CogFi A Cognitive WiFi Channel Hopping Architecture for

  • Slides: 30
Download presentation
Cog-Fi: A Cognitive Wi-Fi Channel Hopping Architecture for Urban MANETs Sung Chul Choi and

Cog-Fi: A Cognitive Wi-Fi Channel Hopping Architecture for Urban MANETs Sung Chul Choi and Mario Gerla WONS 2012 Presentation

Motivation 2

Motivation 2

Motivation 3

Motivation 3

Network Model Mobile node Fixed Interfering source 4

Network Model Mobile node Fixed Interfering source 4

Network Model Network node ch 8 Interfering source 3 5 1 4 • Goal:

Network Model Network node ch 8 Interfering source 3 5 1 4 • Goal: 1 Avoid the channels used by interfering sources using a cognitive multi-channel scheme. 5

Solution Preview q To avoid interfering sources: § Use Cognitive radio technology to sense

Solution Preview q To avoid interfering sources: § Use Cognitive radio technology to sense channel load and discover lightly loaded channels q To maintain P 2 P network connectivity in spite of unpredictable interferers: § Exploit multi channel diversity: a node can receive on multiple channels via Cognitive Channel Hopping § Guarantee neighbor discovery and rendezvous in a finite # of steps(using the QUORUM set) § Design routing algorithm that accounts for “multichannel links” and Channel Hopping 8

Cognitive Channel Hopping q Cognitive Channel Hopping (CCH) § Single-radio, channel-hopping solution in which

Cognitive Channel Hopping q Cognitive Channel Hopping (CCH) § Single-radio, channel-hopping solution in which each node picks its channels based on the load sensed on them f f t f t t 9

CCH: Protocol Operation q A node x periodically triggers Channel Quality Assessment (CQA). §

CCH: Protocol Operation q A node x periodically triggers Channel Quality Assessment (CQA). § A channel availability vector a = {a 1, …, a|C|} is produced. • In this work, ai = 1 - [channel load in i]. § Based on channel availabilities, x picks a channel set Q = {q 1, …, qk} from a predefined Quorum list (any two Q-sets have at least one common element) C = {0, 1, 2, … , 11, 12} § It picks the channel set with the highest combined channel quality, defined as: Example list of channel sets, each with size k = 5. 13

CCH: Protocol Operation q Given Q, x generates two hopping sequences, utx and urx.

CCH: Protocol Operation q Given Q, x generates two hopping sequences, utx and urx. Mrx Mtx Q = {0, 1, 3, 9, 12} k=5 0 1 3 9 12 0 1 3 9 12 0 1 3 9 12 12 0 1 3 9 utx 0 1 3 9 12 0 1 3 … urx 0 1 3 9 12 0 1 9 12 0 … |utx| = |urx| = k 2 = 25 14

CCH: Channel Rendezvous Property q Claim: A channel rendezvous of a pair of nodes

CCH: Channel Rendezvous Property q Claim: A channel rendezvous of a pair of nodes is guaranteed to occur within k 2 slots. Qx = {0, 1, 2} Qy = {2, 3, 4} By the property of a quorum system, there exists at least one common channel. utx(x) … Mtx(x) Mrx(y) 0 1 2 2 3 4 2 4 2 3 2 appears in the same 2 appears exactly column, every row. once in each column. 0 1 2 urx(y) … 2 3 4 2 4 2 3 … … 15

CCH: Channel Rendezvous Property q This still holds when two sequences are not in

CCH: Channel Rendezvous Property q This still holds when two sequences are not in sync. Qx = {0, 1, 2} Qy = {2, 3, 4} By the property of a quorum system, there exists at least one common channel. utx(x) … urx(y) … Mtx(x) Mrx(y) 0 1 2 2 3 2 4 3 4 2 4 3 2 appears in the same 2 appears exactly column, every row. once in each column. 0 1 2 … 2 3 4 2 4 2 3 … 16

CCH: Protocol Operation q When x has no packet to transmit, it follows urx(x).

CCH: Protocol Operation q When x has no packet to transmit, it follows urx(x). q When x has packets to transmit, it follows utx(x) to locate the neighbor. q A channel rendezvous is guaranteed within the length of utx(x), k 2. channel quality assessment slot 0 1 2 3 4 5 6 7 8 9 1 1 0 1 2 channel quality assessment time Has packets to No send to packets to send. more y. tx slot rx slot 18

CCH: Protocol Operation q Within a slot, a conventional RTS/CTS-based packet exchange is made.

CCH: Protocol Operation q Within a slot, a conventional RTS/CTS-based packet exchange is made. § By default, a slot is 10 ms, enough to fit in tens of MAC frames. x y backoff RT S DATA CT S AC K time q Retransmissions occur within a slot and over multiple slots. 19

CCH: Learning of utx and urx q Learning hopping sequences § Every CCH frame

CCH: Learning of utx and urx q Learning hopping sequences § Every CCH frame includes information about the hopping sequences that the transmitter is using. § If node x has received a frame from y, it can later use its cache to predict which channel y will be without scanning channels with utx(x). x y 20

CCH: MAC-level Broadcast q Broadcast function is critical in making upper-layer mechanisms to work

CCH: MAC-level Broadcast q Broadcast function is critical in making upper-layer mechanisms to work (e. g. , 1 routing). ? 2 § Not all neighbors are in the same channel as 3 you! 1 ? 1 2 ? 4 ? q Each broadcast frame is kept in a separate buffer and transmitted in the transmitting channel (specified in utx) in the beginning of the slot, for multiple slots. 21

Solution: Cog-Fi Architecture q Cog-Fi is a cross-layer architecture with these modules: IP Routin

Solution: Cog-Fi Architecture q Cog-Fi is a cross-layer architecture with these modules: IP Routin g MAC CH-LQSR • Make a routing decision. channel load, link rate, BER • Coordinate channel access. • Store and maintain channel status. CCH SNR/BER PHY 802. 11 PHY • Regular 802. 11 PHY. 25

CH-LQSR: Motivation q Conventional on demand routing protocols like AODV and DSR are not

CH-LQSR: Motivation q Conventional on demand routing protocols like AODV and DSR are not well-suited. § Problem 1: Not all hops are equal. S T 26

CH-LQSR: Motivation q Conventional, hop-count based routing protocols like AODV and DSR are not

CH-LQSR: Motivation q Conventional, hop-count based routing protocols like AODV and DSR are not wellsuited. § Problem 1: Not all hops are equal. ps 54 Mb 1 18 Mbp s T S 18 Mb p s 2 bps 11 M 27

CH-LQSR: Motivation q Conventional, hop-count based routing protocols like AODV and DSR are not

CH-LQSR: Motivation q Conventional, hop-count based routing protocols like AODV and DSR are not wellsuited. § Problem 2: Broadcast does not occur simultaneously. S T 28

CH-LQSR: Motivation q Conventional, hop-count based routing protocols like AODV and DSR are not

CH-LQSR: Motivation q Conventional, hop-count based routing protocols like AODV and DSR are not wellsuited. § Problem 1: Not all hops are equal. • Use the channel load and link rates to quantifying the quality of each hop, and factor this in when computing routes. § Problem 2: Broadcast does not occur simultaneously. • Modify Route Discovery procedure. 29

CH-LQSR: ETTCH Metric u e v q Extend ETX and ETT [4, 5]. §

CH-LQSR: ETTCH Metric u e v q Extend ETX and ETT [4, 5]. § p: prob. that the packet transmission is not successful: p = 1 – (1 – pf) · (1 – pr) § s(m): prob. that the packet is delivered at m-th attempt. s(m) = pm – 1 · (1 – p) pf and pr, the forward and backward § The expected transmission count (ETX) packet error probabilities, areof link e = (u, v) is: computed based on the link BER reported from the PHY module. § Factoring in the link bandwidth and packet size, one can define the expected transmission time (ETT) of e as: • c: channel index • Sd: data packet size B, the bandwidth, is computed by taking the • B: bandwidth (data rate) of the channel load and link rates into account. 30

CH-LQSR: ETTCH Metric u e v q (cont'd. from the previous slide) § The

CH-LQSR: ETTCH Metric u e v q (cont'd. from the previous slide) § The multi-channel ETT of e, ETT(e), is: which is the avg. of ETTc(e) values over the channels in the channel set the receiver v is using. § Finally, Channel Hopping ETT of a path P is: 31

CH-LQSR: Protocol Operation q Extension of DSR § Route Discovery involving RREQ/RREP. § Once

CH-LQSR: Protocol Operation q Extension of DSR § Route Discovery involving RREQ/RREP. § Once a route is discovered, source routing is used. A How do I reach T? B S C D E Here I am! T 32

CH-LQSR: Protocol Operation q Extension of DSR § Route Discovery RREQ(T) involving. RREQ(T) RREQ/RREP

CH-LQSR: Protocol Operation q Extension of DSR § Route Discovery RREQ(T) involving. RREQ(T) RREQ/RREP Path: S-A ETTCH: 0. 012 A B S src S Path: S-A-B ETTCH: 0. 032 C RREQ(T) Path: S RREQ(T) D Route. Cache(E) path dest ETTCH S-A-BS-D 0. 076 0. 011 T C E Path: S-A-B-C Path: S-D ETTCH: 0. 076 0. 011 T 33

Cog-Fi: Evaluation Setup q Qual. Net 4. 5 § CCH: implemented as a full-fledged

Cog-Fi: Evaluation Setup q Qual. Net 4. 5 § CCH: implemented as a full-fledged MAC protocol. § CH-LQSR: implemented as a full-fledged routing protocol. q Channel environment § 13 orthogonal channels in the 5 -GHz band. § Interfering source: (x, y, tx_power, channel, active_%). q CCH parameters § § § Use RBAR for rate adaptation [8], using 802. 11 a rates. Channel set size k = 5. Channel switching delay: 80µs. Slot size: 10 ms. CQA Period: 3 seconds. 34

Cog-Fi: Evaluation Setup [COG] for single-radio nodes q List of schemes compared for evaluation

Cog-Fi: Evaluation Setup [COG] for single-radio nodes q List of schemes compared for evaluation CONTROL Symbol Description CCH+CH-LQSR Our Cog-Fi solution. CCH+DSR CCH with DSR. CCH+AODV CCH with AODV. 802. 11 a communicati on contr ol scannin g interferen ce time Single-channel 802. 11 a, routed using DSR. COG A conventional cog radio scheme, with DSR. RH+DSR Random hopping and DSR. 35

Cog-Fi: 25 -node (5 x 5) Grid Topology q 5 saturated 1500 -byte CBR

Cog-Fi: 25 -node (5 x 5) Grid Topology q 5 saturated 1500 -byte CBR streams for 5 random node pairs. 1 … … 5 40 m 7 … 2 3 2 … 4 8 36

Cog-Fi: 100 -node (10 x 10) Grid q 5 saturated 1500 -byte CBR streams

Cog-Fi: 100 -node (10 x 10) Grid q 5 saturated 1500 -byte CBR streams for 5 random node pairs. 1 … … 5 30 m 7 … 2 3 2 … 4 8 37

Summary q Goal: Devise a multi-channel multi-hop mechanism with the following requirements. § Interfering

Summary q Goal: Devise a multi-channel multi-hop mechanism with the following requirements. § Interfering sources should be avoided • A CCH node employs a cognitive radio-like channel sensing to identify lightly loaded channels. § The network connectivity must be maintained. • Exploit multi channel diversity: a node can receiver on multiple channels via Cognitive Channel Hopping • Guarantee neighbor discovery and rendezvous in a finite # of steps(using the QUORUM set) q Performance is further improved by CH-LQSR, ie by using a link metric that factors in channel load and link rates. 39

Thank You! 40

Thank You! 40