1. A Dual-Radio Framework for MAC Protocol Implementation in Wireless Sensor Networks Manjunath D, Mun Choon Chan, and Ananda A L National University of Singapore ICC'2011

2. 2 Outline of This Talk  An overview of the background, problem, and the proposed solution  Analytical and experimental analysis of the proposed solution  Implementation and evaluation of our framework  Conclusions ICC'2011

3. 3 Background: Sensor Network MAC Protocols Categories  Synchronous protocols (e.g., SMAC [Infocom’02])  Asynchronous protocols  Sender-initiated techniques (e.g., BMAC [SenSys’04])  Receiver-initiated techniques (e.g., AMAC [SenSys’10])  Hybrids of synchronous and asynchronous techniques (e.g., SCP [SenSys’06]) ICC'2011

4. 4 Problem  Even after a decade of research, sensor network MAC protocols still spend significant energy in idle-listening and/or control operations  This is true even for the most recent AMAC [Sensys’10]  Such idle-listening and control operations are inevitable  These operations are part of the MAC functionality ICC'2011

5. 5 Proposed Solution Our solution is based on the two key observations  In typical sensor network MAC protocols, there are two categories of operations  Bandwidth-independent operations Durations of idle-listening and control operations are independent of the physical data rate  Bandwidth-dependent operations Durations of transmission and reception of data packets are functions of data rate ICC'2011

6. 6 Proposed Solution  It is well-known that there are two types of sensor network specific radio transceivers 1.Time-wise energy efficient transceivers 2.Bit-wise energy efficient transceivers ICC'2011 ParameterCC1000CC2420 Data rate19.2 Kbps250 Kbps Tx Power31.2 mW52.2 mW Rx/ID Power22.2 mW56.4 mW Time-wise energy efficiency (Power) Tx energy/bit1625 nJ208 nJ Rx energy/bit1156 nJ225 nJ Bit-wise energy efficiency

7. 7 Proposed Solution  Unlike existing multi-radio systems, we do not propose a new MAC protocol  We show that how a given MAC protocol can be re-implemented using dual radios so that in- evitable operations can be efficiently handled  Bandwidth-independent operations are served on time-wise energy efficient transceivers like CC1000  CC2420 like bit-wise energy efficient transceivers are used to serve bandwidth-dependent operations  We demonstrate significant energy savings by re-implementing SMAC, BMAC, and SCP ICC'2011

8. 8 Proposed Solution: Analytical Analysis  In order to motivate the necessity for dual radios, we model BMAC, the most popular sensor network MAC protocol  We analyze by comparing energy consumption of single and dual-radio BMAC protocols  We mainly consider three parameters, number of nodes, load, and packet size ICC'2011

9. 9 Proposed Solution: Analytical Analysis Power consumption vs. Number of nodes ICC'2011

10. 10 Proposed Solution: Analytical Analysis Power consumption vs. Load ICC'2011

11. 11 Proposed Solution: Analytical Analysis Power consumption vs. Packet size ICC'2011

12. 12 Proposed Solution: Experimental Analysis Methodology of the analysis  We analyze a representative protocol from each of the three categories of MAC protocols  Bandwidth-independent and bandwidth- dependent operations are analyzed separately ICC'2011

13. 13 Experimental Analysis of Bandwidth- Independent Operations ICC'2011 -60% Synchronous protocols (SMAC [Infocom’02]) Power consumption of a node with 10% duty-cycling

14. 14 Asynchronous Protocols (BMAC [SenSys’04]) Sender Receiver savings range from 37% to 46% savings is about 60%

15. 15 Hybrid Protocols (SCP [SenSys’06]) Sender Receiver savings range from 50% to 55% savings range from 69% to 81%

16. 16 Operation of RTS/CTS/DATA/ACK Exchange ICC'2011 Experimental Analysis of Bandwidth- Dependent Operations

17. 17 Dual-Radio Framework: Implementation  We re-implement BMAC and SCP by modifying their existing single-radio versions on TinyOS (1.x)  It is not necessary to re-implement SMAC separately  Its dual-radio re-implementation results in a operation similar to dual-radio SCP  Moreover, SMAC and SCP protocols share same synchronization procedure ICC'2011

18. 18 Dual-Radio Framework: Implementation  Energy saving operations are implemented on CC1000  Periodic radio wakeup, channel polling, and Tx and Rx of preambles constitute these operations in BMAC  In SCP, such operations include achieving synchronization, radio wakeup, channel polling, and Tx and Rx of wakeup tones  Data is communicated on CC2420 ICC'2011

19. 19 Dual-Radio Framework: Evaluation  Our dual-radio setup ICC'2011

20. 20 Dual-Radio Framework: Evaluation  Evaluations are carried out in two scenarios 1.Using a classical set-up of a sender and a receiver nodes 2.In a more realistic scenario of multiple nodes where not every node is in the vicinity of another  We compare dual-radio versions against single-radio versions running on CC2420 ICC'2011

21. 21 Dual-Radio Framework: Evaluation  Evaluation results on a pair of sender and receiver nodes ICC'2011 stateenergy savings Wake-up55% to 64% Idle/Rx3% to 46% Tx44% to 45% Total36% to 47% Dual-Radio BMAC stateenergy savings Wake-up43% to 57% Idle/Rx44% to 50% Tx41% to 44% Total44% to 49% Dual-Radio SCP/SMAC

22. 22 Dual-Radio Framework: Evaluation  Evaluation results of dual-radio BMAC on a setup of six nodes ICC'2011 SetupWup (mJ)Id/Rx (mJ)Tx(mJ)Total (mJ) single-radio102.333197.781569.67 dual-radio62.281178.42487.011733.97 dual-radio with power control 52.94403.62467.13 Sum of the energy consumed at all six nodes -65% -81% 4874.03 928.99

23. 23 Conclusions  Idle-listening and control operations of sensor network MAC protocols are inevitable  Our dual-radio framework is efficient in serving such unavoidable operations  The framework is generic to all the mainstream categories of sensor network MAC protocols  The framework is easy-to-implement and significant savings of up to 81% is being observed ICC'2011