Chapter 1 Basic concepts of wireless data networks (cont d) Part 2: Medium access methods for mobile data networks Sept 15 2004 1 Fixed assignment access schemes in voice-oriented networks Frequency division multiple access (FDMA) Different users using different frequency channels Time division multiple access (TDMA) Different users using different time slots Code division multiple access (CDMA) Different users using different codes Time division duplex (TDD) The downlink and uplink of a connection use different time slots Frequency division duplex (FDD) Uplink and downlink use different frequency channels 2
Medium access schemes in voiceoriented networks A FDMA/FDD t uplink downlink A t FDMA/TDD User 1 User 2 User 3 User 4 A TDMA/FDD f1 f2 f3 f4 t uplink f1 f2 f3 f4 downlink t1 t2 f f A f1 f2 f3 f4 t 2 1 f1 f2 f3 f4 downlink uplink f TDMA/TDD downlink uplink f 3 Random access schemes in data networks Data traffic is bursty Fixed assignment access scheme wastes resources when traffic is bursty Random access schemes more flexible and efficient for data services Contention among users for access 4
Pure ALOHA Operation: When a packet arrives, it is encoded for error detection, then transmitted. At the receiver side, check the correctness when a packet received. If error-free, an ACK sent back. The sender will wait an amount of time for ACK. If ACK received within this time interval, transmission is successful. If ACK is not received in time, delay a random amount of time, then retransmit it. 5 An example of pure ALOHA T Successful transmission User 1 User 2 User 3 User 4 2T Random delay No other packet arrival New packet Rescheduled packet T: time for sending a packet : collision 6
Throughput of pure ALOHA A packet is successfully transmitted if no other packets arrive within one frame time of its start. Poisson arrival with mean λ pkt/s, including new and rescheduled packets P(N pkt arrive in time t) = (λt) N e - λ t / N!, N = 0, 1, Time for sending one packet T is constant. G= pkt sent per pkt time T= λ T Throughput (pkt successfully sent per pkt time T) 2G S = S max =0.184 when G=0.5 Ge Advantage: simple; disadvantage: low throughput 7 Slotted ALOHA Time is divided into discrete time intervals, each corresponding to one packet transmission Synchronization needed among different geographically dispersed users Operation When a packet arrives, it will be sent at the beginning of next time slot If ACK received in time, successful. Otherwise, delay this packet for a random interval and retransmitted. Slotted ALOHA is adopted in GPRS 8
Operation of slotted ALOHA User 1 User 2 User 3 User 4 T No other packet arrival T: one time slot Successful transmission Random delay New packet Rescheduled packet : collision 9 Throughput of slotted ALOHA A packet is successfully transmitted if no other packets arrive in the same time slot Poisson arrival with mean λ Time for sending one packet T. G= λt Throughput S = Ge S max =0.368 when G=1 G Twice the throughput of pure ALOHA (or is it really twice?) 10
Dynamic slotted ALOHA Full duplex communication in Mobitex: BS and mobile can transmit simultaneously Mobile can only transmit during certain free cycles consisting of several random slots that are periodically initiated by BS Long messages are transmitted in reserved time slots. 11 Dynamic slotted ALOHA (cont d) Operation: BS sends a free message to indicate a number of random slots for MS to send requests. MS having requests to send chooses one of these random time slots to send request. If the request is received by BS without collision, an Access Grant packet is sent by the BS to the MS indicating the reserved time slots for sending information message. MS sends the information message in the reserved time slot. 12
An example of dynamic slotted ALOHA Free Free AG to MS2 ACK to MS2 BS Message to MS3 1 2 3 4 5 MS Message from MS2 MS1 MS2 ACK from MS3 One message to MS3 MS1 and MS2 want to send requests to BS Access grant is given to MS2 MS2 then sends packet to BS in reserved time slot 13 Operation of this example BS sends a Free message indicating 5 random slots available BS sends a message to MS3 during these random slots (in downlink channel) MS1 and MS2 want to send requests to BS during these random slots. They randomly choose one of these 5 random slots to send requests. MS1 chooses slot 1 and MS2 chooses slot 3. These requests are received without collision by BS. Access Grant is given to MS2. MS2 sends message in reserved slot. BS sends ACK to MS2 indicating the success of the sent message. 14
Carrier sense multiple access (CSMA) Reason: the efficiency of ALOHA is too low as each station operates independently. By sensing channel status, collision can be significantly reduced. Operation: Before sending packet, the MS senses if the channel is idle. If idle, send packet. Otherwise, delay using a back-off algorithm, then sense the channel again. If packet collides, reschedule as in ALOHA. (Why would packets collide with sensing?) 15 An example of CSMA User 1 User 2 User 3 Reschedule delay Busy delay Packet arrival New packet Delayed packet Rescheduled packet : collision 16
Variations of CSMA Non-persistent CSMA: after sensing channel is busy, delay a certain amount of time before sensing the channel again. 1-persistent CSMA: after sensing the channel is busy, continues sensing the channel until the channel is idle and then sends the packet p-persistent CSMA (for slotted channel): after sensing the channel is busy, continues sensing the channel until it is idle. Then the packet is sent with probability p. Channel busy delay Non-persistent ready 1-persistent p-persistent 17 Throughput of CSMA Unslotted, 1-persistent: Slotted, 1-persistent: [ 1+ G + ag(1 + G + ag / 2) ] G G(1 + 2a) (1 e e ) + (1 + ag) e G(1+ 2a) S = ag G(1 + a) ag [ 1+ a e ] G (1 + a)(1 e e ) + ae G(1+ a) S = ag G(1 + a) Unslotted, non-persistent: ag Ge S = G(1 + 2a) + e ag Slotted, non-persistent: G: the same as in ALOHA a= τ/t P τ: maximum propagation delay T P : packet transmission time S age 1 e = ag ag + a 18
Comparison Throughput versus offered traffic load for various random access protocols 19 Comparison (cont d) Capacity versus normalized delay for various random access protocols 20
Comparison (cont d) Delay-throughput behavior of random access protocols 21 CSMA/CA (collision avoidance) Adopted by IEEE 802.11 WLAN Operation: Step 1: When packet arrives, sense the channel. If it is idle, send packet; otherwise, choose a backoff time delay randomly. Then continues sensing the channel. Step 2: When channel becomes idle, delay an IFS (interframe spacing). Then count down the backoff delay. Step 3: If channel remains idle until the delay is 0, send packet. If channel becomes busy during delay countdown, freeze the delay and continues sensing the channel. Go to Step 2. 22
An example of CSMA/CA IFS IFS IFS IFS A Frame B C Frame Frame D E Frame Frame CW CW CW CW Delay after packet ready Back off Remaining back-off 23 Complications caused by wireless channel Hidden terminal problem: two nodes are too far away that they cannot sense the transmission of each other and cause collision. Throughput is decreased 24
Complications (cont d) Capture effect: when two nodes are trying to send packets simultaneously, and the distances to the receiver are different, the packet sent by the closer node may be correctly received, and performance may be improved. 25 Impact of capture effect Impact of capture effect on throughput, BPSK modulation and SNR=20dB 26
Impact of capture effect (cont d) Delay versus throughput of CSMA for BPSK modulation and SNR=20dB 27 Impact of hidden terminals Mobile terminal Access point Area 2 Area 1 Area 1 carrier sensible by red mobile terminal Area 2 carrier non-sensible by red mobile terminal, but still within coverage area of AP => hidden terminal effect Coverage areas of an AP and a mobile terminal in a WLAN 28
Impact of hidden terminals (cont d) With capture R MS < R BS R MS = R BS Without capture R MS : radius of the coverage of MS R BS : radius of the coverage of BS 29 Providing data services in voiceoriented networks Two reasons for providing data services in voice-oriented networks (to improve efficiency and save cost): Some portion of unused resources can be used by data users Short pauses in voice channel can be used for data services 30
Providing data services in FDMA system CDPD (cellular digital packet data) packet data system uses available frequency channel in existing analog FDMA cellular phone network (AMPS) Data rate<=19.2kbps Idle frequency channels are used for data services Talk pauses not exploited Frequency hopping allowed to release current channel to voice users Only one frequency channel for one data user: data rate limited 31 Average number of idle channels in FDMA system Required call blocking lower, more available channels for data services For 1% call blocking and 15 channels, 7 available 32
Average channel AMPS idle period Higher call blocking requirement, longer idle period 33 Providing data services in TDMA system GPRS in GSM Unused time slots are assigned for data services Multiple time slot assignment possible to increase data rate Talk pause not exploited 34
Movable boundary TDMA scheme with silence detection Some slots dedicated for data services only Remaining slots shared by data and voice users, with voice users having higher priority Movable boundary. When the number of voice users increases, the number of dedicated slots for data decreases Movable boundary Voice slots data slots Voice slots data slots frame frame 35 Providing data services in CDMA system Data services in 3G Integration of data and voice services Assign multiple parallel channels or reduce processing gain, data rate increased 36