Open-Source Cyber-Physical Network Infrastructure Tobias Betz - PowerPoint PPT Presentation

Chair of Network Architectures and Services Department of Informatics Technical University of Munich Open-Source Cyber-Physical Network Infrastructure Tobias Betz 2017-12-13 Chair of Network Architectures and Services Department of Informatics Technical University of Munich

Motivation CoAP RPL UDP ICMPv6 IPv6 • Open standards: 6TiSCH network stack 6LoWPAN adaptation layer 6TiSCH 6top IEEE 802.15.4e TSCH MAC IEEE 802.15.4 PHY • Open-source firmware: OpenWSN • Open hardware: OpenMote [1] T. Betz — Open-Source Cyber-Physical Network Infrastructure 2

Background — IEEE 802.15.4e TSCH TSCH: Time-Slotted Channel Hopping A → B E → B 16 channels B → C B → C B → ∗ C B D D → C 101 timeslots A E ≈ 10 ms • Energy-efficient • Synchronized timeslots • Deterministic MAC behavior • Scheduled transmissions • Robust • Channel hopping against interference / multi-path fading T. Betz — Open-Source Cyber-Physical Network Infrastructure 3

Background — 6TiSCH IETF working group: "IPv6 over TSCH mode of IEEE 802.15.4e" C CoAP RPL B D UDP ICMPv6 IPv6 A E 6LoWPAN adaptation layer A B ADD 2 [x,y] 6TiSCH 6top IEEE 802.15.4e TSCH MAC SUCCESS [x,y] IEEE 802.15.4 PHY 6top Protocol (6P) Send requests to L2 neighbors to add / delete / relocate cells in the shared schedule [2]. Allows nodes to dynamically adapt their TSCH schedules. T. Betz — Open-Source Cyber-Physical Network Infrastructure 4

Background — 6TiSCH Scheduling function Runs on each 6TiSCH node, monitors traffic requirements, and triggers 6P transactions [3]. If more bandwidth is needed, a 6P ADD request is sent including the locations of the candidate cells. The default implementation (SF0) chooses random cell locations, which seems inconsistent with deterministic MAC behavior. ⇒ Main work: make SF0 more efficient and deterministic T. Betz — Open-Source Cyber-Physical Network Infrastructure 5

New Implementation — hashSF Modifications from SF0: M.1: Use the addresses of both nodes to generate a hash that maps to a determin- istic sequence of cell locations ⇒ determinism M.2: Use spare candidate cells in 6P requests ⇒ efficiency M.3: Let the RPL parent choose cell locations ⇒ efficiency T. Betz — Open-Source Cyber-Physical Network Infrastructure 6

Modification M.1 — Hashed cell locations hash ( A � B ) = 0xb35f... 0xb3 mod ( nrChannels · nrTimeslots ) = 3 0x5f mod ( nrChannels · nrTimeslots ) = 95 ... 1 15 0 a → 0 b 5 0 c → 0 e 0 e → 0 b channelOffset 4 0 2 3 0 a → 0 b 0 a → 0 b 2 0 b → 0 d 0 a → 0 e 5 1 0 a → 0 b 6 0 0 a → 0 b 5 0 1 2 3 4 10 slotOffset T. Betz — Open-Source Cyber-Physical Network Infrastructure 7

Modification M.2 — Spare cells in 6P transactions A B ADD 1 [x,y,z] SUCCESS [y] Send a larger list of cells in the 6P requests to allow the neighbor to choose from more cell locations. ⇒ 6P requests are more likely to succeed T. Betz — Open-Source Cyber-Physical Network Infrastructure 8

Modification M.3 — Let RPL parent choose cell locations A B ADD 2 [] SUCCESS [x,y] CONFIRM [x,y] Child node has less busy schedule than parent ⇒ chosen cell locations might not be free in parent’s schedule Use 3-step 6P transactions to let RPL parent choose cells ⇒ more likely to succeed T. Betz — Open-Source Cyber-Physical Network Infrastructure 9

Measurements (1/2) n1 n2 n3 n4 • Compare chosen candidate cells for SF0 and M.1 ⇒ determinism T. Betz — Open-Source Cyber-Physical Network Infrastructure 10

Results — Chosen cell locations — SF0 5.0% 5 RPL child n2 4 3.8% 3.3% 3.3% Frequency [%] 3 2.9% 2.9% 2.9% 2.9% n1 2.5% 2.5% n2 2 n3 n4 1 133 108 171 104 120 67 141 69 174 77 0 1 2 3 4 5 6 7 8 9 10 Order of most frequent cell locations at parent n1 T. Betz — Open-Source Cyber-Physical Network Infrastructure 11

Results — Chosen cell locations — M.1 33.0% 32.2% RPL child n2 30 25.7% 25 Frequency [%] 20 n1 15 n2 10 n3 n4 6.7% 5 0.9% 0.7% 0.4% 0.2% 0.2% 1 6 9 17 12 21 5 7 3 0 1 2 3 4 5 6 7 8 9 Order of most frequent cell locations at parent n1 T. Betz — Open-Source Cyber-Physical Network Infrastructure 12

Measurements (2/2) • Compare timestamps of network formations events and ratio of denied 6P requests (SF0, M.1+M.2, M.1+M.3) ⇒ efficiency of 6P transactions n1 n2 n4 n3 T. Betz — Open-Source Cyber-Physical Network Infrastructure 13

Results — Timestamps Time spent in SF 40 30 Elapsed time [s] Median 20 SF0 1.40 s M.1, M.2 2.14 s M.1, M.3 2.30 s 10 0 SF0 M.1, M.2 M.1, M.3 T. Betz — Open-Source Cyber-Physical Network Infrastructure 14

Results — Timestamps Time until reachable 200 Elapsed time [s] 150 Median SF0 59.23 s M.1, M.2 61.91 s 100 M.1, M.3 55.58 s 50 SF0 M.1, M.2 M.1, M.3 T. Betz — Open-Source Cyber-Physical Network Infrastructure 15

Results — Ratio of successful 6P transactions 100 98.7% 88.9% 81.6% 80 Ratio of successful requests [%] 60 40 20 0 SF0 M.1, M.2 M.1, M.3 T. Betz — Open-Source Cyber-Physical Network Infrastructure 16

Conclusion • Deterministic cell locations can be achieved by hashing the nodes’ addresses • Efficiency of 6P transactions can be increased by proposing more candidate cells (M.2) or by letting the busier node choose the cells (M.3) • Faster scheduling function does not significantly affect network formation time T. Betz — Open-Source Cyber-Physical Network Infrastructure 17

Bibliography [1] WoTT OpenMote. http://wott.tlc.unipr.it/site/images/nodes/OpenMote.png. [2] Q. Wang, X. Vilajosana, and T. Watteyne. 6top Protocol (6P), 2017. https://tools.ietf.org/html/draft-ietf-6tisch-6top-protocol-07. [3] Diego Dujovne, Luigi Alfredo Grieco, Maria Rita Palattella, and Nicola Accettura. 6TiSCH 6top Scheduling Function Zero (SF0), 2017. https://tools.ietf.org/html/draft-ietf-6tisch-6top-sf0-05. [4] Simon Duquennoy, Beshr Al Nahas, Olaf Landsiedel, and Thomas Watteyne. Orchestra: Robust Mesh Networks Through Autonomously Scheduled TSCH. In Proceedings of the 13th ACM Conference on Embedded Networked Sensor Systems , SenSys ’15, pages 337–350, New York, NY, USA, 2015. ACM. T. Betz — Open-Source Cyber-Physical Network Infrastructure 18

Backup — Chosen cell locations — M.1 89.3% RPL child 84.5% n3 80 n4 60 Frequency [%] n1 40 n2 n3 n4 20 10.7% 6.0% 3.6% 2.4% 1.2% 1.2% 1.2% 1 2 11 5 17 17 33 35 11 0 1 2 3 4 5 Order of most frequent cell locations at parent n2 T. Betz — Open-Source Cyber-Physical Network Infrastructure 19

Backup — Timestamps 250 SF0 M.1, M.2 M.1, M.3 200 150 Time [s] 100 50 0 50 rebooted sync parent_found cell_added reachable Timestamp T. Betz — Open-Source Cyber-Physical Network Infrastructure 20

Backup — Timestamp differences 20 SF0 175 10 M.1, M.2 40 M.1, M.3 200 10 150 8 30 125 0 150 6 100 10 Time [s] 20 100 75 4 20 50 10 50 30 2 25 40 0 0 0 0 rebooted diff_sync diff_parent_found diff_cell_added diff_reachable T. Betz — Open-Source Cyber-Physical Network Infrastructure 21

Backup — Autonomous Scheduling Function (ASF) 6 slots slotframe #3 #1 RxTxS RxTxS #2 RxTxS RxTxS #1 #2 #3 RxTxS RxTxS #4 #4 RxTxS RxTxS Time (a) Topology (b) Rendez-vous 6 slots slotframe 6 slots slotframe #1 #1 TxS Rx TxS Rx Rx Rx #2 #2 TxS Rx Rx TxS Rx Rx TxS Rx TxS Rx #3 #3 TxS Rx TxS Rx TxS TxS #4 #4 TxS Rx TxS Rx TxS TxS Time Time (c) Receiver-based (d) Sender-based Three types of slotframes as defined in ASF . The network topology consists of four nodes. The cell types are labeled as Tx (transmit), Rx (receive), and S (shared), from [4] T. Betz — Open-Source Cyber-Physical Network Infrastructure 22