ping Sheu

3. 6.1 The Problems of Time Synchronization • Why Need for Time Synchronization (TS)? • Many of the applications of WSN needs the event with time stamp • Ordering of the samples for reporting • Events are reported by multiple nodes • When WSN is energy save enabled, it need all nodes to be in sync in order to be in Idle or Active mode • Medium Access Layer (MAC) Scheduling (ex: TDMA) • Order of messages may change while transmission Jang Ping Sheu

4. Sources of Inaccuracies • A local software clock of node i at time tLi(t) = qiHi(t) + fi • Hi(t): hardware clock of node i at time t • qi :clock drift rate of node i • fi :phase shift of node i • Actual oscillators have random deviations from nominal frequency (drift/skew) • additional pulses or lost pulses over the time of one million pulses at nominal rate (ppm) • Oscillator frequency is time variable • Long-term variation: oscillator aging • Short-term variation: environment (temperature, pressure, supply voltage, ...) Jang Ping Sheu

5. General Properties of TS Algorithms • Physical time vs. logical time • External vs. internal synchronization • Algorithms may or may not require TS with external time scale like UTC • Global vs. local algorithms • Keep all nodes of a WSN synchronized or only a local neighborhood? • Absolute vs. relative time • Many applications need only accurate time difference and it would be sufficient to estimate the drift instead of phase offset Jang Ping Sheu

6. General Properties of TS Algorithms • Hardware vs. software-based mechanisms • A GPS receiver would be a hardware solution, but often too heavy weight/costly/energy-consuming in WSN nodes, and in addition a line-of-sight to at least four satellites is required • A-priori vs. a-posteriori synchronization • Is TS achieved before or after an interesting event?  Post-facto synchronization: is triggered by an external event • Deterministic vs. stochastic precision bounds • Local clock update discipline • No backward jumps of local clocks • No sudden jumps Jang Ping Sheu

7. Performance Metrics and Fundamental Structure • Metrics: • Precision: maximum synchronization error for deterministic algorithms, mean error /stddev /quantiles for stochastic ones • Energy costs, e.g. # of exchanged packets, computational costs • Memory requirements • Fault tolerance: what happens when nodes die? Jang Ping Sheu

8. Fundamental Building Blocks of TS Algorithms • Resynchronization event detection block: • When to trigger a time synchronization round? • Remote clock estimation block: • Figuring out the other nodes clocks with the help of exchanging packets • Clock correction block: • Compute adjustments for own local clock based on remote clock estimation • Synchronization mesh setup block: • Figure out which node synchronizes with each other in a multi-hop network Jang Ping Sheu

9. Constraints for TS in WSNs • Scale to large networks of unreliable nodes • Quite diverse precision requirements, • from micro seconds to tens of seconds • Use of extra hardware is mostly not an option • Low mobility • Often there are no fixed upper bounds on packet delivery delay • Negligible propagation delay between neighboring nodes is negligible • Manual node configuration is not an option Jang Ping Sheu

10. 6.2 Protocols Based on Sender/Receiver Synchronization • In this kind of protocols, a receiver synchronizes to the clock of a sender • The classical Network Time Protocol (NTP) belongs to this class • We have to consider two steps: Pair-wise synchronization • How does a single receiver synchronize to a single sender? • Network wide synchronization • How to figure out who synchronizes with whom to keep the whole network / parts of it synchronized? Jang Ping Sheu

11. Network Time Protocol (NTP) • Synchronizing Physical Clocks • Computer Clocks in distributed system not in consistent • Need to synchronize clocks • External synchronization (ES) • Synchronized with an external reliable time source S • |S - C| < D, where C is computer’s clock • Internal synchronization (IS) • Synchronized with other computer in the distributed system • | Ci - Cj| < D • IS does not imply ES • Clock Ci and Cj may drift together • ES implies IS • Within bound 2D Jang Ping Sheu

12. Network Time Protocol (NTP) • Distributed System Type • Synchronous distributed system • Known upper bound on transmission delay • Simplifies synchronization • One process p1 sends its local time t to process p2 in a message m • p2 could set its clock to t + Ttrans , where Ttrans is transmission delay from p1 to p2 • Ttrans is unknown but min≤Ttrans≤max • Set clock to t + (max - min)/2 then skew ≤ (max - min)/2 • Asynchronous distributed system • Internet is asynchronous system • Ttrans = min + x where x≥ 0 Jang Ping Sheu

13. Network Time Protocol (NTP) Primary servers are connected to UTC sources 1 Secondary servers are synchronized to primary servers Synchronization subnet - lowest level servers in users’ computers 2 2 3 3 3 Reliability from redundant paths, scalable, authenticates time sources • A time service for the Internet - synchronizes clients to UTC (Coordinated Universal Time) Jang Ping Sheu

14. Network Time Protocol (NTP) Server B Ti-2 Ti-1 Time m m' Time Server A Ti-3 Ti • Messages exchanged between a pair of NTP peers • All modes use UDP • Each message bears timestamps of recent events: • Local time of Send and Receive of previous message • Local time of Send of current message • Recipient notes the time of receipt ( we have Ti-3, Ti-2, Ti-1, Ti) • In symmetric mode there can be a non-negligible delay between messages Jang Ping Sheu

15. Network Time Protocol (NTP) • Accuracy of NTP • For each pair of messages between two servers, • NTP estimates an offset oi between the two clocks and a delay di (total time for the two messages, which take t and t’) • Ti-2 = Ti-3+ t + oi(1) and Ti = Ti-1+ t’ – oi(2) • This gives us (by adding the equations (1) and (2)): • di = t + t’ = Ti-2 - Ti-3 + Ti - Ti-1 • Also (by subtracting the equations (1) – (2)): • oi = o + (t’ - t )/2 where o = (Ti-2 - Ti-3 + Ti-1 - Ti)/2 • Using the fact that t, t’ > 0 it can be shown that • o - di /2 ≤ oi ≤ o + di /2 . • Thus oi is an estimate of the offset and di is a measure of the delay Jang Ping Sheu

16. Network Time Protocol (NTP) • Techniques to Improve Accuracy • NTP servers filter pairs <oi, di>, estimating reliability from variation, allowing them to select peers • High variability in successive pairs implies unreliable data • Accuracy depends on the delay between the NTP servers • Accuracy of 10s of millisecs over Internet paths (1 on LANs) • Peer selection • Lower layer peer favoured over higher layer server • Peer with lower synchronization imprecision is preferred • Synchronization imprecision is the sum of variability in data from the server to the root Jang Ping Sheu

17. LTS – Lightweight TS • Overall goal: • Synchronize the clocks of all sensor nodes of a subset of nodes to one reference clock (e.g. equipped with GPS receivers) • Considers only phase shifts • Does not try to correct different drift rates Jang Ping Sheu

18. LTS – Lightweight TS • Two components: • Pair-wise synchronization: • based on sender/receiver technique • Network wide synchronization: • Minimum-height spanning tree construction with reference node as root Jang Ping Sheu

19. LTS – Pairwise Synchronization

20. LTS – Pair-wise Synchronization • Assumptions: • Node i’s original aim is to estimate the true offset O = Δ(t1) = Li(t1) – Lj(t1), where Li(tj) is the local software clock of node i at time tj • During the whole process the drift is negligible  the algorithm in fact estimates Δ(t5) and assumes Δ(t5) = Δ(t1) • There is one propagation delay τ and one packet transmission delay tp between nodes i and j Jang Ping Sheu

21. Li(t5)

22. Li(t5) t5 >= t1+ τ+tp where τ :propagation delay tp :packet transmission time Jang Ping Sheu

23. t5 <= t8- τ- tp where τ :propagation delay tp:packet transmission time Li(t5)

24. The uncertainty is in the interval [Li(t1) +τ+ tp, Li(t8) - τ – tp – (Lj(t6) – Lj(t5)] Li(t5)

25. LTS – Pair-wise Synchronization • Under the assumption that the remaining uncertainty is allocated equally to both i and j, node i can estimate Li(t5) as This exchange takes two packets. If node j should also learn about the offset, a third packet is needed from i to j carrying O Jang Ping Sheu

26. LTS – Pair-wise Synchronization • Sources of inaccuracies: • Medium access delay • Interrupt latencies upon receiving packets • Delays between packet interrupts and time-stamping operation • Delay in operating system and protocol stack Jang Ping Sheu

27. Jang Ping Sheu

28. LTS – Pair-wise Synchronization • Improvements: • Let node i timestamp its packet after the MAC delay, immediately before transmitting the first bit • This would remove the uncertainty due to i’s operating system, protocol stack and the MAC delay!! • Let node j timestamp receive packets as early as possible, e.g. in the interrupt routine • This removes the delay between packet interrupts and time-stamping from the uncertainty, and leaves only interrupt latencies Jang Ping Sheu

29. Number of trials Pair-wise difference in packet reception time (μsec) LTS – Pairwise Synchronization – Error Analysis Jang Ping Sheu

30. LTS – Network Wide Synchronization • This way a spanning tree is created • But one should not allow arbitrary spanning trees • Consider a node i with hop distance hi to the root node • Assume that: • all synchronization errors are independent • Hence, a minimum spanning tree minimizes synchronization errors Jang Ping Sheu

31. Timing-sync Protocol for Sensor Networks (TPSN) • Introduction • We present a Timing-sync Protocol for Sensor Networks (TPSN) that works on the conventional approach of sender-receiver synchronization • Pair-wise-protocol: time-stamping at node i happens immediately before first bit appears on the medium, and time-stamping at node j happens in interrupt routine Jang Ping Sheu

32. Timing-sync Protocol for Sensor Networks (TPSN) • Network Model • The network is “always-on” • Every node maintains 16-bit register as clock • Sensor has unique ID • Build hierarchical topology for the network • Node at level i can connect with at least one node at level i-1 Jang Ping Sheu

33. Timing-sync Protocol for Sensor Networks (TPSN) • Level discovery Phase • Trivial • Synchronization Phase • Pair-wise sync is performed along the edge of hierarchical structure Jang Ping Sheu

34. Timing-sync Protocol for Sensor Networks (TPSN) • Level discovery Phase • The root node is assigned a level 0and it initiates this phase by broadcasting a level_discovery packet • level_discovery packet contains the identity and the level of the sender • The immediate neighbors of the root node receive this packet and assign themselves a level (level = level +1) • This process is continued and eventually every node in the network is assigned a level. On being assigned a level, a node neglects any such future packets. This makes sure that no flooding congestion takes place in this phase Jang Ping Sheu

35. Timing-sync Protocol for Sensor Networks (TPSN) • Synchronization Phase • T1: A is sender, starting sync by sending synchronization_pulse packet to B • T2 = T1 + Δ + d, where • Δ is the clock offset • d is propagation delay • T3: B replies acknowledgement containing T1, T2, T3 • T4: A receives ack. and we have T4 = T3 – Δ + d. • So: • Δ = [(T2 － T1) － (T4 － T3)] / 2 • d = [(T2 － T1) ＋ (T4 － T3)] / 2 Jang Ping Sheu

36. Timing-sync Protocol for Sensor Networks (TPSN) • Synchronization Phase T1: A is sender, starting sync by sending synchronization_pulse packet to B with timestamp T1 T2: B receive the synchronization _pulse packet and timestamping immediately T3: B replies acknowledgement containing T1,T2,T3 T4: A receives an Ack and gets timestamp T1,T2,T3 T2 B At time T4 At time T1 At time T2 At timeT3 T1 A T4 Jang Ping Sheu

37. Timing-sync Protocol for Sensor Networks (TPSN) • Simulation and Comparison Jang Ping Sheu

38. Timing-sync Protocol for Sensor Networks (TPSN) • Simulation and Comparison Jang Ping Sheu

39. Flooding Time Synchronization Protocol (FTSP) Jang Ping Sheu

40. Flooding Time Synchronization Protocol (FTSP) • Introduction • The FTSP synchronizes the time to possibly multiple receivers utilizing a single radio message • Linear regression is used in FTSP to compensate for clock drift Jang Ping Sheu

41. Flooding Time Synchronization Protocol (FTSP) • Network Model • Every node in the network has a unique ID • Each synchronization message contains three fields: • TimeStamp • RootID • SeqNum • The node with the smallest ID will be only one root in the whole network Jang Ping Sheu

42. Flooding Time Synchronization Protocol (FTSP) • The root election phase • FTSP utilizes a simple election process based on unique node IDs • Synchronization phase Jang Ping Sheu

43. Flooding Time Synchronization Protocol (FTSP) • The root election phase • When a node does not receive new time synchronization messages for a number of message broadcast periods • The node declares itself to be the root • Whenever a node receives a message, the node with higher IDs give up being root • Eventually there will be only one root Jang Ping Sheu

44. Flooding Time Synchronization Protocol (FTSP) • Synchronization phase • Root and synchronized node broadcast synchronization message • Nodes receive synchronization message from root or synchronized node • When a node collects enough synchronization message, it estimates the offset and becomes synchronized node Jang Ping Sheu

45. FTSP Time Synchronization • Local and global time • Each node maintains both a local and a global time. • Both the clock offset and skew between the local and global clocks are estimated using linear regression • Hierarchy • Global time is synchronized to the local time of an elected leader • Utilizes asynchronous diffusion: each node sends one synchronization msg per 30 seconds, constant network load • Sequence number, incremented only be the elected leader

46. FTSP Time Synchronization • Robustness • If leader fails, new leader is elected automatically. The new leader keeps the offset and skew of the old global time • Fault tolerant: nodes can enter and leave the network, links can fail, nodes can be mobile, topology can change • Platform independent: uses the time stamping module

47. Flooding Time Synchronization Protocol (FTSP) Timestamp Timestamp Timestamp rootID rootID rootID seqNum seqNum seqNum Root A B C Jang Ping Sheu

48. 1st leader 2nd leader FTSP Experimental Evaluation topology: avg. error: 1.6 μs max. error: 6.1 μs per hop 50% turned off 1st leader turned off random nodes turned off/on all turned back on all turned on

49. Ratio-based Time Synchronization Protocol (RSP) Jang Ping Sheu

50. Ratio-based Time Synchronization Protocol (RSP) • The RSP use two synchronization messages to synchronize the clock of the receiver with that of sender • The RSP also can extend to multi-hop synchronization • The nodes in the wireless sensor network construct a tree structure and the root of this tree is the synchronization root • The global time of the root is flooding out to the nodes through the tree structure Jang Ping Sheu