MAC Research HighlightMAC Research Highlight

Y.C. TsengY.C. Tseng

Outline: 3 Main Research Outline: 3 Main Research IssuesIssues• Analysis:

– G. Bianchi, “Performance Analysis of the IEEE 802.11 Distributed Coordination Function”, IEEE J-SAC, 2000.

– K. Kanodia et al., “Ordered Packet Scheduling in Wireless Ad Hoc Networks: Mechanisms and Performance Analysis”, ACM MobileHoc 2002.

• Protocols:– R. Garces and J. J. Garcia-Luna-Aceves, "Collision Avoidance and R

esolution Multiple Access with Transmission Groups", INFOCOM 2007.

– B. P. Crow, J. G. Kim, & P. Sakai, "Investigation of the IEEE 802.11 Medium Access Control (MAC) Sublayer Functions", INFOCOM'97.

– R. O. Baldwin, N. Davis, and S. F. Midkiff, "A Real-time Medium Access Control Protocol for Ad Hoc Wireless Local Area Networks", ACM MC2R, Vol. 3, No. 2, 1999, pp. 20-27.

• Handover latency reduction:– H. Kim, S. Park, C. Park, J. Kim, and S. Ko, “Selective

Channel Scanning for Fast Handoff in Wireless LAN using Neighbor Graph”, ITC-CSCC 2004, July 2004.

– S. Shin, A. S. Rawat, H. Schulzrinne, "Reducing MAC Layer HandoffLatency in IEEE 802.11 Wireless LANs", ACM MobiWac'04, Oct, 2004.

– C.C. Tseng, K.H. Chi, M.D. Hsieh, and H.H. Chang, “Location-based fast handoff for 802.11 networks”, IEEE Communications letters, vol. 9, issue 4, pp. 304- 306, April 2005.

Research Highlight: DCF Performance Analysis

Ref: G. Bianchi, “Performance Analysis of the IEEE 802.11 Distributed Coordination Function”, IEEE J-SAC, 2000.

Assuming saturation situation (stations always have packets to transmit), the work analyze the DCF performance.

state of a station: (s(t), b(t)) s(t): backoff stage (0, 1, …, m) of the station

CWmax = 2m Wmin

Let Wi = 2i W.

b(t): backoff counter value p: colliding probability (a constant)

State Transition Diagram of Backoff

Some Important Transitionsstart backoff

failure, next stage

failure, max stage

backoff 1 step

succ

essf

ul tr

ans.

Research Highlight: Unfair Access

Ref: K. Kanodia et al., “Ordered Packet Scheduling in Wireless Ad Hoc Networks: Mechanisms and Performance Analysis”, ACM MobileHoc 2002.

As there are multiple wireless links coexisting, some unfairness problem may arise.

Scenario 1: Asymmetric Information throughputs ratio of A to B = 5% : 95% reason: B knows more information than A does

AB

Scenario 2: Perceived Collision throughputs of A : B : C = 36% : 28% : 36% reason: Due to spatial reuse, flow A and C can capture the ch

annel simultaneously, thus causing flow B to reserve consecutive NAVs.

Proposed solution: “Distributed Wireless Ordering Protocol” an ordered distributed packet scheduling for MAC can be based on any reference scheduler, such as FIFI, Virtua

l Clock, Earliest Deadline First.

AB C

Research Highlight:Collision Avoidance and Resolution

Multiple Access with Transmission Groups

R. Garces and J. J. Garcia-Luna-Aceves

INFOCOM’97

Abstract

a CARMA-NTG protocol for accessing wireless media CARMA-NTG = Collision Avoidance and Resolution Multip

le Access Protocol with Non-persisitent Trees and transmission Group

Based on transmission group Once obtaining the medium, a station will have its right to ke

ep on sending. based on RTS/CTS messages

Concept of Cycles

Dynamically divide the channel into cycles of variable length. Each cycle contains a contention period and a group-

transmission period. The group-transmission period is a train of packets sent by

users already in the group. New users contend to join transmission group by

contending during the contention period.

A, B, C Y, A, B, C

X Y Z

Z, Y, A, B, C X, Z, Y, A, B, Cmedia

: contention period

: group trans. period

Each STA Needs to Keep Track of …

To send in the transmission period, each station must know the following environment parameters: the number of members in the transmission group its position within the group the beginning of the each group-transmission period the successful RTS/CTS exchange of new users in the

previous contention period

Group-Transmission Period

A station transmits once the previous station’s packet is received. The spacing is twice the propagation delay.

If this is not heard during this period, assume that the previous station fails its membership is removed from the group the failed station has to contend to join the group later.

A B C A C A C

B contend later

B’s transmission exceeds propagation delay

Contention Period

Contending based on RTS/CTS exchange. The contention period terminates once the first station succ

essfully join the group. Each station runs the NTG scheme (non-persistent tree and

transmission group) Each station keeps the following variables:

a unique ID LowID and HiID: to denote the current contention window i

n the current contention periodcontention window: the allowable ID’s that can contendan ID not within this range can not contend

a stack: the future potential contention windows

NTG Scheme

Initially, LowID=1 and HiID=(max. ID in the system) On RTS conflict, all stations divide (LowID, HiID) into

(LowID, (LowID+HiID)/2) ((LowID+HiID)/2 + 1, HiID) // i.e., binary split

PUSH the first part into STACK Contend if its ID is within the latter part. If no RTS is heard after channel delay, POP the stack and r

epeat recursively.

ONLY stations in the RTS state can persist in trying. new stations: backoff and wait until the next period already-in-group stations: not until they leave the group

Contention Example

A system with 4 stations: n00, n01, n10, n11. n00 and n01 are contending.

n11idle

n10idle

n01RTS

n00RTS

(00, 11)

before 1stcollision

(10, 11)

after 1stcollision

allowed interval

(00, 01)

after idle

(00, 01)

after 2ndcollision

(01, 01)

(00, 00)

after n01success

(00, 11)

n01RTS

n00RTS

n01RTS

packets

(a)

(a) (b)

(b) (c)

(c) (d)

(d)

Short Summary

propose the concept of group transmission Only one RTS/CTS exchange is used for transmitting a train

of packets better fairness than IEEE 802.11

NTG (non-persistent tree group) keeps the contention cost low.

Performance: on high load, similar to TDMA on low load, better than TDMA by getting rid of empty slots

Research Highlight:Polling Issue in IEEE 802.11

Research Highlight:Polling Issue in IEEE 802.11

“Investigation of the IEEE 802.11 Medium Access Control (MAC) Sublayer Functions”, B. P. Crow, J. G. Kim, & P. Sakai, INFOCOM’97.

Problem Statement In the PCF function of IEEE 802.11, it is

NOT specified how to poll STAs. Problem: how to do voice communicatio

n using PCF? Assuming that all voice packets have the s

ame priority. Voice stream characteristic:

ON-and-OFF process ON = talking; OFF = listening

talk silent

low probability

low probability

A “Round Robin” Approach AP keeps track of the list of STAs to be

polled. When CFP begins, the AP polls the STAs s

equentially.If the AP has an MPDU to send, the poll and M

PDU are combined in one frame to be sent. O/w, a sole CF-Poll is sent.

When CFP ends, the AP keeps track of the location where the polling stops.Then resume at the same place in the next CFP.

(cont.) Within a CFP_Repetition_Interval, if an

STA sends no payload in k polls, the STA is dropped from the polling list. k is an tunable parameter

In the next CFP, the STA will be added back to the list again.

Basic Idea: to avoid useless polling.

Simulation results: Smaller k gives better data throughput

(Fig. 14). k = 1~5 does not affect the voice delay

(Fig. 15).

Short Summary An interesting polling mechanism based

on specific applications.

Future directions: how to support other types of media.

Main Idea: “send your next backoff value”

ACM Mobile Computing and Communication Review, 1999, Vol. 3, No. 2, pp. 20-27.

A Real-Time Medium Access Control Protocol for Ad Hoc Wireless Local

Area Networks

Review of IEEE 802.11 The CW (contention window) is initially

CWmin, and is doubled after each failure, until CWmax is reached. BV (backoff value) randomly in [0..CW-1]. The BV is decreased after each idle slot.

Drawback of IEEE 802.11 Can not meet the requirements of real-

time communication. When a packet has missed its deadline,

the packet will still be buffered and sent. Thus, this causes more contention,

collisions, ...more packets may miss their deadlines.

There are 4 rules (next few pages).

Rule 1:Enhanced Collision Avoidance

Announcing the next BV: When a packet is transmitted, the next BV

to be used is placed in a field of the packet. Stations who hear this packet will avoid sel

ecting this BV as their next backoff timer.BV is a random number in [0..CW-1].

Details: Prior to transmitting a packet, a station will

select its next BV from the range of [0..CW-1], excluding those BV’s already chosen by other stations.

A station will indicate in its data packet the next BV value to be used.

A station should keep a table of BV values used by other stations.After an idle slot, a station should decrease its

own BV, as well as others’ BVs in its table.

Example: A: 3 1 8 B: 1 6 ... C: 5 2 (collides with B’s, changed to 3)

B(6) A(1) A(8) C(3) B(...) C(...)0 1 2 3 4 5 6 7 …

Rule 2:Transmission Control

A station must send when its BV value has expired.

If the packet experiences transmission failure, it will be reexamined to see if its deadline has been missed. Note: another backoff still has to be taken.

Rule 3:Contention Window Size

CW is set to 8N, where N is the estimated number of “real-time” stations. N: can be estimated by counting the number of

unique addresses for a period of time. [alternative] N: a function of current channel load. “8” is chosen by instinct.

Note: CW is thus not doubled after a transmission failure (compared the original IEEE 802.11 of doubling

each time).

Rule 4: Collision of BV Due to mobility, transmission error, and

collisions, a station may receive a packet indicating a BV equal to its own BV.The station must select another BV value;

otherwise, collision will occur. To avoid the station being unduly

penalized, the new BV should be selected from [0..CBV-1].CBV = its current BV.I.e., the station is given higher priority.

If all values in [0..CBV-1] are chosen, then we double it (i.e., [0..2*CBV-1]).

Collision Ratio RT-MAC is quite stable in collision prob.

with respect to the number of stations.

Short Summary A new RT-MAC protocol.

broadcasting the next BV value BV depends on the current number of

stations

Results: The network behavior is quite stable in terms

of mean delay, missed deadline ratio, and collision ratio.

The mean delay is quite independent of the number of stations.

Research Highlights How to reduce handover time?

How to reduce handover time?

Channel scanning in 802.11 is very time-consuming if all channels need to be scanned. If scanning one channel ta

kes 30 ms, the toally 300-400 ms is needed.

Research Highlight:Fast Channel Scanning by Neighbor Graph Ref: H. Kim, S. Park, C. Park, J. Kim, and S. Ko, “Selectiv

e Channel Scanning for Fast Handoff in Wireless LAN using Neighbor Graph”, ITC-CSCC 2004, July 2004.

Method: A concept called neighbor graph (NG) is proposed. From the

NG provided by an external server, a MH only needs to scan the channels that are used by its current AP’s neighbors. About 10 ms are needed to scan a specific neighbor.

Research Highlight:Fast Channel Scanning by Caching

Ref: S. Shin, A. S. Rawat, H. Schulzrinne, "Reducing MAC Layer HandoffLatency in IEEE 802.11 Wireless LANs", ACM MobiWac'04, Oct, 2004.

Method: MH maintains a cache which contains a list of APs adjacent t

o its current AP. The cached data was established from its previous scanning.

Only the two APs with the best RSSI were cached.

During handoff, the cached APs are searched first. If this fails, scanning is still inevitable.

Research Highlight:Fast Channel Scanning by Location Information

Ref: C.C. Tseng, K.H. Chi, M.D. Hsieh, and H.H. Chang, “Location-based fast handoff for 802.11 networks”, IEEE Communications letters, vol. 9, issue 4, pp. 304- 306, April 2005.

Method: MH can predict its movement path and select the potential A

P. A location server is needed to provide information of APs. So a MH can re-associate with its new AP directly without g

oing through the probe procedure. However, this scheme relies on a precise localization method.

Other ReadingsOther Readings

• Medium Access ControlMedium Access Control– R. Garces and J.J. Garcia-Luna-Aceves, “Floor Acquisition Multiple AR. Garces and J.J. Garcia-Luna-Aceves, “Floor Acquisition Multiple A

ccess with Collision Resolution,” Proc. ACM/IEEE MobiCom 96, Rye, ccess with Collision Resolution,” Proc. ACM/IEEE MobiCom 96, Rye, New York, November 11-12, 1996.New York, November 11-12, 1996.

– Z. Tang and J.J. Garcia-Luna-Aceves, “Hop-Reservation Multiple AcZ. Tang and J.J. Garcia-Luna-Aceves, “Hop-Reservation Multiple Access (HRMA) for Ad-Hoc Networks,” Proc. IEEE INFOCOM '99, New Ycess (HRMA) for Ad-Hoc Networks,” Proc. IEEE INFOCOM '99, New York, New York, March 21--25, 1999.ork, New York, March 21--25, 1999.

– V. Bharghavan, A. Demers, S. Shenker and Lixia Zhang, “MACAW: A V. Bharghavan, A. Demers, S. Shenker and Lixia Zhang, “MACAW: A Media Access Protocol for Wireless LAN's,” Proceedings of SIGCOMMedia Access Protocol for Wireless LAN's,” Proceedings of SIGCOMM 94, pp.212-225.M 94, pp.212-225.

– P. Karn, “MACA - A New Channel Access Method for Packet Radio,” P. Karn, “MACA - A New Channel Access Method for Packet Radio,” ARRL/CRRL Amateur Radio 9th Computer Networking Conference, AARRL/CRRL Amateur Radio 9th Computer Networking Conference, April 1990, pp.134-140.pril 1990, pp.134-140.

– Romit Roy Choudhury, Xue Yang, Ram Ramanathan, and Nitin VaidyRomit Roy Choudhury, Xue Yang, Ram Ramanathan, and Nitin Vaidya, “Using Directional Antennas for Medium Access Control in Ad Hoa, “Using Directional Antennas for Medium Access Control in Ad Hoc Networks,” ACM International Conference on Mobile Computing ac Networks,” ACM International Conference on Mobile Computing and Networking (MobiCom), September 2002.nd Networking (MobiCom), September 2002.