1. On Detecting Termination in Cognitive Radio Networks Shantanu Sharma 1 and Awadhesh Kumar Singh 2 1 Ben-Gurion University of the Negev, Israel 2 National Institute of Technology, Kurukshetra, India

2. Outline Introduction Problem Statement and Our Contribution T-CRAN Protocol An Example Conclusion 2

3. Cognitive Radio Networks (CRNs) – A collection of heterogeneous cognitive radio nodes (or processors), called secondary users – The cognitive radio nodes (CRs) have sufficient computing power and power backup to operate on multiple heterogeneous channels (or frequency bands) in the absence of the licensed user(s), termed as primary user(s), of the respective bands – CRs have learning, efficiency, intelligence, reliability, and adaptively capability to scan and operate on different channels Introduction 3

4. Available Channel – A channel that is not currently occupied by a primary user is called an available channel Neighboring Nodes – Any two nodes that are in the transmission range of each other and tuned to a common available channel during an identical time interval Introduction 4

5. Termination Detection – A termination detection (TD) protocol is used to announce termination of a normal computation or an underlying computation TD is not a trivial task Two properties of a TD protocol – No false termination detection (safety) – Eventual termination detection (liveness) Introduction 5

6. Active and Passive nodes – The nodes in the active state are called active nodes – The nodes in the passive state are called passive nodes – The active nodes execute an assigned computation, and usually, after completion of the computation, they become passive – A passive node can become active on reception of a message from an active node Introduction 6

7. Two Types of TD Protocol Initiation Delayed initiation Introduction 7 Concurrent initiation

8. Why TD is challenging in CRNs – Network structure and communication links – Reaction to a communication link break – No definitive logical structure Introduction 8

9. State-of-the-Art Y.-C. Tseng and C.-C. Tan. Termination detection protocols for mobile distributed systems. IEEE Trans. Parallel Distrib. Syst., 12(6):558–566, 2001 H. Kurian, A. Rakshit, and G. Singh. Detecting termination in pervasive sensor networks. In ISADS, pages 323–332, 2009. P. Johnson and N. Mittal. A distributed termination detection algorithm for dynamic asynchronous systems. In ICDCS, pages 343–351, 2009 S. Katiyar and S. Karmakar. A simple scheme for termination detection in delay tolerant networks. In ICCSN, pages 478–482, 2011. 9

10. State-of-the-Art Why the existing protocols are unable to work in CRNs – Existence of an initiator node until termination declaration – Work on a single pre-decided channel – Do not consider the presence of some special users (like primary users) 10

11. Outline Introduction Problem Statement and Our Contribution T-CRAN Protocol An Example Conclusion 11

12. A termination detection protocol for cognitive radio networks Problem Statement 12

13. T-CRAN : Termination detection protocol for Cognitive Radio Networks – Based on credit distribution and aggregation Virtual tree-like structure – A node may surrender its credit to any node (not necessarily to its parent node) – Not mandatory for the initiator of the protocol to stay involved until the termination of the computation. Our Contribution 13

14. Outline Introduction Problem Statement and Our Contribution T-CRAN Protocol An Example Conclusion 14

15. Credit distribution and aggregation based 3-phase protocol A node initiates a computation and the T-CRAN protocol2 with a fixed credit value, C, and such a node is called the chief executive node, C E C E may distribute the computation among its neighboring nodes, called the child nodes When a node finishes its computation, the node’s state becomes passive, and the node surrenders its credit T-CRAN Protocol 15

16. Credit Distribution – Step 1: Initiation and distribution of a computation and the T-CRAN protocol – Step 2: Reception of a COMputation message at a passive node – Step 3: Reception of COMputation messages at an active node T-CRAN: Phase 1 16

17. Credit Aggregation – Step 4: Credit surrendering by active nodes C E can also surrender its credit This a beauty of the protocol – Step 5: Three-way handshake T-CRAN: Phase 2 17 Virtual tree-like structure No need of an identical root node

18. Termination Detection – Step 6: Termination announcement T-CRAN: Phase 3 18

19. Starts concurrently with Phase 1 The appearance of a PU can be visualized similar to the network partitioning – Primary user(s) affected CRN (CRN P ) – Non-primary user(s) affected CRN (CRN N ) T-CRAN: Primary User Detection 19

20. Step 7: Node failure due to the appearance of PUs – An affected node, CR j, is detected by all the neighboring nodes – All the neighboring nodes of CR j inform C E about such a situation using a PaN message Step 8: Recovery of the affected nodes – Once an affected node, CR j, becomes a non-affected node, CR j informs C E and all its neighboring nodes, whose states are active – C E first checks whether the computation has terminated – If not, then C E informs CR j about the ongoing computation (with CR j ’s credit value) T-CRAN: Primary User Detection 20

21. Strong termination – Passive state of all the CRs + no in-transit message Weak termination – Two CRNs, namely CRN P and CRN N + no in-transit message Local termination – Termination of the computation at a node Global termination – Termination of the computation at all the nodes. We like to have: global weak termination or global strong termination T-CRAN: Termination Detection 21

22. Outline Introduction Problem Statement and Our Contribution T-CRAN Protocol An Example Conclusion 22

23. 6 nodes in the system Local channel set for every node is given in Table, where boldface characters show the currently tuned channel at the respective nodes An Example 23 Local Channel SetsNodes 2, 3, 5 1 3, 5, 6, 9 2 5 3 5 4 5, 7, 9 5 5, 9 6

24. An Example 24 First consider without the appearance of a PU Protocol initiation

25. An Example 25 First consider without the appearance of a PU Credit distribution

26. An Example 26 First consider without the appearance of a PU Credit surrender by 5

27. An Example 27 First consider without the appearance of a PU Credit surrender by 3

28. An Example 28 First consider without the appearance of a PU Credit surrender by 6

29. An Example 29 First consider without the appearance of a PU Credit surrender by 4

30. An Example 30 First consider without the appearance of a PU Credit surrender by the root node

31. An Example 31 Now consider the presence of a PU

32. An Example 32 Now consider the presence of a PU

33. An Example 33 Now consider the presence of a PU

34. An Example 34 Now consider the presence of a PU

35. Outline Introduction Problem Statement and Our Contribution T-CRAN Protocol An Example Conclusion 35

36. Conclusion T-CRAN Protocol – A termination detection protocol for an asynchronous multi- hop cognitive radio networks – Based on credit distribution and aggregation approach – Capable enough to work on heterogeneous channels – Can also handle multiple computations simultaneously – Can also be implemented in dynamic networks, e.g., cellular, mobile ad hoc networks, and vehicular ad hoc networks Virtual tree-like structure – A node may surrender its credit to any node (not necessarily to its parent node) – Reduces the waiting time to announce termination – Not mandatory for the initiator of the protocol to stay involved until the termination of the computation 36

37. Shantanu Sharma 1 and Awadhesh K. Singh 2 1 Department of Computer Science, Ben-Gurion University of the Negev, Israel to_shantanusharma@ieee.org 2 Department of Computer Engineering, National Institute of Technology Kurukshetra, India aksinreck@ieee.org Presentation is available at http://www.cs.bgu.ac.il/~sharmas/publication.html