Medium Access Control Protocols Wireless Networking Medium Access Control Protocols

Topics Contention-based channel access Scheduled channel access IEEE 802.11 as the main example, plus others Basics and focus on algorithms in class; readings for more details on protocol and mechanisms Scheduled channel access Directional antennas Contention-based and scheduled MACs Multiple channels Modeling and performance analysis Spring 2006 UCSC CMPE257

Channel Access Schemes Contention based schemes ALOHA, CSMA/CA (FAMA, MACA, MACAW, IEEE 802.11) : with/without RTS/CTS handshakes. Difficulties: Throughput, fairness, multipoint communication support, QoS, synchronization. Scheduled schemes FDMA/TDMA/CDMA in multi-hop networks: graph coloring problem — UxDMA. Topology independent scheduling Topology-dependent scheduling Difficulties: Efficiency vs consistent state information for scheduling; synchronization Spring 2006 UCSC CMPE257

IEEE 802.11 Standard PHY/MAC standard for wireless LANs First standardized in 1997 Meet great success starting in 1999 Several working groups IEEE 802.11a: Operates in the 5GHz band; provides up to 54 Mbps; uses OFDM 802.11b: Operates in 2.4 GHz band; provides 1,2, 5.5 and (nearly always) 11 Mbps; uses high rate DSSS 802.11g: Operates in 2.4GHz band; provides up to 54 Mbps; uses OFDM 802.11e: Quality of service (QoS) enhancement (still active) 802.11i: Security enhancement 802.11s: Mesh-networking support Spring 2006 UCSC CMPE257

Requirements Single MAC to support multiple PHYs PHYs: Handle ``hidden terminal’’ problem Provisions time-bounded service PHYs: Direct sequence Frequency hopping Infrared (never implemented) Spring 2006 UCSC CMPE257

Architecture Infrastructure mode Ad hoc mode Basic Service Set (BSS) Access Point (AP) and stations (STA) take different roles Distribution system (DS) interconnect multiple BSSs to form a single network (not specified in the standard). Ad hoc mode Independent Basic Service Set (IBSS) Single-hop (the standard makes this assumption either explicitly or implicitly) Spring 2006 UCSC CMPE257

Access Points Stations select an AP and “associate” with it Support roaming (not part of the standard) Provide other functions time synchronization (beaconing) power management support point coordination function (PCF) Traffic typically (but not always) flows through AP direct communication possible (in IEEE 802.11e) Spring 2006 UCSC CMPE257 13

802.11 Protocol Entities LLC MAC Sublayer MAC PLCP Sublayer PHY MAC Layer Management MAC Sublayer MAC Station Management PHY Layer Management Glossary: PLCP: Physical Layer Convergence Procedure PMD: Physical Medium Dependent PLCP Sublayer PHY PMD Sublayer Spring 2006 UCSC CMPE257

802.11 Protocol Architecture MAC Entity basic access mechanism fragmentation encryption MAC Layer Management Entity synchronization power management roaming MAC MIB Physical Layer Convergence Protocol (PLCP) PHY-specific, supports common PHY SAP provides Clear Channel Assessment (CCA) signal (carrier sense) SAP: Service access point Spring 2006 UCSC CMPE257 8

802.11MAC in Detail Channel access mechanism Distributed Coordination Function (DCF) Carrier sense multiple access (CSMA) with immediate MAC-level ACK RTS/CTS exchange (optional) Point Coordination Function (PCF) Polled access through AP and distributed access Contention-free period (CFP) and contention period (CP) Seldom implemented in practice Synchronization and power management Spring 2006 UCSC CMPE257 15

CSMA/CA in 802.11 Backoff-Window Next Frame Free access when medium is free longer than DIFS DIFS Contention Window PIFS DIFS SIFS Busy Medium Backoff-Window Next Frame Slot time Defer Access Select Slot and Decrement Backoff as long as medium is idle. Channel priorities based on “free channel intervals”: SIFS (Short InterFrame Spacing): Used to give nodes in an ongoing dialogue the chance to transmit. Example is RTS-CTS handshake, ACK to data, or consecutive data packets. DIFS (DCF InterFrame Spacing ): Any station may attempt to transmit after DIFS elapses. PIFS (PCF InterFrame Spacing): Base station can send poll or beacon if no station accesses channel after PIFS interval. EIFS (Extended InterFrame Spacing): Longer than DIFS and used for base station to report a bad frame DIFS: Distributed Inter-Frame Space SIFS: Short Inter-Frame Space PIFS: Point Inter-Frame Space Spring 2006 UCSC CMPE257 5

CSMA/CA in 802.11 Reduce collision probability where mostly needed. Free access when medium is free longer than DIFS DIFS Contention Window PIFS DIFS SIFS Busy Medium Backoff-Window Next Frame Slot time Defer Access Select Slot and Decrement Backoff as long as medium is idle. Reduce collision probability where mostly needed. Stations are waiting for medium to become free. Select Random Backoff after a Defer, resolving contention to avoid collisions. Backoff algorithm stable at high loads. Exponential Backoff window increases for retransmissions. Backoff timer elapses only when medium is idle. Implement different fixed priority levels. To allow immediate responses and PCF coexistence. DIFS: Distributed Inter-Frame Space SIFS: Short Inter-Frame Space PIFS: Point Inter-Frame Space Spring 2006 UCSC CMPE257 5

CSMA/CA + ACK Defer access based on Carrier Sense. DIFS Src Data SIFS Dest Ack DIFS Contention Window Other Next MPDU Defer Access Backoff after Defer Defer access based on Carrier Sense. CCA from PHY and Virtual Carrier Sense state. Direct access when medium is sensed free longer than DIFS, otherwise defer and backoff. Receiver of directed frames to return an ACK immediately when CRC correct. When no ACK received then retransmit frame after a random backoff (up to maximum limit). Spring 2006 UCSC CMPE257 6

RTS/CTS Based Access DIFS Src RTS Data Dest CTS Ack CW Other NAV (RTS) Next MPDU NAV (CTS) Defer Access Backoff after Defer Duration field in RTS and CTS frames distribute Medium Reservation information which is stored in a Net Allocation Vector (NAV). Defer on either NAV or "CCA" indicating Medium Busy. Use of RTS / CTS is optional but must be implemented. Use is controlled by a RTS_Threshold parameter per station. To limit overhead for short frames. Spring 2006 UCSC CMPE257 9

Frame Format MAC Header format differs per type Bytes: 2 2 6 6 6 2 6 0-2312 4 Frame Duration Sequence Frame Addr 1 Addr 2 Addr 3 Addr 4 CRC Control ID Control Body 802.11 MAC Header Bits: 2 2 4 1 1 1 1 1 1 1 1 Protocol To From More Pwr More Type SubType Retry WEP O Version DS DS Frag Mgt Data Frame Control Field MAC Header format differs per type (control, management, or data frames). The subtype indicates RTS, CTS, ACK, etc. To/From DS bits indicate that frame comes from/goes to intercell distribution system More Frag bit indicates that more fragments for frame follow Retry bit states that frame is a retransmission Pwr Mgt bit is used by station to put nodes in/out of sleep state More data bit states that sender has more frames for receiver WEP states if wired equivalent privacy algorithm was used to encrypt frame O bit states that frames with bit set must be processed in order. Spring 2006 UCSC CMPE257 15

Frame Format 802.11 MAC Header Bytes: 2 2 6 6 6 2 6 0-2312 4 Frame Duration Sequence Addr 1 Addr 2 Addr 3 Addr 4 Payload CRC Control Control Duration states how long frame and ack will occupy channel. Present in control frames and used for NAV Four addresses are in standard IEEE 802 format Sequence control is used to number frames; 16 bits identify frame, and 4 the fragment in a frame. Control frames have only 1 or 2 addresses, no data and no sequence control fields. Spring 2006 UCSC CMPE257 15

Address Field Description Addr 1 = All stations filter on this address. Addr 2 = Transmitter Address (TA) Identifies transmitter to address the ACK frame to. Addr 3 = Dependent on To and From DS bits. Addr 4 = Only needed to identify the original source of WDS (Wireless Distribution System) frames. Spring 2006 UCSC CMPE257 16

Comments on CSMA/CA IEEE 802.11 cannot avoid collisions of data packets (see [FAMA97]). CSMA/CA works fine when hidden terminals are very few. For most single-hop wireless LANs, RTS/CTS is not useful (turned off by default in practice) Spatial reuse is reduced in multi-hop networks Why ACK each frame? Why “sender initiated”? Where does the use of “free channel” intervals come from? Is RTS-CTS inherently good or bad? Is synchronization in MANETs important? Spring 2006 UCSC CMPE257 15

Synchronization in 802.11 Timing Synchronization Function (TSF) Used for Power Management Beacons sent around well known intervals All station timers in BSS are synchronized Used for Point Coordination Timing TSF Timer used to predict start of Contention Free burst Used for hop timing for frequency hoping (FH) PHY TSF timer used to time dwell interval All Stations are synchronized, so they hop at same time. Spring 2006 UCSC CMPE257 21

Synchronization Approach All stations maintain a local timer. TSF goals: Keep timers from all stations in synch AP to control timing in infrastructure networks Distributed function for Independent BSS Timing conveyed by periodic beacon transmissions Beacons contain timestamp for the entire BSS Timestamp from beacons used to calibrate local clocks Not required to hear every beacon to stay in synch Beacons contain other management information also used for Power Management and roaming Spring 2006 UCSC CMPE257 22

Infrastructure Beacon Generation Beacon Interval "Actual time" stamp in Beacon X X X X Time Axis Beacon Busy Medium Simple for infrastructure networks, because AP is the only one sending beacons. Beacons scheduled at beacon intervals. Transmission may be delayed by CSMA deferral. subsequent transmissions at expected beacon interval, not relative to last beacon transmission next beacon sent at Target Beacon Transmission Time (TBTT) Timestamp contains timer value at transmit time. Spring 2006 UCSC CMPE257 23

TSF in Ad Hoc Networks Beacon generation in IBSS is distributed and peer-to-peer process aimed at beacon avoidance first, and adopting the most current time second Each TBTT, node computes a random delay uniformly distributed in [0, (2aCWmin)(aSlotTime)], e.g., 31x20us for DSSS Node waits for random delay period Node cancels its pending beacon and remaining random delay if a beacon arrives before timer expires If random delay timer expires, node transmits beacon with a timestamp equal to its TSF timer value When receiving a beacon, node sets its TSF equal to the time stamp of the beacon if the value of the timestamp is more recent than the TSF value. Spring 2006 UCSC CMPE257

TSF in Ad Hoc Networks Problems, Issues? asynchronism There is no need to adopt the “latest” time in order to synchronize nodes for the purpose of channel access activities. Beacon avoidance in the presence of many nodes may lead to network being out of synchronization (TSF timer values at nodes differ by more than a threshold delta) Fastest-station asynchronism: Clock of fastest node is ahead of clocks of rest of nodes by more than delta Global asynchronism: A percentage of the n(n-1)/2 pairs of nodes are out of synchronization. asynchronism time H L H E(H)/[E(H)+E(L)] = (1-p)^tau p = prob. that node sends beacon successfully tau = delta/(diff. clock accuracy x beacon interval length) Spring 2006 UCSC CMPE257

Solutions? Do not require clocks to always move forward Have larger windows for beacon contention Give priority of beacon transmission to nodes with most recent clock value. …others? [HW] Spring 2006 UCSC CMPE257

Scheduled Access Problem description: Topology dependence: Given a set of contenders Mi of an entity i in contention context t, how does i determine whether itself is the winner during t ? Topology dependence: In unit disc graph model, two-hop neighbor information required to resolve contentions. In ad hoc networks, two-hop neighbors are acquired by each node broadcasting its one-hop neighbor set. Spring 2006 UCSC CMPE257

Goals Safety: In our case, this amounts to collision avoidance (avoid the hidden terminal problem) under a given set of assumptions. Liveness: No node is precluded from accessing the channel for an indefinite period of time Fairness: The channel is assigned to competing entities in proportion to their needs Timeliness: Through additional mechanisms, a node can be granted guarantees on how long it needs to wait between consecutive channel access times. Efficiency: Attain the highest possible maximum throughput Locality and distributed operation: No need for global network information. Simplicity: The algorithm(s) used for scheduling are simple enough to verify Spring 2006 UCSC CMPE257

Neighbor-Aware Contention Resolution (NCR) In each contention context (time slot t ): Compute priorities i is the winner for channel access if: 6 9 5 4 2 a b c d e Contention Floor Spring 2006 UCSC CMPE257

Channel Access Probability: Dependent on the number of contenders in the neighborhood. Channel access probability: Bandwidth allocation general formula to i Spring 2006 UCSC CMPE257

NAMA: Node Activation Multiple Access (Broadcast) Channel is time-slotted. Transmissions are broadcasts via omni-directional antenna: all one-hop neighbors can receive the packet from a node. The contenders of a node for channel access are neighbors within two hops because of direct and hidden terminal contentions. Spring 2006 UCSC CMPE257

Algorithm Spring 2006 UCSC CMPE257

Illustration of NAMA A B C D E F H G 8 6 5 3 4 1 2 9 Spring 2006 UCSC CMPE257

How Do We Share Context? Scheduling requires all nodes in two-hop neighborhood of each receiver to have consistent state. Neighbor protocol must disseminate such state How fast can it be done? How should it be integrated with time sync and the frame structure of the protocol? Each time slot has part of it dedicated to context? Part of the frame dedicated to context sharing? Spring 2006 UCSC CMPE257

Other Channel Access Protocols Other protocols using omni-directional antennas: LAMA: Link Activation Multiple Access PAMA: Pair-wise Activation Multiple Access Protocols that work when uni-directional links exist. Node A can receive node B’ s transmission but B cannot receive A’ s. Protocols using directional antenna systems. Spring 2006 UCSC CMPE257

Channel Access Probability Analysis of NAMA The channel access probability for a single node i is given by We are interested in average probability of channel access in multi-hop ad hoc networks. Spring 2006 UCSC CMPE257

Ad Hoc Network Settings Equal transmission range; Each node knows its one- and two-hop neighbors — Mi . Nodes are uniformly distributed on an infinite plane with density . A node may have different numbers of neighbors in one-hop and two-hop. Spring 2006 UCSC CMPE257

Counting One-Hop Neighbors The prob of having k nodes in an area of size S is a Poisson distribution: Average one-hop neighbors is: Note: the mean of r.v. with Poisson dist is Spring 2006 UCSC CMPE257

Counting Two-hop Neighbors Two nodes become two-hop neighbors if they share at least one one-hop neighbor. Average number in B(t): Spring 2006 UCSC CMPE257

Counting Two-hop Neighbors Probability of becoming two-hop: Prob of a node staying at tr is 2t. Summation of nodes in ring (r,2r) times the corresponding prob of becoming two-hop --- number of two-hop neighbors: Spring 2006 UCSC CMPE257

Total One- and Two-hop Neighbors Sum: This is average number of one-hop and two-hop neighbors. Spring 2006 UCSC CMPE257

Average Probability of Channel Access Apply Poisson distribution with the mean (number of one- and two-hop neighbors) Spring 2006 UCSC CMPE257

Plotting Channel Access Probability Spring 2006 UCSC CMPE257

Delay per Node Delay is related with the probability of channel access and the load at each node. Channel access probability can be different at each node. Delay is considered per node. Spring 2006 UCSC CMPE257

Packet Arrival and Serving: M/G/1 with server vacation: Poisson arrival (exponential arrival interval), service time distribution (any), single server. FIFO service strategy: head-of-line packet waits for geometric distributed period Yi with parameter 1-qi ̶ qi is the channel access probability of node i. Spring 2006 UCSC CMPE257

Service Time: Service time: Xi = Yi + 1. The mean and second moment of service time: Server vacation: V=1, Spring 2006 UCSC CMPE257

Delay in The System Pollaczek-Kinchin formula: Take in Xi and Vi : Delay in the system: (q>) Spring 2006 UCSC CMPE257

Plotting System Delay Spring 2006 UCSC CMPE257

System Throughput Multi-hop networks have concurrent transmissions >1. The system can carry as many packets at a time as all nodes can be activate. Simple! Spring 2006 UCSC CMPE257

Comparisons with CSMA CSMA/CA by Analysis Different slotting: NAMA long slots CSMA CSMA/CA short slots CSMA(CA) assumptions: Heavy load (always have packets waiting) Channel access regulated by back-off probability p’ in each slot. Convert the load to comparable one in NAMA. Spring 2006 UCSC CMPE257

Convert Load in CSMA(CA) to the Load in NAMA Each attempt to access channel is a packet arrival p’. Packet duration is geometric with average 1/q. Two state Markov chain to compute the load. Spring 2006 UCSC CMPE257

NAMA Load Relation: i=b is the load for each node. qm is the channel access probability of each node. Spring 2006 UCSC CMPE257

Protocol Throughput Comparison Spring 2006 UCSC CMPE257

Simulations Two scenarios: Compare with UxDMA: Fully connected: 2, 5, 10, 20 nodes. Multi-hop network: 100 nodes randomly placed in 1000x1000 area. Transmission range: 100, 200, 300, 400. Compare with UxDMA: Spring 2006 UCSC CMPE257

Fully Connected Network (Throughput) Spring 2006 UCSC CMPE257

Fully Connected Network (Delay) Spring 2006 UCSC CMPE257

Multi-hop Network (Throughput) Spring 2006 UCSC CMPE257

Multi-hop Network (Delay) Spring 2006 UCSC CMPE257

Conclusions NCR ensures collision-free transmissions. Only two-hop topology information is needed. HAMA performs better than static scheduling algorithms (UxDMA). HAMA performs better than contention-based protocols. The use of directional antennas can improve performance further. (Next topic) Spring 2006 UCSC CMPE257

Comments Scheduled-access protocols are evaluated in static environments and what about their performance in mobile networks? Neighbor protocol will also have impact on the performance of these protocols Need comprehensive comparison of contention-based and scheduled access protocols. Spring 2006 UCSC CMPE257

Beyond Elections and Graph Coloring… Elections and graph coloring are not enough to support all traffic or be efficient (example is a string of nodes). How do we support traffic with deadlines? Reservations How do we combine reservations & elections? CATA as a “pre-example” What if we can do more than transmit or receive one packet at a time? Network coding Channel division Multi-user detection Spring 2006 UCSC CMPE257

Synchronization and Power Management Finding and staying with a WLAN Synchronization functions TSF Timer, Beacon Generation Power Management Goal is sleeping without missing messages Power Management functions periodic sleep, frame buffering, Traffic Indication Map Spring 2006 UCSC CMPE257

Power Management Mobile devices are battery powered. Power management is important for mobility. Current LAN protocols assume stations are always ready to receive. Idle receive state dominates LAN adapter power consumption over time. How can we power off during idle periods, yet maintain an active session? 802.11 Power Management Protocol goals: Allow transceiver to be off as much as possible Transparent to existing protocols Flexible to support different applications possible to trade off throughput for battery life Spring 2006 UCSC CMPE257 24

Power Management in 802.11 Two power modes: active and power saving (PS). Protocols for Infrastructure, i.e., network with an access point (AP) Ad hoc networks. Spring 2006 UCSC CMPE257

Power Management in 802.11 Infrastructure network Node in active mode behaves “normally” and node in PS wakes up periodically to check for packets from AP. Node notifies AP when it changes modes. AP transmits beacons every beacon interval and nodes must monitor those frames. AP sends a traffic indication map (TIM) in each beacon, which states PS nodes with buffered frames in the AP. When PS node hears its ID, it remains active for the remaining of the beacon interval. Spring 2006 UCSC CMPE257

Power Management Approach Allow idle stations to go to sleep station’s power save mode stored in AP APs buffer packets for sleeping stations. AP announces which stations have frames buffered Traffic Indication Map (TIM) sent with every Beacon Power Saving stations wake up periodically listen for Beacons TSF assures AP and Power Save stations are synchronized stations will wake up to hear a Beacon TSF timer keeps running when stations are sleeping synchronization allows extreme low power operation Independent BSS also have Power Management similar in concept, distributed approach Spring 2006 UCSC CMPE257 25

Infrastructure Power Management TIM-Interval DTIM interval Time-axis Busy Medium TIM TIM DTIM TIM TIM DTIM Broadcast Broadcast AP activity Broadcast frames are also buffered in AP. all broadcasts/multicasts are buffered broadcasts/multicasts are only sent after DTIM DTIM interval is a multiple of TIM interval Spring 2006 UCSC CMPE257 26

Infrastructure Power Management TIM-Interval DTIM interval Time-axis Busy Medium TIM TIM DTIM TIM TIM DTIM Broadcast AP activity Broadcast PS Station Broadcast frames are also buffered in AP. all broadcasts/multicasts are buffered broadcasts/multicasts are only sent after DTIM DTIM interval is a multiple of TIM interval Stations wake up prior to an expected (D)TIM. Spring 2006 UCSC CMPE257 27

Infrastructure Power Management TIM-Interval DTIM interval Time-axis Busy Medium TIM TIM DTIM TIM TIM DTIM Broadcast AP activity Broadcast PS Station PS-Poll Tx operation Broadcast frames are also buffered in AP. all broadcasts/multicasts are buffered broadcasts/multicasts are only sent after DTIM DTIM interval is a multiple of TIM interval Stations wake up prior to an expected (D)TIM. If TIM indicates frame buffered station sends PS-Poll and stays awake to receive data else station sleeps again Spring 2006 UCSC CMPE257 28

References [R01] S. Ramanathan, A unified framework and algorithm for channel assignment in wireless networks, ACM Wireless Networks, Vol. 5, No. 2, March 1999. [BG01] Lichun Bao and JJ, A New Approach to Channel Access Scheduling for Ad Hoc Networks, Proc. of The Seventh ACM Annual International Conference on Mobile Computing and networking (MOBICOM), July 16-21, 2001, Rome, Italy. [BG02] Lichun Bao and JJ, Hybrid Channel Access Scheduling in Ad Hoc Networks, IEEE Tenth International Conference on Network Protocols (ICNP), Paris, France, November 12-15, 2002. Spring 2006 UCSC CMPE257

References [IEEE99] IEEE Standard for Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications, IEEE Std 802.11-1999. [TK84] H. Takagi and L. Kleinrock, Optimal Transmission Range for Randomly Distributed Packet Radio Terminals, IEEE Trans. on Comm., vol. 32, no. 3, pp. 246-57, 1984. [WV99] L. Wu and P. Varshney, Performance Analysis of CSMA and BTMA Protocols in Multihop Networks (I). Single Channel Case, Information Sciences, Elsevier Sciences Inc., vol. 120, pp. 159-77, 1999. [WG02] Yu Wang and JJ, Performance of Collision Avoidance Protocols in Single-Channel Ad Hoc Networks, IEEE Intl. Conf. on Network Protocols (ICNP ’02), Paris, France, Nov. 2002. Spring 2006 UCSC CMPE257

Acknowledgments Parts of the presentation are adapted from the following sources: Mustafa Ergen, UC Berkeley, http://www.eecs.berkeley.edu/~ergen/docs/IEEE-802.11overview.ppt Greg Ennis, Symbol Technologies, http://grouper.ieee.org/groups/802/11/Tutorial/archit.pdf Phil Belanger, Aironet and Wim Diepstraten, Lucent Technologies, http://grouper.ieee.org/groups/802/11/Tutorial/MAC.pdf Abhishek Karnik, Dr. Ratan Guha, University Of Central Florida, http://www.cs.ucf.edu/courses/cda4527/WLAN/80211.ppt Spring 2006 UCSC CMPE257

CMPE 257: Wireless Networking SET MAC-2: Medium Access Control Protocols Spring 2006 UCSC CMPE257

NAMA Improvements Inefficient activation in certain scenarios. For example, only one node, a, can be activated according NAMA, although several other opportunities exist. —— We want to activate g and d as well. a f g c d e h b 10 1 6 4 7 3 8 5 Spring 2006 UCSC CMPE257

Node + Link (Hybrid) Activation Additional assumption Radio transceiver is capable of code division channelization (DSSS —— direct sequence spread spectrum) Code set is C . Code assignment for each node is per time slot: i .code = i .prio mod |C | Spring 2006 UCSC CMPE257

Hybrid Activation Multiple Access (HAMA) Node state classification per time slot according to their priorities. Receiver (Rx): intermediate prio among one-hop neighbors. Drain (DRx): lowest prio amongst one-hop. BTx: highest prio among two-hop. UTx: highest prio among one-hop. DTx: highest prio among the one-hop of a drain. Spring 2006 UCSC CMPE257

HAMA (cont.) Transmission schedules: BTx —> all one-hop neighbors. UTx —> selected one-hops, which are in Rx state, and the UTx has the highest prio among the one-hop neighbors of the receiver. DTx —> Drains (DRx), and the DTx has the highest prio among the one-hops of the DRx. Spring 2006 UCSC CMPE257

HAMA Operations Suppose no conflict in code assignment. Nodal states are denoted beside each node: Node D converted from Rx to DTx. Benefit: one-activation in NAMA to four possible activations in HAMA. a f g c d e h b 10-BTx 1-DRx 6-Rx 4-DRx 7-UTx 3-DRx 8-Rx 5-DTx Spring 2006 UCSC CMPE257

Other Channel Access Protocols Other protocols using omni-directional antennas: LAMA: Link Activation Multiple Access PAMA: Pair-wise Activation Multiple Access Protocols that work when uni-directional links exist. Node A can receive node B’ s transmission but B cannot receive A’ s. Protocols using direct antenna systems. Spring 2006 UCSC CMPE257

Comparison of Channel Access Probability focus on NAMA & HAMA comparison. Spring 2006 UCSC CMPE257

Protocol Throughput Comparison Spring 2006 UCSC CMPE257

Comments Scheduled-access protocols are evaluated in static environments and what about their performance in mobile networks? Neighbor protocol will also have impact on the performance of these protocols Need comprehensive comparison of contention-based and scheduled access protocols. Spring 2006 UCSC CMPE257

References [R01] S. Ramanathan, A unified framework and algorithm for channel assignment in wireless networks, ACM Wireless Networks, Vol. 5, No. 2, March 1999. [BG01] Lichun Bao and JJ, A New Approach to Channel Access Scheduling for Ad Hoc Networks, Proc. of The Seventh ACM Annual International Conference on Mobile Computing and networking (MOBICOM), July 16-21, 2001, Rome, Italy. [BG02] Lichun Bao and JJ, Hybrid Channel Access Scheduling in Ad Hoc Networks, IEEE Tenth International Conference on Network Protocols (ICNP), Paris, France, November 12-15, 2002. Spring 2006 UCSC CMPE257

References [IEEE99] IEEE Standard for Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications, IEEE Std 802.11-1999. [TK84] H. Takagi and L. Kleinrock, Optimal Transmission Range for Randomly Distributed Packet Radio Terminals, IEEE Trans. on Comm., vol. 32, no. 3, pp. 246-57, 1984. [WV99] L. Wu and P. Varshney, Performance Analysis of CSMA and BTMA Protocols in Multihop Networks (I). Single Channel Case, Information Sciences, Elsevier Sciences Inc., vol. 120, pp. 159-77, 1999. [WG02] Yu Wang and JJ, Performance of Collision Avoidance Protocols in Single-Channel Ad Hoc Networks, IEEE Intl. Conf. on Network Protocols (ICNP ’02), Paris, France, Nov. 2002. Spring 2006 UCSC CMPE257

MAC Protocols Using Directional Antennas Basic protocols Directional Virtual Carrier Sensing (DVCS) Directional MAC (D-MAC) in UDAAN Spring 2006 UCSC CMPE257

MAC Protocols Using Directional Antennas The MAC protocols so far assume that a node’s transmissions reach all of its neighbors. With powerful antenna systems, it is possible to limit transmissions and receptions to desired directions only. This can increase spatial reuse and reduce interferences to neighbors nodes. Caveat: Not all neighbor nodes defer access. Directional receiving is not always desired. Nodes may move in and out of range. Nodes don’t always have the complete info. of the channel. Nodes may miss some broadcast information. Spring 2006 UCSC CMPE257

Omni-Directional and Directional Transmissions  Node B Node B Node C Node C Node A Node A Omni-directional transmission Directional transmission Spring 2006 UCSC CMPE257

Directional Antenna Models Antenna systems Switched beam – fixed orientation Adaptive beam forming – any direction Simulation models: Complete signal attenuation outside the directional transmission beamwidth () ``Cone plus ball’’ model Directional transmissions have higher gains Possible to use power control to reduce the gain Various medium access control schemes have been proposed and/or investigated (see Refs). We’ll extend the analysis in our future work. Spring 2006 UCSC CMPE257

Basic Scheme One OTOR (omni-transmit, omni-receive) The usual omni RTS/CTS based collision avoidance All packets are transmitted and received omni-directionally. IEEE 802.11 MAC protocol uses such scheme. Spring 2006 UCSC CMPE257

Basic Scheme Two DTOR (directional-transmit, omni-receive) Packets are transmitted directionally. Packets are received omni-directionally. Increased spatial reuse (+) and collisions (-). Talks btw. A & B, C & D can go on concurrently; – More collisions may occur; + Spatial reuse is increased; + Nodes spend less time waiting. You may wonder why we use omni-receive: because it is possible that somehow after we send CTS which cannot be received correctly by the receiver, we’ll wait for Data in that direction for a long time. Omni-receive helps the node to know other activities in the channel. Spring 2006 UCSC CMPE257

Basic Scheme Three DTDR (directional-transmit, directional-receive) All packets are transmitted and received directionally. Aggressive spatial reuse Talks btw. A & B, C & D can go on concurrently; – More collisions may occur; – Channel status info. is incomplete; + Aggressive spatial reuse; + Nodes spend less time waiting. Spring 2006 UCSC CMPE257

Basic Scheme Four MTDR (mixed transmit, directional receive) CTS packets are transmitted omni-directionally while other packets are transmitted directionally. Tradeoff between spatial reuse and collision avoidance D sends RTS to C directionally; C replies with omni-CTS; + A and G defer their access and won’t cause collisions; – However, A cannot talk with B at the same time. Omni-CTS is used to notify other nodes to defer their access. Spring 2006 UCSC CMPE257

Predictions from the Analysis [WG03] The DTDR scheme performs the best among the schemes analyzed. Increased spatial reuse and reduced interference through directional transmissions. Directional receiving cancels much interferences from neighbors and hidden terminals. Throughput of the DTDR scheme with narrow beamwidth θ has a slightly increase when N increases. Spatial reuse effect is more conspicuous. Scalability problem is mitigated. Spring 2006 UCSC CMPE257

Simulation Results [WG03] Higher-gain directional transmissions have negative effects on throughput and delay. More nodes are affected. Influence of side lobes can be almost canceled out if: The level of side lobes is reasonably low through the advancement of antenna systems. Carrier sensing threshold is raised such that nodes are less sensitive to channel activities. Spring 2006 UCSC CMPE257

Advanced Schemes Directional Virtual Carrier Sensing ([TMRB02]) Angle-of-Arrival (AoA) information available Nodes record direction information and defer only to non-free directions (directional NAV) UDAAN ([RRSWP05]) Switched beam antenna Experimental system was built to test the effectiveness of directional antenna systems Omni-CTS is used to notify other nodes to defer their access. Spring 2006 UCSC CMPE257

Details on Directional NAV Physical carrier sensing still omni-directional Virtual carrier sensing be directional – directional NAV When RTS/CTS received from a particular direction, record the direction of arrival and duration of proposed transfer Channel assumed to be busy in the direction from which RTS/CTS received Spring 2006 UCSC CMPE257

Directional NAV (DNAV) Nodes overhearing RTS or CTS set up directional NAV (DNAV) for that Direction of Arrival (DoA) D CTS C X Y Spring 2006 UCSC CMPE257

Directional NAV (DNAV) Nodes overhearing RTS or CTS set up directional NAV (DNAV) for that Direction of Arrival (DoA) D C DNAV X Y Spring 2006 UCSC CMPE257

Directional NAV (DNAV) New transmission initiated only if direction of transmission does not overlap with DNAV, i.e., if (θ > 0) B D DNAV θ A C RTS Spring 2006 UCSC CMPE257

D-MAC Forced Idle is to avoid starvation FI-Busy ``aggressive’’ Tight integration with power control Spring 2006 UCSC CMPE257

Directional Neighbor Discovery Three kinds of links (neighbors) N-BF, without beam forming T-BF, using only transmit-only beamforming TR-BF, using transmit and receive beamforming Two methods for discovery Informed discovery Blind discovery Spring 2006 UCSC CMPE257

Directional Packet Transmission B D-O transmission B’s omni receive range D-D transmission A B B’s directional receive beam Spring 2006 UCSC CMPE257

Related topics Neighbor protocol and topology management Energy efficiency Routing Spring 2006 UCSC CMPE257

References [KSV00] Ko et al., Medium Access Control Protocols Using Directional Antennas in Ad Hoc Networks, in IEEE INFOCOM 2000. [NYH00] Nasipuri et al., A MAC Protocol for Mobile Ad Hoc Networks Using Directional Antennas, in IEEE WCNC 2000. [R01] R. Ramanathan, On the Performance of Ad Hoc Networks with Beamforming Antennas, ACM MobiHoc '01, Oct. 2001. [TMRB02] Takai et al., Directional Virtual Carrier Sensing for Directional Antennas in Mobile Ad Hoc Networks, ACM MobiHoc ’02, June 2002. [CYRV02] Choudhury et al., Medium Access Control in Ad Hoc Networks Using Directional Antennas, ACM MobiCom '02, Sept. 2002. [WG03] Yu Wang and JJ, Collision Avoidance in Single-Channel Ad Hoc Networks Using Directional Antennas, in IEEE ICDCS '03. [RRSWP05] Ramanathan et al., Ad Hoc Networking With Directional Antennas: A Complete System Solution, IEEE JSAC 2005. Spring 2006 UCSC CMPE257

Acknowledgments Parts of the presentation are adapted from the following sources: Prasant Mohapatra, UC Davis, http://www.cs.ucdavis.edu/~prasant/ECS257/NOTES/Adhoc-Sensor.ppt Spring 2006 UCSC CMPE257

MAC Protocols for Networks with Multiple Channels Motivation and issues Summary of approaches Examples Open research problems Spring 2006 UCSC CMPE257

Motivation for Using Multiple Channels Multiple orthogonal channels available in IEEE 802.11 3 channels in 802.11b 12 channels in 802.11a Multiple channels available in general Utilizing multiple channels can improve throughput and reduce delays Allow simultaneous transmissions in the same 2-hop neighborhood. 1 defer 1 2 Single channel Multiple Channels Spring 2006 UCSC CMPE257

Issues with Multiple Channels Senders and receivers need to “meet” in one of many channels Hidden and exposed terminals are now problems involving more than one channel Similar problems than with space division Each node may have one or multiple radios Half duplex nodes, even with multiple radios per node: At best, a node can receive or transmit over one or multiple radios, but not both Synchronization is hard to avoid w/o a dedicated control channel. Spring 2006 UCSC CMPE257

Approaches* Dedicated Control Channel Split Phase Common Hopping Dedicated control radio & channel for all control messages DCA[Wu2000], DCA-PC[Tseng2001], DPC[Hung2002]. Split Phase Fixed periods divided into (i) channel negotiation phase on default channel & (ii) data transfer phase on negotiated channels MMAC[J.So2003], MAP [Chen et al.] Common Hopping All non-busy nodes follow a common, well-known channel hopping sequence -- the control channel changes. HRMA[Tang & JJ 98], CHMA, CHAT, RICH [Tzamaloukas & JJ] Parallel Rendezvous Each node publishes its own channel hopping schedule SSCH [Bahl04], McMAC [So et al] * Mo, So and Walrand, “Comparison of Multi-Channel MAC Protocols,” MSWIM 05. Spring 2006 UCSC CMPE257

Approach 1: Dedicated Control Channel Dedicated control radio and control channel DCA[Wu2000], DCA-PC[Tseng2001], DPC[Hung2002]. [Control Chan]: S ---- RTS (Suggested Data Chan.) --> R [Control Chan]: S <-- CTS (Agreed Data Chan.) ---- R [Control Chan]: (optional) S --- Reservation (broadcast) ---> All [Data Chan]: Sender --- Data Packet ---> Receiver Spring 2006 UCSC CMPE257

Dedicated Control Channel Rendezvous & contention occur on the control channel. Channel Ch3 Data Ack Node 1+2 Ch2 Data . . . Ack Ch1 Rts (2,3) Cts (2) Rsv (2) Rts (3) Cts (3) Rsv (3) Time Legend: Node 1 Node 2 Note 3 Node 4 Spring 2006 UCSC CMPE257

Nasipuri’s Protocol Assumes N transceivers per host Capable of listening to all channels simultaneously Sender searches for an idle channel and transmits on the channel [Nasipuri99WCNC] Extensions: channel selection based on channel condition on the receiver side [Nasipuri00VTC] Disadvantage: High hardware cost (today!)* * In the future (~5 to 10 years), having 4 radios per node will be affordable Spring 2006 UCSC CMPE257

Approach 2: Split-Phase Time is divided into (equal) periods. Each period consists of 2 phases: control (channel negotiation), data transfer. Examples ae: MMAC[J.So2003] (UIUC), MAP [Chen2003] Channel negotiation happens on a default channel. Nodes negotiate the channels to use. RTS/CTS/Data on the separate channels Spring 2006 UCSC CMPE257

Split-Phase Channel Channel negotiation on a common channel Ch2 Rts Cts Data Ack Ch1 Ch0 Hello (1,2,3) Ack (1) Rsv (1) Rts Cts Data Ack Hello (2,3) ... Time Data Transfer Phase Control Phase Legend: Node 1 Node 2 Note 3 Node 4 Spring 2006 UCSC CMPE257

MMAC (So and Vaidya) Assumptions Each node is equipped with a single transceiver The transceiver is capable of switching channels Channel switching delay is approximately 250us Per-packet switching not recommended Occasional channel switching not to expensive Multi-hop synchronization is achieved by other means Spring 2006 UCSC CMPE257

MMAC Idea similar to IEEE 802.11 PSM Divide time into beacon intervals At the beginning of each beacon interval, all nodes must listen to a predefined common channel for a fixed duration of time (ATIM window) Nodes negotiate channels using ATIM messages Nodes switch to selected channels after ATIM window for the rest of the beacon interval Spring 2006 UCSC CMPE257

Preferred Channel List (PCL) Each node maintains PCL Records usage of channels inside the transmission range High preference (HIGH) Already selected for the current beacon interval Medium preference (MID) No other vicinity node has selected this channel Low preference (LOW) This channel has been chosen by vicinity nodes Count number of nodes that selected this channel to break ties Spring 2006 UCSC CMPE257

Channel Negotiation In ATIM window, sender transmits ATIM to the receiver Sender includes its PCL in the ATIM packet Receiver selects a channel based on sender’s PCL and its own PCL Order of preference: HIGH > MID > LOW Tie breaker: Receiver’s PCL has higher priority For “LOW” channels: channels with smaller count have higher priority Receiver sends ATIM-ACK to sender including the selected channel Sender sends ATIM-RES to notify its neighbors of the selected channel Spring 2006 UCSC CMPE257

Channel Negotiation A B C D Time Common Channel Selected Channel Beacon B C D Time ATIM Window Spring 2006 UCSC CMPE257 Beacon Interval

Channel Negotiation A B C D Time Common Channel Selected Channel ATIM- RES(1) ATIM A Beacon B ATIM- ACK(1) C D Time ATIM Window Spring 2006 UCSC CMPE257 Beacon Interval

Channel Negotiation A B C D Time Common Channel Selected Channel ATIM- RES(1) ATIM A Beacon B ATIM- ACK(1) ATIM- ACK(2) C D ATIM ATIM- RES(2) Time ATIM Window Spring 2006 UCSC CMPE257 Beacon Interval

Channel Negotiation A B C D Time Common Channel Selected Channel ATIM- RES(1) ATIM RTS DATA Channel 1 A Beacon B Channel 1 ATIM- ACK(1) CTS ACK ATIM- ACK(2) CTS ACK Channel 2 C Channel 2 D ATIM ATIM- RES(2) RTS DATA Time ATIM Window Spring 2006 UCSC CMPE257 Beacon Interval

Simulation Model ns-2 simulator Transmission rate: 2Mbps Transmission range: 250m Traffic type: Constant Bit Rate (CBR) Beacon interval: 100ms Packet size: 512 bytes ATIM window size: 20ms Default number of channels: 3 channels Compared protocols 802.11: IEEE 802.11 single channel protocol DCA: Wu’s protocol MMAC: Proposed protocol Spring 2006 UCSC CMPE257

Wireless LAN - Throughput 2500 2000 1500 1000 500 2500 2000 1500 1000 500 MMAC MMAC DCA DCA Aggregate Throughput (Kbps) 802.11 802.11 1 10 100 1000 1 10 100 1000 Packet arrival rate per flow (packets/sec) Packet arrival rate per flow (packets/sec) 30 nodes 64 nodes MMAC shows higher throughput than DCA and 802.11 Spring 2006 UCSC CMPE257

Multi-hop Network – Throughput 2000 1500 1000 500 1500 1000 500 MMAC MMAC DCA DCA Aggregate Throughput (Kbps) 802.11 802.11 1 10 100 1000 1 10 100 1000 Packet arrival rate per flow (packets/sec) Packet arrival rate per flow (packets/sec) 3 channels 4 channels Spring 2006 UCSC CMPE257

Throughput of DCA and MMAC (Wireless LAN) 4000 3000 2000 1000 4000 3000 2000 1000 6 channels 6 channels 2 channels Aggregate Throughput (Kbps) 2 channels 802.11 802.11 Packet arrival rate per flow (packets/sec) Packet arrival rate per flow (packets/sec) DCA MMAC MMAC shows higher throughput compared to DCA Spring 2006 UCSC CMPE257

Analysis of Results DCA MMAC Bandwidth of control channel significantly affects performance Narrow control channel: High collision and congestion of control packets Wide control channel: Waste of bandwidth It is difficult to adapt control channel bandwidth dynamically MMAC ATIM window size significantly affects performance ATIM/ATIM-ACK/ATIM-RES exchanged once per flow per beacon interval – reduced overhead Compared to packet-by-packet control packet exchange in DCA ATIM window size can be adapted to traffic load Spring 2006 UCSC CMPE257

Further Work Needed Dynamic adaptation of ATIM window size based on traffic load for MMAC Efficient multi-hop clock synchronization Better uses of data segment Multipoint communication support Spring 2006 UCSC CMPE257

Approach 3: Common Hopping All idle nodes follow the same channel hopping sequence E.g., HRMA[Tang98], CHMA[Tzamaloukas2000], CHAT[Tzamaloukas2000] [Common Channel]: S ---- RTS ---> R [Next Common Channel]: everyone else [Same Channel]: S <-- CTS ---- R [Same Channel]: S ---- Data ---> R All return to “Current Common Channel” after sending/receiving Spring 2006 UCSC CMPE257

3. Common Hopping Idle nodes hop together in “common channel” Channel RTS (b to a) Ch1 Cts, Data, Ack Ch0 RTS (c to d) 1 2 3 4 5 6 7 8 9 10 11 Time Enough for one RTS Legend: Node a Node b Note c Node d Spring 2006 UCSC CMPE257

Main Limitations The time any dialogue can last must be shorter than the time it takes for the common hopping sequence to revisit the channel being used. Approach is useful only if sufficiently large numbers of channels are available. Time sync is needed. Spring 2006 UCSC CMPE257

Approach 4: Parallel Rendezvous Nodes choose their own hopping sequences. Nodes publish the seeds of their hopping sequences so nodes can track each other. Key: Parallel rendezvous on multiple channels Examples: SSCH [Bahl et al.]: senders transmits only when receivers are on the same channel. McMAC [So & Walrand]: senders can deviate from their published schedule temporarily to transmit. Spring 2006 UCSC CMPE257

McMAC Every node has a its own random hopping sequence called the“home channel”. Hopping freq. is a parameter. To Tx, sender S leaves its home channel to meet its receiver R with some probability p_tx. If R’s channel is busy or R is away from home, S tries again later. Otherwise, S and R exchanges Data/Ack. Spring 2006 UCSC CMPE257

McMAC (no hopping) t=1 2 3 4 5 6 ... Ch 1 Ch 2 Ch 3 Ch 4 ? ? Sender needs to know the home channel of the receiver, but time sync. is not needed. Spring 2006 UCSC CMPE257

McMAC (with hopping) t=1 2 3 4 5 6 7 8 9 Ch1 Ch2 Original schedule Spring 2006 UCSC CMPE257

McMAC (with hopping) t=1 2 3 4 5 6 7 8 9 Ch1 Ch2 Original schedule 2. RTS/ CTS/ Data 4. Hopping resumes 1. Data arrives 3. Hopping stopped during data transfer Spring 2006 UCSC CMPE257

Qualitative Comparison (True? Let’s Discuss!) Control Channel Split- Phase Common Hopping Parallel Rendez-vous # Radios 2 1 Contention Bottleneck Y N Time Sync. Loose Very Tight Param. Track neighbors Spring 2006 UCSC CMPE257

Simulation Parameters N: # nodes M: # channels L: avg. packet length T_sw: channel switch time T_slot: slot time = RTS/CTS S: link speed/channel (Mbps) Spring 2006 UCSC CMPE257

Simulation Scenarios 20 40 3 12 1024B /10240B (5/50 slots) Similar to 802.11b Scenario #2 Similar to 802.11a N: # nodes 20 40 M: # chans 3 12 L: Avg Pkt Len 1024B /10240B (5/50 slots) (6.8/68 slots) T_sw: Switch Time 100us T_slot: Slot Time 810us 200us S: Link Speed 2Mbps 6Mbps Spring 2006 UCSC CMPE257

Spring 2006 UCSC CMPE257

Dedicated Control Channel (short pkts) Spring 2006 UCSC CMPE257 Time (slot)

Dedicated Control Channel (long pkts) Spring 2006 UCSC CMPE257 Time (slot)

Split Phase (30 control / 120 data slots) Channel Spring 2006 UCSC CMPE257 Time (slot)

Split Phase (12 control / 48 data slots) Channel Spring 2006 UCSC CMPE257 Time (slot)

Common Hopping (Short Pkts) Channel Spring 2006 UCSC CMPE257 Time (slot)

Common Hopping (Long Pkts) Channel Spring 2006 UCSC CMPE257 Time (slot)

McMAC hopping (Short Pkts) Channel Spring 2006 UCSC CMPE257 Time (slot)

McMAC hopping (Long Pkts) Channel Spring 2006 UCSC CMPE257 Time (slot)

Spring 2006 UCSC CMPE257

Dedicated Control Channel (short pkts) Spring 2006 UCSC CMPE257 Time (slot)

Common Hopping (short pkts) Channel Spring 2006 UCSC CMPE257 Time (slot)

McMAC hopping (short pkts) Channel Spring 2006 UCSC CMPE257 Time (slot)

Conclusions by Mo, So, and Walrand Dedicated control channel: works surprisingly well esp. for long packets! Split-phase: depends heavily on the control/data phase durations Common-hopping: “fragmentation” problem Parallel rendezvous: good potential, but synchronization is an issue. Spring 2006 UCSC CMPE257

Research Opportunities Few schemes based on contention have not addresses collision issues. Efficient time sync is a requirement for multi-channel MACs, unless a dedicated control channel is used. Can we map prior approaches for distributed code assignment for CDMA networks to multi-channel MACs by making a code to equal a hoping sequence? Is topology-dependent scheduling inherently better than multiple rendezvous? Impact of updating 2-hop neighborhood What happens when radios become really cheap and a node can receive multiple concurrent transmissions? (Same if we have MUD with MIMO) What scheduling advantages do we gain? What MACs can we propose? Spring 2006 UCSC CMPE257