300 likes | 317 Views
Note: in last class I said that the noise floor in 802.11b was -110 dBm) That was wrong, it was 100 dBm. Thermal = 173 dBm/Hz Bandwidth = 10log10(20e6) = 10log10(2) + 10log10(1e7) = 73 dB Giving 100 dBm. Existing MACs for Wireless nets. Basic 802.11 media access control - 802.11 DCF.
E N D
Note: in last class I said that the noise floor in 802.11b was -110 dBm) That was wrong, it was 100 dBm. Thermal = 173 dBm/Hz Bandwidth = 10log10(20e6) = 10log10(2) + 10log10(1e7) = 73 dB Giving 100 dBm Existing MACs for Wireless nets
Basic 802.11 media access control - 802.11 DCF • DCF – distribution coordination function • Basic medium access protocol • All implementations must support DCF
Wireless MAC • Aloha – inefficient • CSMA – listen before sending • Hidden node problem • Expose node problem Hidden node exposed node
Medium Access Collision Avoidance (MACA/MACAW ) • RTS-CTS-Data-ACK • Exposed nodes The media supports both B and C transmitting simultaneously. But the MAC does not The media supports both A and D transmitting simultaneously. But the MAC does not The above arguments are not true if we require ACKs. In wireless networks, it is very difficult to not have ACKs. And we will see that in some cases, higher data rates are permitted if acks are used. In this case, all transmission are two-way, and so there is no exposed node.
Basic Problems of MACA • Overhead (we’ll see this later) • If the CTS is not decoded, it does not mean that the node will not interfere. • Example • Suppose that the noise floor is -100 dBm and the receiver needs 7 dB or more SNR or SNIR. And suppose 40 loss in the first meter • HA,B = -64 dB • A transmits at 15 dBm • So B receives A’s transmissions at -89 dBm (well above the SNR limit, we could even use a higher data rate) • HC,B=-70 dB • B transmits CTS at 15 dBm • C receives CTS at -95 dBm and is unable to decode • C transmits and B receives interference at -95 dBm. • The SNIR at B is 6 dB, not enough for 1Mbps (much less the higher data rate that was selected). data A B C
Interference with RTS/CTS • More details • C will transmit if it does not hear B’s CTS. The received signal strength at B is the same that C could not receive, so it cannot be a very strong signal • How strong can it be? • Suppose that C is listening, there is interference from many transmitters so it cannot hear anything B’s CTS even though the signal is very strong. Then C could cause very high interference. • Solution: clear channel assessment (one type) if C hears a strong signal, even if it cannot decode it, it cannot transmit for a short period, to allow B to receive (even if B is not receiving) • The interference at B is the same as the signal strength received by C from B (but unable to decode). • Option 1 • If C hears even a weak signal, then don’t transmit • Drawback: this will cause exposed node • Option 2 • If C hears a stronger signal, then don’t transmit • Drawback: B might receive strong interference • 802.11 defines a strong signal as around -75 dBm. Which is quite strong, and implies that B can expect interference of nearly -75 dBm! It is difficult to imagine a setting where this is useful for interference from 802.11 (it could be from some other source). In 802.11, if the received signal strength is -75 dBm, it should likely be decodable (unless sent at high data rate, but CTS are always sent at low data rate) • In conclusion, C will receive B’s CTS when the SNIR is high enough. If the interference at C is strong, then C could cause strong interference.
Overhead in 802.11 (ignoring backoff) DIFS RTS SIFS SIFS Data SIFS CTS ACK Check that the channel is idle Packet arrives from network layer RTS arrives at receiver CTS arrives at sender data arrives at receiver done NAV
Frame sizes (and contents) MAC frames are called MPDU (MAC protocol Data Unit) • Frame formats • RTS (20B) • Frame control (2B) • Protocols version (2bits) (802.11 = 00, so if not 00, then the following means something different) • Type (2 bits)- management, control,data, reserved • subtype (4bits), e.g., control|RTS, management|beacon,… • To DS (DP) (1 bit) • From DS (1bit) • More frag (1b) • This is a Retry (1b) • Pwr mgt (1b) (=1 means the STA will go to power save (i.e., periodic sleep) after the frame is received) • More data (1b) – if a STA is in power save mode, then this is sued to signify that more data is waiting, so maybe the STA should wake up and get the data!) • Protected frame (1b) (WEP, etc.) • Order (1b) • Duration / ID (2B) (but MSB is zero, so 15 bits for the duration) in microsec (rounded up) (ID is used in power save mode) • RA receiver address (6B) • TA transmitter address (6B) • FCS – CRC-32 (4B) • CTS (14B) • Frame control • Duration = duration in RTS – time to transmit CTS – 1 SIFS • RA (no transmit address) • FCS • ACK (14B) • Frame control • Duration = 0 unless it is an ACK one for several more frames of a fragmented packet • RA • FCS • Data frame 34B overhead • Frame control (2B) • Duration (2B) • Address 1 (6B) • Address 2 (6B) • Address 3 (6B) (homework, what are these addresses used for) • Sequence control 2B (for fragmenting) • Address 4 (6B) • Frame body, up to 2312B • FCS 4B
Physical Protocol Data Unit (802.11b) • PLCP (phy layer convergence protocol = phy with MAC frame inside) • PLCP preamble (144 bits) • SYNC (128b) – to let the physical layer acquire sync (transmitter actually begins transmitting before this. This period is called ramp up where the transmitter goes from 0 zero to full/desired power. This typically takes ~ 1us) • SFD (16b)– like a radio ID, IEEE 802.11 has SFD-F3A0hex • PLCP header (48 bits) • Signal (8b) – specifies the modulation • Service (8b) – reserved • Length (16b) in micro seconds • CRC (16b) • PLCP SYNC and header always take 192 us • MPDU (MAC protocol data unit) or PSDU (PLCP service Data unit)
Short PLCP (optional) • Preamble – 72 bits at 1 Mbps • PLCP header – 48 bits at 2Mbps • The preamble and header at 96 us • MPDU / PSDU at 2, 5.5, … Mbps • Since the PLCP is sent at 2 Mbps, the channel must be able to support 2 or more Mbps • In practice the short header does not work well unless the channel is very good.
802.11 g • PLCP preamble and header are similar to 802.11b • PSDU • Long Training sequence • 1.6 us guard interval • Long training symbol (3.2 us) • Long training symbol (3.2us) (again) • Total = 8 us • OFDM signal • Guard interval 0.8 us (why why why is it reduced to 8 us?) • Signal 3.2 us • Total 4 us • Transmitted at 6 Mbps OFDM modulation – which is the worst performing modulation – it has the same SNR-BER relationship at 9 and 12 Mbps. • Data • 6 us of quiet time (time to decode) to SIFS is 16 us
802.11 overhead • Four 192b from PLCP • At 1Mbps = 192 microsec so 768 microsec • 82B from RTS, CTS, ACK, data • 656 mic sec at 1Mbps • 328 micro sec at 2Mbps • 131 at 5Mbps • 60 at 11Mbps • 12 at 54Mbps • 1 DIFS, 3 SIFS, 4 propagation delays • 30 microsec + 3*20 micro + 4*2 microsec = 98 microsec • Total overhead time • 1552 mic sec at 1Mbps • 1194 micro sec at 2Mbps • 997 at 5Mbps • 925 at 11Mbps • 878 at 54Mbps • Data duration • 40B packet • 320 mic at 1Mbps => efficiency data duration over all duration = 17% • 160 2Mpbs => 11% • 64 => 6% • 29 => 3% • 6 at 54 Mbps => 0.6% • 1500B packet • 12000 mic at 1Mbps => 89% • 6000 2Mpbs => 83% • 2400 => 70% • 1090 => 54% • 222 at 54 Mbps => 20% • Small packets and high data rates are a problem. Increasing packet size would help efficiency (similar things occur in wired TCP). • 802.16 can transmit multiple frames with one set of overhead
802.11 overhead w/o RTS/CTS • two 192b from PLCP • At 1Mbps = 192 microsec so 384 microsec • 14B from RTS, CTS, ACK, data • 112 mic sec at 1Mbps • 56 micro sec at 2Mbps • 22 at 5Mbps • 10 at 11Mbps • 2 at 54Mbps • 1 DIFS, 1 SIFS, 1 propagation delays • 30 microsec + 20 micro + 2 microsec = 52 microsec • Total overhead time • 548 mic sec at 1Mbps • 492 micro sec at 2Mbps • 458 at 5Mbps • 446 at 11Mbps • 438 at 54Mbps • Data duration • 40B packet • 320 mic at 1Mbps => efficiency data duration over all duration = 36% • 160 2Mpbs => 24% • 64 => 12% • 29 => 6% • 6 at 54 Mbps => 1% • 1500B packet • 12000 mic at 1Mbps => 96% • 6000 2Mpbs => 92% • 2400 => 83% • 1090 => 70% • 222 at 54 Mbps => 33%\ If the packet is smaller than RTSThreshold, then RTS/CTS are not used
802.11 without RTS/CTS • See write-up
Contention • When a collision does occur, the transmitters must back-off • Also, if a node desires to transmit (i.e., a frame is received from the upper layer), and it finds the channel busy, it should no transmit as soon as the channel is free. • Random backoff • Backoff time = Random()*SlotTime • Random uniform on [0,CW] • CWMin(7) <= CW <= CWMax (1023) – depending on the previous contention experienced during this transmission • SlotTime = 20 us (CW=1023 -> ~10 ms of idle time between transmission attempts. At 1Mbps, this is a bit more than one 1500B packet.) • The backoff is a timer, for ever slot time that is DIFS after the channel is not busy, the timer decrements. • When the time reaches 0, the node attempts to transmit. • If the transmission fails, CW is doubled (but is never greater than CWMax) • After a successful transmission or the so many attempts were tried that the packet was dropped, CW = CWMin
Max number of transmissions • If an RTS fails to generate a CTS, then the Short Retry count is incremented, until is reaches the short_retry_limit • If the pkt size is below RTSThreshold (so no RTS is transmitted), and fails to generate an ACK, then the Short retry count is incremented • When small data pkt is successful, the short retry count is reset to zero • If pkt size is above RTSThreshold and fails to result in an ACK, then the long retry is incremented. • If the pkt is successfully transmitted, then long retry count is reset to 0 • If a pkt is dropped due to too many failed attempts, the retry counters are set to zero
Using ACKs to increase data rate • Envelop analysis • Suppose that the pkt error prob for 10 Mbps is 0 and the pkt error prob for 20 Mbps is 0.25. • The effective data rate at 10 Mbps is 10 Mbps, and at 20, is 0.75*20 > 10, so 20 is better, even though there will be many losses and retransmissions. • When a transmission fails, it must wait DIFS and then back off • Suppose an empty channel • DIFS, Data, SIFS, ACK • If Fails, then wait, [0,7]*SlotTime and repeat the above • If Fails again, then wait, [0,13]*SlotTime and repeat the above
PCF – point coordination function • Provide contention free (CFP) access • Also, allows contention (CP) • AP is master – it polls STAs • The beacon is used to set STA’s NAV • The AP does not have to use PCF. But if it does, the STAs must obey. But they will anyway, since it obeys DCF • When a STA joins, it announces whether it can be polled. • The AP may select to only transmit data during the CF and force STAs to use the CP to transmit • All transmission must be ACKed. • If no ACK is received, the AP can retransmit the next time the AID comes up (AID = association ID) • The AP can retransmit after waiting at least PIFS
PCF – centrally controlled access • When the AP wants to transmit, it waits for the current transmission to complete and then waits for PIFS and begins to transmit (the PCF may have to wait a long time to begin the CFP). Since the PIFS is smaller than the DIFS, the AP will always get the channel over the mobile hosts. • When the AP has control of the channel it is called the contention-free period (CFP). • Then the AP is not in a CFP, then hosts can use the DCF to transmit. • The CFP begins with the AP sending a beacon. • During the CFP the AP polls the mobile hosts • Beacon • Timestamp (64b), beacon interval (units are 1024microsec) (16b) and capability info • CFP max duration – every mobile host saves this info into its NAV. • Service set ID (SSID), supported rates, phy parameters, CF parameters, IBSS parameters, traffic indication map • 802.11d • Country info and hopping pattern parameters • 801.22e • QBSS and EDCA parameters • 802.11g • ERP info • 802.11h • Power constraint, supported channels, channel switch announcement and quiet info • 802.11i • RSN info
CFP All nodes update their NAV to 32000 us beacon Data+CF-poll Data+CF-ACK from station1 Data+CF-ACK+CF-poll Data+CF-ACK from station1 Data+CF-ACK+CF-poll PIFS PIFS PIFS SIFS SIFS PIFS Trans error Data+CF-ACK+CF-poll Data+CF-ACK from station1 Data+CF-poll PIFS No ack SIFS No data sent, but data was received Data+CF-ACK+CF-poll ACK+CF from station PIFS SIFS Station had no data to send, so AP regains control after PIFS CF-ACK+CF-poll Data+CF-poll PIFS One problem is that the mobile station may send very large packets at slow rates and hence use the channel for a long time PIFS CF-ACK+CF-end End of CFP All nodes update their NAV PIFS
Tim and DTim • A CFP will occur after a beacon. • But it does not occur after every beacon • Every few beacons is a DTIM and every few DTIMs is a CFP • In the TIM is a bit string with a 1 if the STA with the AID corresponding to the bit number has a data packet ready to delivery from the AP. • This string is 2008 bits long. • Node not in the list can sleep • Since beacons are transmitted after listening, the CFP might be a bit late • If the channel found to be busy, then the AP backoff between 0 and CWMin. This is to reduce collisions with other APS
Transmit Power control (TPC) in 802.11 – 802.11h • 802.11a is at 5GHz, so it might interfere with other users. 802.11a is not the primary user of 5GHz. • To control the impact of 802.11a on these other spectrum users, 802.11a must attempt to reduce the power and the AP must monitor the total power over all mobile nodes. • Note that all 802.11a cards must have the capability to control transmission power. • Power control is communicated via a new frame called the action frame • The mobile maintains a variables local power constraint and country RF power constraint channel TX power. The later depends on the country and the channel. The former is received from the AP. The mobiles transmit power is country RF power constraint channel TX power - local power constraint . • The mobile will report to the AP • Its min and max possible transmission power • The current transmit power • Current average link margin • The AP will also demand that the mobiles are quiet so radars can be detected • 802.11h supports changing to a different channel if there is not enough power allowance on the current channel. Also, a mobile might be rejected from joining a BSS if there are power problems (but this is not specified in the spec)
802.11 types • 802.11a,b,g • 802.11h = 802.11a with power/spectrum control • 802.11d international operation • 802.11f mobility, inter-AP communication • 802.11e QoS • 802.11i security enhancements • 802.11s mesh
Features in 802.11e • 8 priority classes • Hybrid coordination function • Contention free period (CFP) and contention period • During the contention period, the AP may still gain control and poll nodes. • During CP the AP may poll all nodes but with specific priority with a single poll. In this way the nodes with high priority can compete among each other and not with low priority traffic. • During the controlled contention, nodes can only send requests for transmission ops (TXOPs) • The requests are replied to with acks from the ap • The AP polls nodes and issues TXOPs (transmit opportunities) • The CFP ends with at the time specified in the initial beacon or when a CFP end packet is sent • Within packet header is a place for node’s queue size as well as a place to request a TXOP. • Thus, the AP knows all queue sizes (but it only know the aggregate of each nodes queue size, not each priority class queue.
Features of 802.11e • Enhanced DCF • Each nodes has 4 or 8 queues – one for each priority class • The queues compete for the channel internally • If the queue collide, then they increase their CW. But high priority queues increase there CW more slowly than low priority queues • Block ACKs • Instead of ACKing every packet, it is possible to ack several packets. • The block ack contains the list on packets ACKed. • The block ACK can be requested • Or it can automatically occur at the end of the transfer, in which case the block ack is acked. • No ACK
BSS • A BSS is the set of nodes associated with an AP • An IBSS is an ad hoc BSS • A DS (distribution system) connects the BSS to other BSS and forms a EBSS • But usually the BSS do not directly communicate and so the BSSs are independent, and don’t form a EBSS. • But this might not be the case in future networks • To support mobility • To support QoS/hi-capacity
Associated • First a client must be associated • Then it become authenticated
802.16 / WiMax • The original WiMAX standard, IEEE 802.16, specifies WiMAX in the 10 to 66 GHz range. • 802.16a added support for the 2 to 11 GHz. • IEEE 802.16 provides up to 50 km (31 miles) • data rates up to 70 Mbit/s • "last mile" connectivity
Frame Control Header (FCH), Frame Control Header (FCH), specifies the burst profile and the length of one or more DL bursts The DL-MAP, UL-MAP, DL Channel Descriptor (DCD), UL Channel Descriptor (UCD), and other broadcast messages that describe the content of the frame are sent at the beginning of these first bursts. The remainder of the DL subframe is made up of data bursts to individual SS’s Each data burst assigned a burst profile that specifies the code algorithm, code rate, and modulation level that are used for those data transmitted within the burst