System Model

Figure 2.1: Indoor SOHO WLAN Infrastructure Environment

2.1.1 System Architecture

For an indoor SOHO WLAN environment shown in figure 2.1, there is a QoS access point (QAP) having only two interfaces, an Ethernet interface and an 802.11e interface. The Ethernet/ADSL link supports the data traffic and the real time service traffic sent from or to

the Internet, and the 802.11e interface is used to transmission of the packets in the wireless environment. In this thesis, the end host is called the quality of service station (QSTA) for which we intend to provide the quality of service (QoS) facility. In the infrastructure environment of Small Office, Home Office (SOHO), we assume that all video, voice, and data packets are sent through the QAP even though in some cases, when both QSTAs are within in the coverage area of an QAP, voice and data packets could not be sent directly between the QSTAs.

AIFSdata AIFS_vo

Figure 2.2: An example of a superframe timing

2.1.2 System Operation

In a superframe, it can be divided into two periods: controlled access phase (CAP) and contention period (CP) as shown in Fig. 2.2. In the CAP, the QAP polls the QSTAs to get the uplink voice frames or transmits the downlink voice frames and the downlink video packets directly without acknowledge based on HCF controlled channel access. In the CP, the non-real time data frames and intended contending voice frames of QSTAs would use different priorities to contend the channel based on EDCF.

The CAP within a superframe can be divided into two transmission phase: Voice CAP, and Video CAP. In addition, the voice service within a Voice CAP also can be sequentially served into three transmission period: bi-direction voice transmission, uplink-only (UL-only)

voice transmission, and downlink-only (DL-only) voice transmission. In a superframe, there is at most one voice frame for uplink transmission of a QSTA, and also the QAP has at most one downlink voice frame for each QSTA with the conversation. At the beginning of the Voice CAP, the QAP decides the polling sequence based on the polling list. The first voice service of polling sequence is bi-direction voice transmission. If the QAP has a downlink voice frame queued for transmission to the QSTA in the polling list at the QAP, the downlink voice frame and QoS(+)CF-Poll can be combined and transmitted as a single frame by the QAP after the channel is sensed to be idle for a PIFS interval. After receiving the frame, the destination of the QSTA can transmit an uplink voice frame to the QAP after a SIFS interval.

This frame exchange procedure is called bi-direction voice transmission. After bi-direction voice transmission, for the QSTA in polling list but the QAP does not have a voice frame queued for it, the QAP sends a sole QoS(+)CF-Poll frame to the QSTA after the channel is sensed to be idle for a PIFS interval, and waits to receive an uplink voice frame from the QSTA after a SIFS interval. This frame exchange procedure is called uplink-only (UL-only) voice transmission. If the QAP receives a QoS Null frame from the QSTA during bi-direction voice transmission or UL-only voice transmission, the QAP regards the QSTA entering into silence period and removes the QSTA from the polling list in the QAP. After UL-only voice transmission, the QAP consecutively sends the remaining downlink voice frames with no acknowledge after waiting a PIFS channel idle interval during downlink-only (DL-only) voice transmission.

In the Video CAP, the QAP sends the multiple video packets belong to the QSTA in a burst and Round Robin manner after the channel is sensed to be idle for a PIFS interval.

During the frame interval, the QAP receives several video packets of the frame interval belong to the QSTA from Ethernet/ADSL link and the QAP may transmit these video packets during the following three CAPs in a burst without acknowledge which are separated by SIFS interval in the video CAP. These video packets in the QAP are lasting availably for the following three CAPs or until they are transmitted completely. While these video packets are not completely transmitted by QAP during the following three CAPs, the QAP would not transmit these packets which are dropped by the QAP, and the QAP will transmit the at most packets of next frame interval. During the following two CAPs, if there are enough residual time in the CAP bound for a video packet transmission, the QAP sends the packet belong to the QSTA which is at service turn and also sends the remaining packets at next video CAP in first after the channel is sensed to be idle for a PIFS interval. The video packets belong to a

QSTA may be transmitted in two consecutive video CAPs. Each video burst transmission on the wireless medium is corresponding to several video packets arriving at the QAP during a frame interval intended to a QSTA. After the QAP transmits a burst of video packets to a QSTA, the QAP continues to transmit another video burst of the next QSTA in a cyclic manner (that is so called Round Robin) after a PIFS channel idle interval. In addition, after all bursts of the frame interval are transmitted, the QAP consecutively transmits the bursts of the next frame interval if they are all arrived at QAP. In addition, until all bursts of video packets arriving at QAP before the superframe are transmitted completely by the QAP, and QAP sends a CF-end frame as termination of the video CAP. Fig. 2.2 shows an example of a superframe timing.

In the CP, the voice frames for uplink but not polled during the voice CAP in the same superframe will contend with data frames. The data frames including QAP downlink data frames and QSTA uplink data frames. The QAP should assign different AIFS, CW , and _min CWmax for the contending uplink voice service and data service. After the QAP receives an uplink contending successful voice frame from the QSTA, the QAP re-adds the QSTA in the polling list. When the counter of a data frame contending for transmission counts down to zero and there is enough time for a complete data frame exchange in CP, the QSTA would transmit the data frame. Also, it happens to a data frame having zero-counter in the beginning of the CP because there is not enough access time for the frame exchange in the last CP. In the beginning of the CP, if a data frame is first transmitted, the voice service would delay for at least a data frame transmission period in spite of the collision happened or not. In order to guarantee the voice delay as small as possible, the AIFS for voice service is set to equal to PIFS, denote as AIFS_vo =PIFS. And the CW of a voice frame, denote as _min CW_{min_vo}, and contending voice packet number of the superframe which may be zero if N_NP is equal to zero. Also the CW_{min_vo} and CW_{max_vo} is the smallest integer greater or equal to the mean value of NC. Here, we guarantee that the contending uplink voice service has a higher priority

than non-real-time data service. That is the contending voice frames are always served before the data frames. In a CP, when a data frame of a QSTA is served in the CP, it stands for contending voice frames has been completely served and hereafter only the data frames contend with each other in the CP. In order to avoid the uplink contending voice frame to be terminated by non-real-time data frames, the AIFS of non-real-time data service, denote as AIFSdata is given by

max_vo data

for the first frame transmitted of a QSTA in a CP others

where the “+1” is for contending uplink voice frames avoiding collision with a data frame having zero-counter. The minimum and maximum contention window of a data frame,CW_{min_data} and CW_{max_data}, could be variable set by the retransmit times of the data frame. And CW_{min_data} and CW_{max_data} of the first frame of a QSTA in a superframe are the same and not changed even during the retransmissions. If a data frame is retransmitted five times and it is still not successful, the data frame is dropped by the QSTA.