Tema 6.- Redes inalámbricas Ad Hoc. Control de Energía

1. Tema 6.- Redes inalámbricas Ad Hoc. Control de Energía Introducción Trabajo Relacionado Propuestas del Grupo GRC Estudios Preeliminares Minimum Drain Rate Arquitectura Mobile Network Framework Power Management in IEEE 802.11

2. Limitaciones de las redes ad hoc [Corson99] • Limitaciones debidas al medio inalámbrico • Radio de transmisión limitado(Encaminamiento complejo) • Errores de transmisión/perdidas de paquetes • Seguridad restringida • Limitaciones debidas al carácter de las estaciones • Topología dinámicas • Energía reducida ¿ Cuales son los mayores retos en el área de las redes ad-hoc ? [Corson99], S. Corson and J. Macker, “Mobile Ad hoc Networking (MANET): Routing Protocol Performance Issues and Evaluation Considerations”, RFC 2501, January 1999.

3. Retos de las redes ad hoc • ¿Como encaminar paquetes entre estaciones? • ¿Como hacerlo de forma eficiente? • Otros retos: • Configuración • Descubrimiento de servicios • Calidad de servicio • TCP over MANETs • Seguridad y privacidad • ... “A pesar de los recientes avances en el área de la capacidad de las baterías, el Consumo de energía, es todavía un factor de diseño crítico”[Tsuda2000] [Tsuda00], Shingo Tsuda, “Battery Technologies Based on the Requirements for Applications and Future Movement”, 10th Int. Meeting on Lithium Batteries, June 2, 2000, Como (Italy).

4. 4 7 9 8 1 5 6 2 3 Capacidad de batería # 7 Motivación • Dependencia de baterías Batería limitada Capacidad de batería # 2 ¡¡ Red sin capacidad de encaminamiento !!

5. Power Management in IEEE 802.11 • Mobile devices are battery powered • Enhancement of battery life enhances network lifetime • Idle receive state dominates LAN adapter power consumption over time • Power consumption of ORiNOCO WLAN Card How can we power off during idle periods and yet maintain an active session?

6. Trabajo Relacionado: Solutions addresing the power-saving issue in MANETS: • Transmission power control [Wu2000] • To reduce interferences and improve throughput on the MAC layer • To determine the best network topology • Adjust the power tx according to the distance with the next node • Power-aware routing [Kim2003] • Use different metrics based on power consumption • Nodes’ lifetime, conections’ lifetime • Remaining battery, supported traffic • Low-power mode [IEEE80211-99] • Low power sleep modes at the physical layer • IEEE 802.11 power saving mechanishm • Bluetooth provides sniff, hold, and park low-power modes [Wu2000] S.L. Wu, Y.C. Tseng, J.P. Sheu, Intelligent medium access for mobile ad hoc networks with BusyTones and power control, IEEE Journal on Selected Areas in Communications 18 (2000) 1647–1657. [Kim2003]Dongkyun Kim, J.J. Garcia-Luna-Aceves, Katia Obraczka, Juan Carlos Cano, Pietro Manzoni, "Routing Mechanisms for Mobile Ad Hoc Networks based on the Energy Drain Rate", IEEE Transactions on Mobile Computing, Vol. 2, No. 2, April-June 2003, pp. 161-173. [IEEE80211-99]LAN MAN Standards Committee of the IEEE Computer Society, IEEE Std 802.11-1999, Wireless LAN medium access control (MAC) and physical layer (PHY) specifications,IEEE, 1999.

7. Propuestas del Grupo GRC (I) [Cano2000] J. C. Cano and P. Manzoni, “A Performance Comparison of Energy Consumption for Mobile Ad Hoc Networks Routing Protocols,'' Proceedings of the 8th IEEE/ACM MASCOTS 2000, August 2000 • Protocols evaluated in this work • DSR, AODV, DSDV and TORA • Previous performance studies • [Broch98], evaluated DSR, AODV, DSDV, and TORA • packet loss, routing message overhead and route length • [Joha99], evaluated DSR, AODV, and DSDV • packet loss, routing overhead, throughput and delay • [Das00], evaluated DSR and AODV • packet loss, delay and routing load [Broch98], J. Broch, D. A. Maltz, D. B. Johnson, Y-C. Hu, and J. Jetcheva. “A Performance Comparison of Multi-hop Wireless Ad Hoc Networks.” In Proceedings of the 4th Int. Conference on Mobile Computing and Networking (ACM MOBICOM’98), pp. 85-97, October 1998. [Joha99], P. Johansson, T. Larsson, N. Hedman, B. Mielczarek, and M. Degermark, “Scenario-based Performance Analysis of Routing Protocols for Mobile Ad-Hoc Networks,” In Proceedings of the ACM/IEEE International Conference on Mobile Computing and networking, pp: 195-206, August, 1999. [Das00], S. R. Das, C. E. Perkins, and E. M. Royer, “Performance Comparison of Two On-Demand Routing Protocols for Ad Hoc Networks,” In Proceedings of the IEEE Infocom, March 2000.

8. Modelo de consumo de energía • Energía consumida por una estación depende de: • Consumo del interfaz • Características del interfaz de red • Operación • Suministro de energía • Tiempo de transmisión del paquete • Tamaño del paquete • Ancho de banda • Lucent WaveLAN 2.4 GHz, 11Mbps • Recepción 240mA • Transmisión 280mA • V = 5 voltios • tp = (ph/2*106 + pd/11*106) • Ph = tamaño cabecera • Pd = tamaño datos E(p)= i * v * tp Etx(p)= 280mA * v * tp Erx(p) = 240mA * v * tp

9. Methodology • Basic Scenario • Mobility scenario • 25 mobile nodes • 500m x 500m • 15m/s as a maximum speed • random waypoint model [Johnson99] • Traffic model • 20 CBR sources • 4 packets/s (512 bytes each one) • Sensitivity analysis • Parameters evaluated • Mobility patterns • Traffic pattern • Node number • Area shape • Data sending pattern

10. Simulations Results: The Basic Scenario • The energy due to received traffic is NOT proportional to sent traffic • Receiving process: Receiving actual data plus overhearing • Routing energy consumption • DSR performs better that DSDV and AODV (caching mechanism) • TORA high energy consumption • Due to the IMEP periodic HELLOpackets [Corson98] [Corson98], M. S. Corson, S. Papademetriou, P. Papadopoulos, V. D. Park, and A. Qayyum, “An Internet MANET Encapsulation Protocol (IMEP) Specification,” Internet Draft, draft-ietf-manet-imep-spec-01.txt, August 1998. Work in progress.

11. Conclusions We measured and compared the energy consumption behavior of four different MANET routing protocols Methodology Basic scenario Sensitivity analysis • Results: • Pure on-demand protocols (DSR and AODV) perform better • Proactive protocols seem to perform betterin stressed scenarios • DSR normally performs better than AODV • Source routing could be a good option ... • TORA achieve highly reliability by consuming more energy than othersproposal • Future Work ...

12. Propuestas del Grupo GRC (II) [Kim2003] Dongkyun Kim, J.J. Garcia-Luna-Aceves, Katia Obraczka, Juan Carlos Cano, Pietro Manzoni, "Routing Mechanisms for Mobile Ad Hoc Networks based on the Energy Drain Rate", IEEE Transactions on Mobile Computing, Vol. 2, No. 2, April-June 2003, pp. 161-173. • Objectives at the network layer: • Select the best path to route data packets • Minimizing the total transmission power needed to route data packets • Maximizing the lifetime of all individual nodes • Power-aware route selection schemes: • MTPR (Minimum Transmission Power Routing) [Scott96] • MMBCR (Min-Max Battery Cost Routing) [Singh98] • CMMBCR (Conditional Max-Min Battery Capacity Routing) [Toh01] • MDR (Minimum Drain Rate Routing) [Kim2003] [Scott96], K. Scott and N. Bambos, "Routing and channel assignment for low power transmission in PCS," IEEE ICUPC, 1996. [Singh98], S. Singh, M. Woo, and C. S. Raghavendra, "Power-aware with Routing in Mobile Ad Hoc Networks," IEEE/ACM Mobicom, 1998. [Toh01], C.-K. Toh, "Maximum Battery Life Routing to Support Ubiquitous Mobile Computing in Wireless Ad Hoc Networks," IEEE Communications Magazine, June, 2001.

13. Minimum Total Transmission Power Routing (MTPR) • Objective • Reduce the total energy consumed along a route • MTPR Algorithm • Let the total transmission power along a route rd = n0, n1, …, nd • MTPR selects the optimal route (ro) as: • MTPR drawbacks • It makes us worry about the lifetime of each node Where T(ni, nj) denotes the energy consumed in transmitting over the hop (ni, nj). Where r* is theset of all possible routes.

14. Min-Max Battery Cost Routing (MMBCR) • Objective • Avoid the use of those routes with nodes having the least battery capacity. (It tries to make sure that no node is overused) • MMBCR Algorithm • Let ci(t) the battery capacity of node ni at time t • fi(t) = 1/ci(t) , Battery cost function of node ni • MMBCR selects the optimal route (ro) as: • MMBCR drawbacks • It does not minimize the total transmission power over a route • It tries to distribute energy consumption evenly, but can make many nodes consume their energy because of longer paths. , Route cost of a route rj = n0, n1, …, nd Where r* is theset of all possible routes.

15. Conditional Max-Min Battery Capacity Routing (CMMBCR) • Objective • Prolong the lifetime of nodes and reduce the total transmission power consumed per packet • CMMBCR Algorithm • , Route cost of a route rjwhere ci(t) is the battery capacity of node i . • Let Q the set of all possible paths between src and dst • Let A the set of paths between src and dst that holds: • Where  agiven threshold • CMMBCR route selection: • If AQ != , select a route in AQ, using MTPR protocol • Otherwise , select a route in Q using MMBCR protocol • CMMBCR drawbacks • The performance depends on the value of  • It is not easy to select  according to the two objectives above

16. Motivation • We cannot guarantee that the nodes with high remaining power will survive without power breakage, even when heavy traffic is passing through the node. j i Remaining Battery Power Remaining Battery Power(RBP) • Compare Drain Rates

17. A New Cost Function Based on Drain Rate • Traffic load information + Residual Battery Capacity • Candidate Traffic load information • Number of packets buffered in the node’s queue • It is not trivial to devise an efficient cost function that combines the buffer information with the remaining battery power. • Drain Rate (DR) • The rate at which energy gets dissipated at a given node • A new cost function using drain rate • When the remaining battery of node (i) is exhausted

18. Minimum Drain Rate (MDR) Routing • Objective • Extend both nodal battery life and connection lifetime • MDR Algorithm • Let cicost valueof node ni • MDR selects the optimal route (rM) as: • MDR drawbacks • It does not minimize the total transmission power over a route • Solution : • Conditional Version of MDR similar to CMMBCR in threshold based way , Route cost of a route rp = n0, n1, …, nd Where r* is theset of all possible routes.

19. Implementation Details • We use DSR as the base routing protocol • DSR Protocol • Route Discovery Process • Route Maintenance • We modified the Route Discovery process of DSR to implement MTPR, MMBCR and MDR • Observations • Each intermediate node inserts in the Source Route information on it energy cost • The source node will select the optimal route among all the possible routes • The source node periodically refreshes its routes and gets new routes that represent current power condition • We avoid using DSR cache optimization

20. N2 N1-N2 + N1-N2-N5 + N8 N5 N1 + N1-N3-N4-N7 + N1-N3-N4 + N1 Source N1-N3-N4 + N7 N1 + N4 N1-N3 + N1-N3-N4-N6 + N3 N6 N1-N3-N4 + Modified Route Discovery Process Destination N1-N2-N5-N8: E.Cost N1-N2-N5-N4-N7-N8: E.Cost N1-N2-N5-N4-N6-N7-N8: E.Cost N1-N3-N4-N7-N8: E.Cost N1-N3-N4-N5-N8: E.Cost N1 Route Cache RREQUEST RREPLY • Source node • Select theoptimal route • Periodicallyrefreshes its Route Cache

21. Energy Consumption Model • Lucent WaveLAN 2 Mb/s • Receiving mode (240 mA), Transmitting mode (280 mA) • Energy consumed when nodes Send/Receive a packet (5 volt) • NIC characteristics, Packet size, Bandwidth, Energy supply Energy consumed at node ni • With overhearing: • Without overhearing: Energy model • Etx(p)= (280mA*5V*size(p))/2*106 • Erx (p) = (240mA*5V*size(p))/2*106 • Eo (p) = Erx (p)

22. Resultados obtenidos son coherentes con los presentes en la literatura • Parecen indicar que este tipo de protocolos prolongan el tiempo de operación de las estaciones de una red adhoc Resultados obtenidos • Expiración de las estaciones • MMBCR utiliza la capacidad actual de las baterías para distribuir las rutas utilizadas entre las estaciones • MDR evita el consumo excesivo de energía de estaciones críticas, prolongando el tiempo de vida de las estaciones con respecto a MTPR y MMBCR

23. Simulations Results: Static Dense Network • Node expiration time • MTPR cannot guarantee the lifetime of nodes. • MMBCR evenly distributes the energy consumption among nodes with residual capacity. • Connection expiration time • MTPR allows the connection to survive longer than MMBCR due to abundant alternative routes. • MMBCR suffers the short lifetime of connections. • When considering overhearing, all approaches behave similar • MDR offers a good tradeoff among Node and Connection expiration time

24. Simulations Results: Dynamic Dense Network (10 m/s) • Node expiration time • MTPR presents the worst performance due to the concentration of traffic. • When considering overhearing, all the protocols behave almost the same. • Connection expiration time • MTPR behaves the best, because it can use many alternative routes. • MTPR has the highest variation among the connection expiration times • Again, MDR offers a good tradeoff among Node and Connection expiration time

25. Investigating the Energy Consumption • The energy consumption highly depends on overhearing activities • Overhearing problem • Techniques to reduce the energy expenditure are needed • Switching off Network Interface Cards

26. Análisis detallado de las soluciones propuestas • Los resultados obtenidos dependen del modelo de consumo de energía utilizado !!! Un análisis detallado: • El consumo del interfaz inalámbrico depende del modo de operación • Recepción, Transmisión, Idle y Sleep • El proceso de recepción incluye dos actividades • Recepción de datos propios + Sobre-escucha (overhearing)

27. Tasas de consumo de interfaces comerciales Tarjeta ORINOCO/IEEE Turbo 11 Mb PC Card Transmisión  1400 mW Recepción y overhearing  1200 mW Sleep  Varios modos de bajo consumo 75 mW Idle  ¿Especificaciones?  Estudios experimentales  1100mW [Feeney01] Tx Idle Rx Idle [Feeney01], L. Feeney and M. Nilsson, “ Investigating the Energy Consumption of a Wireless Network Interface in an Ad Hoc Networking Environment,” IEEE INFOCOM 2001, Anchorage.

28. Nuevo modelo de consumo de energía del NIC • Se ha incorporado al modelo anterior el consumo debido a las operaciones en modo Idle y de overhearing de las estaciones • ORINOCO/IEEE Turbo 11 Mb PC, 11Mbps Etx(p)= 1400 mW * tp Erx(p) = 1200 mW * tp Eover(p) = 1200 mW * tp Eidle(t)= 1100 mW * t

29. Con la tecnología 802.11 estos protocolos por si solos no permiten reducir el consumo de energía de las estaciones de la red Nuevos resultados • Expiración de las estaciones • Todos los mecanismos obtienen resultados similares

30. Conclusiones obtenidas • Tasa de consumo: 1:12:14:20 Sleep : idle : Recepción : Transmisión • Los interfaces de red se encuentran la mayor parte del tiempo en modo Idle • La única forma de reducir el consumo es utilizar técnicas basadas en la utilización del modo Sleep • Sin embargo, en una red ad hoc, una estación siempre puede ser requerida como estación encaminadora, y la utilización del modo Sleep podría afectar negativamente a las actividades de encaminamiento

31. Conclusiones obtenidas • Tasa de consumo: 1:12:14:20 Sleep : idle : Recepción : Transmisión Se requieren técnicas que integren aspectos de encaminamiento con algoritmos de reducción de consumo y algoritmos de control de topología que permitan la utilización del modo Sleep sin afectar a las actividades de encaminamiento

32. Power Management in IEEE 802.11 Standard

33. Power Management in IEEE 802.11 • Power management modes • Active mode (AM) • Power Save mode (PS) • Power management in 802.11 • In infrastructure network vs. ad hoc network • Basic Idea • AP or source hosts buffer packets for hosts in PS mode. • AP or sources send TIM periodically. • TIM = traffic indication map (a partial virtual bitmap associated with station id) • TIM is associated with beacon. • Hosts in PS mode only turn on antenna when necessary. • Hosts in PS mode only “wake up” to monitor TIM.

34. TIM types and Synchronization • TIM Types • TIM : • Transmitted with every beacon (for Unicast) • Delivery TIM (DTIM): • Transmitted less frequently (every DTIM_interval) • For sending buffered (broadcast packets) • Ad hoc TIM (ATIM): • Transmitted in ATIM-Window by stations who want to send buffered packets • Structured the same as TIM • Synchronization • Time Synchronization Function: assures AP and Power save stations are synchronized. • A timestamp used to be synchronized • TSF timer keeps running when stations are sleeping

35. Basic Idea:An Illustration Example • Each node configures its wake-up period

36. Power Saving Sequences • 802.11 stations shut down the radio transceiver and sleeping periodically to increase battery life. • During sleeping periods, access points buffer any unicast frames for sleeping stations. • These frames are announced by subsequent Beacon frames. • To retrieve buffered frames, newly awakened stations use PS-Poll frames. • Access Sequences • Immediate response • AP can respond immediately to the PS-Poll. • PS-Poll frame contains an Association ID in the Duration/ID field so AP can determine which frames were buffered for the MS. • Immediate response with fragmentation • If the buffered frame is large, it may require fragmentation. • Deferred response • After being polled, the AP may decide to respond with a simple ACK. • AP may do regular DCF activities • the PS station must remain awake until it is delivered

37. Power Management in Infrastructure Mode • All traffic for MSs must go through APs, so they are an ideal location to buffer traffic. • APs are aware of MSs’ power management state. • APs have two power management-related tasks. • Determine whether a frame should be delivered • Announce periodically which stations have frames waiting for them. • Stations only need to power up • receiver to listen to the buffer status and • transmitter to transmit polling frames after being informed.

38. Assumptions and Models • Assumptions: • TIM interval (beacon interval) and DTIM interval are known by all hosts • Requires time synchronization • Stations in PS mode are known or can be predicted. • Two Operational Models: • Under DCF (contention-based) • Under PCF (contention-free)

39. Under DCF (Infrastructure Mode) • Basic assumption: • Use CSMA/CA to access the channel • RTS, CTS, ACK, PS-Poll are used to overcome the hidden-terminal problem Operations of TIM (in DCF) • AP periodically broadcasts beacon with TIM. • Hosts in PS must wake up to check TIM. • Check for their IDs. • If found having packets buffered in AP, send PS-Poll to AP (by contention?). • Otherwise, go back to PS mode. • AP replies PS-poll with ACK. • The receiver must remain in active mode until it receives the packet. • AP uses CSMA/CA to transmit to stations.

40. Listen Interval • Listen Interval • The number of Beacon periods for which the mobile station may choose to sleep. • One of the key parameter used in estimating the resources required to support an association. • Longer listen intervals require more buffer space on the AP. • If a MS fails to check for waiting frames after each listen interval, they may be discarded without notification. • To retrieve buffered frames, MSs use PS-Poll control frames. • The buffered frames must be positively acknowledged before it is removed from buffer.

41. Traffic Indication Map • If multiple frames are buffered for a MS, then the More Data bit in the Frame Control field is set to 1. • MSs can then issue additional PS-Poll to AP until More Data bit is set to 0. • After transmitting the PS-Poll, a mobile station must remain awake until either • The polling transaction has concluded or • The bit corresponding to its AID is no longer set in the TIM • When multiple stations have buffered frames, stations use the random backoff algorithm before transmitting the PS-Poll.

42. Buffered Frame Retrieval Process for Two Stations • Station 1 has a listen interval of 2 while Station 2 has a listen interval of 3.

43. Delivering Multicast and Broadcast Frames: the Delivery TIM (DTIM) • Frames are buffered whenever any station associated is sleeping. • Buffered broadcast and multicast frames are saved used AID = 0. • APs set the first bit in the TIM to 0. • At a fixed number of Beacon intervals, a DTIM is sent. • Buffered broadcast and multicast traffic is transmitted after a DTIM Beacon. • If TIM indicates frame buffered • station sends PS-Poll and stays awake to receive data. - else station sleeps again

44. Under PCF (Infrastructure Mode) • Basic Assumption: • Point coordinator uses CF-Polling to access the channel. • AP only maintains the CF-Pollable stations. AP must poll stations in PS mode first • AP broadcasts beacon with TIM. • Hosts in PS mode checks TIM for their IDs. • If there are buffered packets in AP, the host must remain in Active Mode until being polled. • O/w, the station goes back to PS mode. • Then AP polls those PS stations. • When being polled, the station (in PS mode) sends PS-Poll to AP. • Then AP sends buffered packets to the station.

45. Buffered Frame Retrieval in PCF Beacon_ Interval TIM TIM TIM Data Poll Poll Data AP STA 1 in PS mode ACK PS-poll STA 2 in PS mode ACK PS-poll

46. Operations of DTIM in PCF • All CF-pollable stations need be in Active Mode when AP broadcasts DTIM. • Immediately after DTIM, AP sends out the buffered broadcast/multicast packets.

47. Power Management in Ad Hoc Mode • Power management is less efficient. • Far more of the burden is placed on the sender to ensure that the receiver is active. • Receiver must also be more available and cannot sleep for the same lengths of time. • ATIM: Announcement TIM • The ATIM frame is a message to keep the transceiver on because there is a pending data frame. • All stations in an IBSS listen for ATIM frames during specified periods after Beacon transmissions. • Stations that do not receive ATIM frames are free to conserve power.

48. ATIM Usage

49. PS in Ad Hoc Mode • Assumptions: • ATIM interval (beacon interval) & ATIM window are known by all hosts • Each station predicts which stations are in PS mode • ¿How, not mentioned in the standard? • The network is fully connected • Basic Method: • CSMA/CA is used to access the channel. • RTS, CTS, ACK, PS-Poll are used to overcome hidden terminal.

50. Operation of ATIM (1) • All stations should be in active mode during ATIM window. • The station which completes its backoff procedure broadcasts a beacon. • Sending beacon is based on contention. • Any beacon starts the ATIM window. • Once a beacon is heard, the rest beacons are inhibited. • In ATIM window, each source station having buffered packets to be sent contends to send out its ATIM. • If a host finds it is in the ATIM name list, • Send an ACK to the sender. • Remain in the ACTIVE mode throughout the beacon interval. • If the host is not in the name list, • it can go back to the PS mode.