802 11 e Wireless LANs Speaker Fu-Yuan Chuang Advisor Ho-Ting

802. 11 e Wireless LANs Speaker：Fu-Yuan Chuang Advisor：Ho-Ting Wu Date： 2007. 01. 02 1

Outline Introduction to IEEE 802. 11 e Tuning of 802. 11 e Network Parameters Adaptive Contention-Window MAC Algorithms for Qo. S-Enabled Wireless LANs 2

Introduction to IEEE 802. 11 e New terminology QAP – Qo. S Access Point QSTA – Qo. S Station HC – Hybrid Coordinator A new mechanism defined in IEEE 802. 11 e -- Hybrid Coordination Function (HCF) HCF is implemented by all QAPs and QSTAs HCF has two access mechanisms Contention based ○ Enhanced distributed channel access (EDCA) Controlled channel access ○ HCF Controlled Channel Access (HCCA) 3

Architecture 4

Comparison of 802. 11 and 802. 11 e CAP controlled access phase : A time period when the HC maintains control of the medium after gaining medium access by sensing the channel to be idle for a PIFS duration 5

Transmission Opportunity (TXOP) Under HCF, TXOP is basic unit of transmission TXOP types EDCA TXOP ○ Obtained by a QSTA winning an instance of EDCA contention during the CP HCCA TXOP ○ Obtained using the controlled channel access Polled TXOP ○ by a non-AP QSTA receiving a Qo. S (+)CF-Poll frame during the CP or CFP 6

Enhanced Distributed Channel Access (EDCA) EDCA defines four Access Categories (AC) Voice Video Best Effort Background EDCA supports 8 User Priority (UP) values Priority values (0 to 7) identical to the IEEE 802. 11 D priorities Rules One UP belongs to one AC (Access Category) Each AC may contains more than one UP Traffic of higher UP will be transmitted first within one AC 7

EDCA-Access Category (AC) Four access categories (ACs) that support the delivery of traffic with differentiated UPs An AC is an enhanced variant of the DCF which contends for TXOP using the following parameters: CWmin[AC], CWmax[AC], AIFS[AC]. Each AC queue functions as an independent DCF STA and uses its own backoff parameter. In EDCA, the Contention-Window (CW) size and the Inter-frame space (IFS) is AC dependent 8

EDCA Details Each AC has its own Interframe space – AIFS Back off Counter CWmin, CWmax, CW TXOP limit QSTAs obtains these info from beacon frames Each QSTA implements own queues for each AC If internal collision happens, the frame with higher priority will be sent 9

EDCA Parameter Set element ACM : admission control mandatory ACI : Access category identify 10

Contentions among Different ACs in EDCA Contention among EDCAFs (AC, AIFS, CWmin , CWmax ) to win a TXOP 11

Default Values for Each AC AIFS[AC] = AIFSN[AC] × a. Slot. Time + a. SIFSTime. DIFS=2*a. Slot. Time +a. SIFSTime AC AIFSN CWmin/CWmax AC_VO 2 3 -7 AC_VI 2 7 -15 AC_BE 3 15 -1023 AC_BK 7 15 -1023 12

Tuning of 802. 11 e Network Parameters Juliana Freitag, Nelson L. S. da Fonseca, and Jos´e F. de Rezende, “Tuning of 802. 11 e Network Parameters, ” IEEE Communications Letters , Volume 10, Issue 8, Aug. 2006 Page(s): 611 - 613 13

Introduction A novel mechanism for tuning the access parameters of 802. 11 e QAP and QSTAs To solve the asymmetry problem To produce balanced uplink and downlink delays The network can operate under much higher loads 14

Introduction - asymmetry problem the existing asymmetry between the uplink and downlink delays which occurs when using the 802. 11 e contention method The QAP is responsible forwarding all traffic to/from QSTAs Since both QAP and QSTAs have the same probability of accessing the medium, the queues in the QAP can rapidly build up, increasing the downlink delay 15

16

The Proposed Approach The adjustment of TXOP value is used to improve throughput of the classes at the QAP leading to more balanced delay 17

Idea - TXOP 假設每個QSTA在每個Class中有一個flow，第 i個Class的TXOP允許q個frames被傳送 則在QAP中第i個Class的TXOP的值調整為可 以傳送k*q個frames (k為第i個class中downlink flow的數目) 可讓downlink throughput接近uplink throughput 18

19

Mechanism Load最小的Class將其TXOP設為 0 只可傳送一個frame 其他class根據一個ratio調整TXOP The ratio between the load that has arrived at their queues and the load that has arrived at the queue of the class with the lowest load 可讓load大的queue獲得較大的傳送時間 20

21

CW 當STA數目不多時，low CWmin 可以減少idle 的時間並增進channel utilization 當STA數目過多時，high CWmin 可以避免 collision As the number of stations with active flows of a certain class increases, the CWmin value of this class should increase as well as that of all classes with lower access priority 22

Mechanism 若STA(在class i中有active flows)數目大於 CWmin(i)，則CWmin(i) is increased to the next power of 2 minus 1 反之，則CWmin(i) value is reduced to the immediate lower power of 2，縮短idel的時間 23

24

25

Adaptive Contention-Window MAC Algorithms for Qo. S-Enabled Wireless LANs Samer El Housseini, Hussein Alnuweiri, “Adaptive Contention-Window MAC Algorithms for Qo. S-Enabled Wireless LANs, ” wireless Networks, Communications and Mobile Computing, 2005 International Conference on , Vol. 1, pp. 368 - 374, June 2005 26

Saturation Throughput 27

The Throughput Derivative Algorithm after each CW change, the TD Algorithm employs measurements of the throughput taken in the AP CWmax is also changed while keeping the same ratio (CWmax / CWmin) 28

The Throughput Derivative Algorithm The throughput derivative is taken over the present and a few past measurements of the throughput If the derivative is positive The AP continues to increase the CW If the derivative is negative The AP decrease the CW 29

The Throughput Derivative Algorithm 30

References IEEE-802. 11 WG, “IEEE Standard for Information technology Telecommunications and information exchange between systems - Local and metropolitan area networks - Specific requirements Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) specifications Amendment 8: Medium Access Control (MAC) Quality of Service Enhancements, ” November 2005 Juliana Freitag, Nelson L. S. da Fonseca, and Jos´e F. de Rezende, “Tuning of 802. 11 e Network Parameters, ” IEEE Communications Letters , Volume 10, Issue 8, Aug. 2006 Page(s): 611 – 613 Samer El Housseini, Hussein Alnuweiri, “Adaptive Contention. Window MAC Algorithms for Qo. S-Enabled Wireless LANs, ” wireless Networks, Communications and Mobile Computing, 2005 International Conference on , Vol. 1, pp. 368 - 374, June 2005 31