Survey of admission control in IEEE 802.11e wireless LANs

報告人：李宗穎

2

Outline

Introduction Background for IEEE 802.11e Admission Control for IEEE 802.11e Conclusion

3

Introduction (1/2)

802.11e standard provides a very powerful platform for QoS supports in WLANs

This report provide an extensive survey of advances in admission control algorithms/protocols in IEEE 802.11e WLANs

4

Introduction (2/2)

The purpose of admission control is to limit the amount of traffic admitted into a particular service class so that the QoS of the existing flows will not be degraded, while at the same time the medium resources can be maximally utilized

5

Background for 802.11e

EDCA (enhanced distributed channel access) Contention based

HCCA (HCF controlled channel access) Centralized control NOT much research work on the admission

control issue in HCCA

6

Background for 802.11e

7

8

802.11e EDCA mode

9

Admission Control Research

Admission Control for EDCA Measurement-based Model-Based

Admission Control for HCCA Do NOT covered in this report

10

Distributed Admission Control (1/2) Step 1 : via beacons the QAP announces the

transmission budget Step 2 : measure the amount of time occupied by the

transmission of each AC during each beacon interval transmission budget for an AC is depleted

new flow will not be able to obtain any transmission time existing flows will not be able to increase their

transmission time[ref.] Y. Xiao and H. Li, “Evaluation of Distributed Admission Control for the IEEE 802.11e EDCA,” IEEE Commun. Mag., vol. 42, no. 9, 2004, pp. S20–S24.

[ref.] Y. Xiao and H. Li, “Voice and Video Transmissions with Global Data Parameter Control for the IEEE 802.11e Enhance Distributed Channel Access,” IEEE Trans. Parallel Distrib. Sys., vol. 15, no. 11, 2004, pp. 1041–53.

11

Distributed Admission Control (2/2)

Shortcoming it is difficult to avoid network performance vibr

ation because a station always adjusts its transmission parameters at every beacon interval

this scheme does not provide direct relationships between those QoS requirements from applications

12

Two-Level Protection and Guarantee Mechanism (1/3)

First level protection each existing voice or video flow from new and

other existing QoS flows Second level protection

the existing QoS flows from best effort traffic

[ref.] Y. Xiao, H. Li, and S. Choi, “Protection and Guarantee for Voice and VideoTraffic in IEEE 802.11e Wireless LANs,” Proc. IEEE INFOCOM ’04, vol. 3,Hong Kong, Mar. 2004, pp. 2152–62.

13

Two-Level Protection and Guarantee Mechanism (2/3)

tried-and-known mechanism a new voice/video flow is first accepted

tentatively, and then tries to measure throughput and delay performance for some beacon intervals

early-protection mechanism the budget is below a certain threshold, new

flows are not allowed to enter

14

Two-Level Protection and Guarantee Mechanism (3/3)

too many best effort data transmissions can also degrade the existing QoS flows since many collisions might occur increase the initial contention window size (dyn

amic control EDCA parameter) for best-effort traffic

the problems of DAC is performance oscillation and lack of direct QoS relationships with applications

15

Virtual MAC and Virtual Source Algorithm (1/2)

The VMAC schedules virtual packets on the radio channel in the same way as real packets

it does not transmit anything but estimates the probability of collision if the virtual packet were “really” sent

[ref.] M. Barry, A. T. Campbell, and A. Veres, “Distributed Control Algorithms forService Differentiation in Wireless Packet Networks,” Proc. IEEE INFOCOM’01, vol. 1, Anchorage, AK, 2001, pp. 582–90.

[ref.] A. Veres et al., “Supporting Service Differentiation in Wireless Packet NetworksUsing Distributed Control,” IEEE JSAC, vol. 19, no. 10, 2001, pp. 2081–93.

16

Virtual MAC and Virtual Source Algorithm (2/2)

Advantage The advantage of these virtual algorithms is

that they do not cost any channel bandwidth Disadvantage

However, they need a lot of extra processing in each mobile host

17

Harmonica (1/2)

the AP periodically samples the link layer quality indicator (LQI) parameters, which include drop rate, link layer end-to-end delay, and throughput, for each class

select the channel access parameters that best match the QoS requirements

[ref.] L. Zhang and S. Zeadally, “HARMONICA: Enhanced QoS Support withAdmission Control for IEEE 802.11 Contention-based Access,” Proc. IEEERTAS ’04, Toronto, Canada, May 2004, pp. 64–71.

18

Harmonica (2/2)

the HARMONICA will select a traffic class i that best matches its QoS requirement and then execute an admission control processThe relative adaptation has reached a stable state BEthroughput – Reqthroughput > BEMin

how to find the optimal increment or decrement of the channel access parameters is still a challenging problem

19

Threshold-Based Admission Control (1/2)

Using relative occupied bandwidth Boccu = (TBusy/T) × 100%

Boccu < Blo : Admit the inactive AC with the highest priority

Boccu > Bup : Stop the transmission of the lowest active AC during the next period of T

[ref.] D. Gu and J. Zhang, “A New Measurement-based Admission ControlMethod for IEEE 802.11 Wireless Local Area Networks,” Mitsubishi Elec.Research Lab., Tech. rep. TR-2003-122, Oct. 2003.

20

Threshold-Based Admission Control (2/2)

Using average collision The average collision ratio is defined as Rc=Nc/Nt

Nc is the number of collisions that have occurred

Nt is the total number of transmissions

Similarly, there are two thresholds: the lower threshold Rlo and the upper threshold Rup

21

Threshold-Based Admission Control

If the NUC (Network Utilization Characteristic) of all the flows (NUC_total) is below the set NUC threshold (NUC_threshold), the flow can be admitted very easy to implement and can guarantee the QoS of h

igh priority flows when the medium is heavily loaded the issue of fairness is not considered difficult to set the NUC_threshold value

[ref.] S. Nor, A. Mohd, and C.Cheow, AN ADMISSION CONTROL METHOD FOR IEEE 802.11e, Network Theory and Applications, 2006.

22

Resource Sharing-Based Admission Control (1/2)

bandwidth is reserved for a particular traffic AC and also be shared among them

[ref.] A. Andreadis, G. Benelli, and R. Zambon, An Admission Control Algorithm for QoS Provisioning in IEEE 802.11e EDCA, 3rd ISWPC. Santorini, Greece, 2008.

AC_VO (20%) AC_VI (40%)AC_VO &

AC_VI (30%)

AC_BE & AC_BK (10%)

23

Resource Sharing-Based Admission Control (2/2)

Proposed Scheme Utilization percentages of the channel

UAC_VO≤0.5 & UAC_VI≤0.7 & UAC_VO+UAC_VI≤ 0.9

Reject new flow conditions UAC_VO ≥ 0.5, UAC_VI≥ 0.7, UAC_VO+UAC_VI≥0.9

Characteristic simplicity and fair to all different traffics the static partition of bandwidth which could lead to

an inefficient utilization of resources

24

Markov Chain Model-based Admission Control (1/3) Predicted achievable throughput for each flow, which is calcu

lated by

Psi is the probability of a successful transmission for flow i Pc, Ps and Pidle are the overall collision, overall successful tran

smission and overall idle probability Tc and Ts are the collision time and successful transmission ti

me, E[P] is the data payload

[ref.] D. Pong and T. Moors, “Call Admission Control for IEEE 802.11 ContentionAccess Mechanism,” Proc. IEEE GLOBECOM’03, vol. 1, San Francisco, CA,Dec. 2003, pp. 174–78.

25

Markov Chain Model-based Admission Control (2/3)

based on the two-state Markov Chain model proposed in [19], the transmission probability for flow i can be derived as

pi is the long-term collision probability for flow i,

W is the CWmin size for flow i, and b is the maximum backoff stage

[19] G. Bianchi, “Performance Analysis of the IEEE 802.11 Distributed Coordination Function,” IEEE JSAC, vol. 18, no. 3, 2000, pp. 535–47

26

Markov Chain Model-based Admission Control (3/3)

There are several problems in this algorithm the analytical model is derived under saturation

conditions, where each station always has packets to transmit

This research does not take account of virtual collision between different AC queues in one station

27

Contention-Window-Based Admission Control

The key idea of this scheme is to adjust the CW values for different stations so that the goals of admission control can be fulfilled

IEEE 802.11e WLAN is operating with a CW set {CW1, … , CWn} that meets the throughput requirements {Ri, … , Rn} for all stations

[ref.] A. Banchs, X. Perez-Costa, and D. Qiao, “Providing Throughput Guaranteesin IEEE 802.11e Wireless LANs,” Proc. 18th Int’l. Teletraffic Cong.,Berlin, Germany, Sept. 2003.

28

G/G/1-Based Admission Control (1/2)

channel utilization (cu) which are the decision criteria and average data rate (Rmean), peak data rate (Rpeak) and average packet length (PKl) are used to characterize

a flow’s bandwidth requirement as follows: cu=(R/PKl) T∗ suc, where R is the traffic rate

bandwidth requirement of a flow can be translated into (cumean, cupeak).

[ref.] X. Chen, H. Zhai, and X. Tian, Supporting QoS in IEEE 802.11e Wireless LANs, IEEE Transactions on Wireless LANs Communications, 2006.

29

G/G/1-Based Admission Control (2/2)

all admitted real-time flows into two parameters (cuA,mean, cuA,peak) and also estimate the average delay Di using the G/G/1 model

admit a new QoS flow, three requirements need to be satisfied cuA,mean + cui,mean < CUrt

cuA,peak + cui,peak < CUmax

average delay Di less than the delay bound Di

30

Parameters-Based Admission Control (1/2)

combination of the MAC parameters is from an heuristic real-time algorithm first, computes the minimum aggregated

bandwidth required by all flows Second, using this value and the achievable

maximum physical bandwidth, the data rate, the MAC parameters are adjusted based on a set of predefined thresholds

[ref.] B. Bellalta, M. Meo, and M. Oliver, VoIP Call Admission Control in WLANs in Presence of Elastic Traffic, IEEE Journal of Communications Software and Systems, 2007.

31

Parameters-Based Admission Control (2/2)

Characteristic This scheme sufficiently considers all EDCA

parameters and fairness between uplink and downlink

the values of threshold are difficult to set and some assumptions exist which is inaccurate to algorithm

32

Threshold-Based Admission Control

When new flow of ACi requests admission QAP estimates the equivalent number of compe

ting entities of class i predicts the achievable bandwidth and one-hop delay of the new flow

the analytical model of a non-saturated is for IEEE 802.11 DCF which is not accurate to admission control for EDCA

[ref.] B. Bensaou, Z. Kong, and D. Tsang, A Measurement- Assisted, Model-Based Admission Control Algorithm for IEEE 802.11e, The International Symposium on Parallel Architectures, Algorithms, and Networks. Sydney, Australia, 2008.

33

Conclusion There is many challenge in wireless admission co

ntrol How to model the heterogeneous wireless networks How to optimally map the QoS requirement between d

ifferent network layer How to dynamic change parameter according to cross-

layer conditionsMeasurement Model

Complexity Simple Complex

Theoretical Foundation Worse Better

Network Utilization High Low

Flexibility Good Poor