Point Coordination Function

sensed idle. Once the back-off time reaches zero, the station gets the right to access the medium. If two or more stations finish the back-off countdown procedure simultaneously, they may start to send frames in the same time. Unfortunately, a collision will occur. In this situation, the sender will not receive an ACK from the receiver. After a time period ACKtimeout, the sender assumes that a collision or some failure happened. For this reason, the stations need to do back-off process again to schedule a retransmission.

  Figure 2.3: DCF timing sequence. 

2.1.2 Point Coordination Function (PCF)  

Beside the DCF, IEEE 802.11 also offered an optional access mechanism method, PCF for supporting “contention-free” services. This scheme is often used to transmit the real-time frames. The PCF scheme shall work on an infrastructure wireless networks, it is because that the point coordinator (PC) should be implemented on an access point (AP). The AP manages the access to the medium in the contention free period (CFP) by polling stations sequentially.

When the PCF mode is used, medium access time will be divided into periodical beacon intervals. The beacon intervals are consisting of two parts, the contention free period (CFP) and the contention period (CP). In a CP, STAs need to contend for the medium for transmitting their frames. While in a CFP, a PC maintains a poll list which recoded the information of registered stations and polls each of them according to the order in the list.

When a station is being polled, it starts to transmit frames after a SIFS (see Figure 2.4). For the sake of channel utilization, during the CFP the AP shall use appropriate data subtypes based on the following rules:

‹ Data + CF-Poll, Data + CF-ACK + CF-Poll, CF-Poll, and CF-ACK + CF-Poll shall only be sent by an AP.

‹ Data, Data + CF-ACK, Null, and CF-ACK may be sent by an AP or by the CF-Polled STAs.

  Figure 2.4: PCF timing sequence. 

  The time period for an AP to generate beacon frames is called target beacon transmission time (TBTT). Usually, PCF uses a round-robin polling based scheduler to poll each station sequentially according to the order of the polling list, but priority-based polling mechanisms can also be used if different QoS levels are requested by different stations.

2.2 Introduction of IEEE 802.11e

There are two MAC access mechanisms provided by the IEEE802.11. However, both of these two mechanisms do not distinguish different services for different traffic streams. In other words, IEEE 802.11 does not support Quality of Service (QoS) requirement. For solving this problem, the IEEE 802.11 Working Group E has proposed a new MAC function, called HCF (Hybrid Coordination Function). The IEEE 802.11e MAC architecture is shown in Figure 2.5.

  Figure 2.5: IEEE 802.11e MAC architecture [1]. 

HCF mainly offers two new access methods, one is contention-based mechanism called Enhanced Distributed Channel Access (EDCA) and the other is a contention-free mechanism called HCF Controlled Channel Access (HCCA). These two mechanisms will be introduced in the following sections.

2.2.1 EDCA (Enhanced Distributed Channel Access)

IEEE 802.11e EDCA is an extension of legacy 802.11 DCF. The difference between

them is whether they can support QoS or not. EDCA provides differentiated, distributed channel access for packets with eight different user priorities ranged from zero to seven. In EDCA mode, packet traffics are classified into four different FIFO queues (see Figure 2.6), named Access Categories (ACs). By the way, one or more user priorities are mapped to one Access Category (AC). The mapping between eight user priorities and ACs is shown in Table 2.1.

IEEE 802.11e proposed the use of different Inter-Frame Spaces (IFSs) according to different ACs. Every AC behaves as a virtual station with its own parameter set, including

arbitration inter-frame space (AIFS), minimum contention window size (CWmin), maximum contention window size (CWmax) and TXOP duration, to differentiate the traffic. These parameters are announced by the QAP periodically via beacon frames. Instead of using DIFS, EDCA uses Arbitration IFS (AIFS). The value of AIFS is determined by Equation (2).

AIFS = AIFSN × aSlotTime + SIFS (2)

Where the value of AIFS Number (AIFSN) is an integer greater than zero and is depending on each AC. Besides, the values of aSlotTime and SIFS are defined in the physical layer.

The IFS relationship and related terms is shown in Figure 2.7. It shows that the AC with the smallest AIFS will have the highest priority. Because each AC can be viewed as a virtual station, an AC with higher priority has larger chance to access the medium than those with lower priority ACs. However, the back-off timers of different ACs in a station may reach zero at the same time, this leads to virtual collision. If a virtual collision occurred, it can be resolved by sending the packet with higher priority. But the packet with lower priority is considered as encountering a collision, thus the corresponding AC needs to double its contention window. That is, after a success transmission, the contention widow of corresponding AC is reset to CWmin. However, the contention window is doubled after each unsuccessful transmission.

  Figure 2.7: IFS relationship and related terms. 

IEEE 802.11e EDCA is proposed for supporting QoS and it provides differentiated, distributed channel access for packets with different priorities. However, the resetting mechanism of contention window may cause large variations of contention window size, and increase the probability of collisions, especially in the situation of heavy traffic.

2.2.2 The procedure of QoS flow setup

The Traffic Specification (TSPEC) is used to describe the traffic characteristics and the QoS requirements of a traffic stream. The main purpose of the TSPEC is to reserve resources within the HC and modify the HC’s scheduling behavior. It also allows other parameters associated with the traffic stream to be specified, these include a traffic classifier and acknowledgment policy [1]. TSPEC contains the information such as nominal MSDU size, maximum MSDU size, minimum service interval, maximum service interval, data rate, delay bound and so on. A maximum required service interval (RSI) refers to the maximum duration between the start of successive TXOPs that can be tolerated by the requesting applications.

This information is useful to set up a QoS connection. The format of TSPEC is shown in Figure 2.8.

  Figure 2.8: TSPEC element format [1]. 

When a new QoS flow is created, QSTA uses ADDTS request with TSPEC to request an admission of the QoS flow. The QAP replies with an ADDTS response to the QSTA regarding

whether the request is admitted or not into this QIBSS. Figure 2.9 shows the sequence of messages occurring at a traffic stream setup. The QAP will put this traffic stream into its polling list and poll it periodically in HCCA mode. After the transmission of traffic stream is ending up, QSTA uses the DELTS request to inform the HC to remove this traffic stream from its polling list.

Figure 2.9: The sequence of messages occurring at a traffic stream setup [1]. 

2.2.3 HCCA (HCF Controlled Channel Access)

The HCCA is a centralized access mechanism controlled by the Hybrid Coordinator (HC), it resides in the QoS Access Point (QAP), and it can be viewed as an extension of PCF in which polling is only allowed during contention free period (CFP); but in HCCA mode, QAP can poll stations during either CFP and contention period (CP). Each QoS station (QSTA) may have up to eight established Traffic Streams; a traffic stream is characterized by a Traffic Specification (TSPEC) which is negotiated between a QSTA and a QAP. Mandatory fields of the TSPEC include: Mean Data Rate, Delay Bound, and Nominal PDU Size. For all established streams the QAP is required to provide a service that is compliant with the negotiated TSPEC under controlled operating conditions. 802.11e compliant stations must be

able to process the additional frames which are reported in Table 2.2.

Table 2.2: QoS frames. 

QoS frames  QoS piggybacked frames 

QoS Data  QoS Data + CF‐Ack 

QoS CF‐ACK  QoS Data + CF‐Poll 

QoS Null  QoS Null + CF‐Ack 

QoS CF‐Poll  QoS Data + Cf‐Poll + CF‐Ack 

Figure 2.10 shows an example of IEEE 802.11e beacon interval used in HCCA. During a beacon interval, a QAP is allowed to start several contention-free bursts called Controlled Access Phases (CAPs). A CAP is a time interval during which the QAP may either transmit MSDUs of established downlink traffic streams or poll one or more QSTAs by specifying the maximum duration of the transmission opportunity (TXOP). A QSTA is never allowed to exceed the TXOP limit imposed by the QAP, including inter-frame spaces and acknowledgments. If a polled QSTA has no data to send, then the QSTA responds with a QoS-Null frame.

  Figure 2.10: An example of 802.11e beacon interval used in HCCA. 

Upon receiving a TSPEC, the HC invokes it scheduler in order to perform admission control and scheduling. The admission control function is used to determine whether resources are available to serve the requested TSPEC; and the scheduling function is used to determine the manner the HC will poll traffic streams so as to satisfy their QoS requirements.

As a result of this operation, the HC may decide to accept, reject or propose an alternative TSPEC to the requesting QSTA.

In IEEE 802.11e sample HCCA scheduler [1], the schedule for an admitted stream is calculated in two steps. The first step is to calculate the scheduled SI. In the second step, the HC would calculate the TXOP duration with scheduled SI for the streams.

The calculation of the scheduled service interval is done in two steps. First, the scheduler calculates the minimum, SImin, of all maximum SIs (MSI) for all admitted traffic streams.

Second, the scheduler chooses a value that is a sub-multiple of beacon interval and is smaller than SImin. This value is the scheduled SI for all admitted TSs.

The next procedure is the calculation of the TXOP duration for an admitted traffic stream.

First, the HC calculates the number of MSDUs that arrived with the mean data rate during the

SI by _i ^ρi MSDU size of stream i. Then the HC calculates the TXOP duration as the maximum of time to transmit Ni frame with rate Ri and time to transmit one maximum size MSDU with

transmission rate Ri by _i max ^{i i} ,

When a new traffic stream requests registration, the admission control process can be done in three steps. The HC needs to use above two formulas to calculate the values of Ni and TXOPi after first two steps. Finally, the HC determines whether the stream can be admitted when the following inequality is satisfied,

1

Where TXOPnew is the calculated TXOP duration of the new stream and the second term, is the sum of TXOP durations of existed streams, BI indicates the beacon interval and Tcp is the time used for EDCA traffic. Moreover, the HC needs to ensure that it doesn’t allocate TXOP durations that exceed dot11CAPlimit. This algorithm ensures that a new incoming traffic stream doesn’t occupy all the remaining time period.

2.3 Comparison between WLAN MACs

In the previous sections, we know that there are two types of MAC access functions in WLAN. One is contention-based, such as DCF and EDCA; the other is poll-based which includes PCF and HCCA.

The advantage of contention-based access functions is that they are adaptive to the migration in the network condition and are suitable for high load of traffic. Another advantage is the simplicity in implementation. The complexity of contention-based access function is lower than the poll-based access functions due to its distributed access characteristics.

However, this kind of access function has some disadvantages. For example, when multiple transmissions contend for the same channel, the problems of collision and hidden-node may be encountered. The random back-off mechanism is another problem for provisioning QoS, it can’t guarantee QoS for real-time traffic.

On the other hand, there are several advantages in using poll-based mechanism; one is eliminating the hidden-node problem; another one is QoS guarantee, this is due to the characteristics of contention-free and centralized-based mechanism. Here the channel utilization is much better than that of contention-based functions because of the reduction of back-off time overhead and collision problem. However, the implementation complexity of

poll-based functions is higher than contention-based functions. And the poll-based functions need to be implemented in the infrastructure wireless networks. The overhead of polling is also an issue in using poll-based access functions.

2.4 Multimedia traffic attributes

In this section, we will introduce the QoS traffic attributes such as number of frames per second (fps), frame duration and required bandwidth.

In implementation, VoIP traffic generally can be classified into two types. One is the variable bit rate (VBR), with which the codec generates VoIP packets while in talking; the other is constant bit rate (CBR), with which the codec generates VoIP packets whether the user is talking or not. CBR codec generates VoIP packets in the fixed interval. In the Table 2.3, we can see that VoIP packets feature some characteristics: high transmission interval and small payload size. Also an interesting point is that the different VoIP codec may use the different transmission frequencies.

Table 2.3: Speech Codec standards 

In a common video data flow, the frame generation of codec is 30 fps (frame per second) with NTSC format. It means that the receivers need to handle more than 30 packets from video source per second. Generally speaking, the sensitivity of delay for a video is lower than that for a voice, therefore the priority of transmission for the video is lower than the voice.

Some of video applications are Constant Bit Rate (CBR), but for the sake of video quality, some video applications use Variable Bit Rate (VBR).

2.5 Related Works

According our survey, most of existing EDCA enhanced schemes, such as CWmin, TXOP duration, persistence factor, and AIFSN were trying to adjust the parameters of EDCA.

In [22] and [23], authors are trying to tune the CWmin. In [22], it offered an Adaptive EDCF scheme. The QAP tunes the CW sizes for different classes after receiving the average collision rate measures from different QSTAs. The study in [23] recommends using different values of CWmin and CWmax for different priorities, in which higher priority has lower CWmin and CWmax values than those in lower priorities. Some adaptive schemes were proposed to change the persistence factor (PF), the study in [24] proposes a method based on the back-off increase function. In the original DCF, the CW is multiplied by a PF of 2 after each collision. The method in this work is that the higher priority traffic has a lower PF. In [25], it suggests a way that adapts CW according to channel conditions and adjusts its value depending on the network utilization rather than generating a new random CW.

About HCCA, the reference scheduler of 802.11 task group assumes that all types of traffic are constant bit rate (CBR), so the queue length increases linearly according to the constant application data rate. However, many real-time applications, such as video, are variable bit rate (VBR) traffic. Hence, the basic HCCA scheduler is not suitable for VBR

traffic. Some improvements are offered in works [8] and [15]. Fair Hybrid Coordination Function (FHCF) [8] scheme tries to address VBR traffic by adjusting the TXOP of each flow using queue length estimations. In [15], it offers a bound-based earliest due date (SETT-EDD) scheduling algorithm and uses the additional information from applications. The SET-EDD scheduler also takes into account the impact of link adaptation. Regarding the survey of polling scheme, we found some polling methods are trying to find a mechanism in HCCA to avoid unnecessary polling in order to reduce the polling overhead [7]. Some works [3] [5]

were trying to change the polling order for the VoIP traffic according to the state of traffic streams.

Chapter 3

Proposed Scheme

In the previous chapter, we know that the DCF and EDCA are neither effective nor efficient to support delay-sensitive voice traffic. It is because their contention-based nature and binary exponential back-off mechanism can’t guarantee the delay bound of multimedia packet transmissions. For this reason, in order to guarantee the delay requirement of time-sensitive services, the HCCA function is preferred for real-time applications in WLAN, in which the AP polls each voice source periodically. The IEEE802.11e standard proposes a reference HCCA scheduler that is efficient for the traffic flows with strict CBR characteristics.

However, lots of applications such as VoIP, and video which have VBR characteristics and the reference scheduler could not adapt to this kind of traffic.

In this chapter, we propose a highly efficient polling (HEP) scheme based on HCCA to provide better QoS in wireless LAN. This proposed scheme is used to eliminate the problem of polling overhead, access latency, jitter and utilization of channel. The HEP scheme operates on the MAC layer in the AP to manage the polling schedule. Table 3.1 shows the comparison of our proposed polling scheme with round-robin polling scheme.

Table 3.1: The comparisons of polling schemes. 

Scheme  Round‐robin polling  Proposed polling scheme 

Polling order  Sequential (fixed)  Dynamic 

Complexity of poll scheduling  O(1)  O(n)^*1 

Searching time  Less than 1 µs^*2  Less than 1 µs^*2 

Complexity of implementation  Low  Low 

*1: n is the number of traffic stream

*2: based on n < 100

The HEP scheme is fully compatible with current 802.11e MAC, that is, it works on both HCCA and EDCA modes. Further, the HEP scheme takes not only the real-time but also the non-real-time traffic streams into consideration. HEP will put the time-sensitive streams into HCCA mode and put other non-real-time streams into EDCA mode. Hence, the guarantee of QoS for real-time streams can be realized.

Besides the EDCA mode, HEP can also support bi-directional communications in HCCA mode. When the polled stream is determined, the HC would check the queue of download stream. If download queue of real-time stream is available, HC will send QoS data with CF-Poll to the corresponding QSTA; otherwise, the HC will only send a QoS CF-Poll to QSTA. Moreover, the HC could also send data to the QSTAs after the polling procedure as long as the CAP is not over.

3.1 Main Architecture

The proposed HEP scheme maintains a polling list for keeping the information of traffic streams. The elements of polling list will be introduced in Section 3.2.1. HEP consists of three parts, Polling Order Selector (POS) module, Estimated Time Management (ETM) module, and Silence Handler (SH) module.

The POS module is responsible for choosing the suitable entry from the polling list, and to determine TXOP duration for traffic stream. When CFP or CAP starts, the POS module needs to query the next polling station information from the polling list, and send its decision to the frame sending module to poll the selected QSTA.

When HC’s MAC receives frames from stations, it will decode the header of frames. If

the MAC identifies the frame type as data, it needs to decode the extended QoS control field, and send an event message to ETM module. The ETM module will adjust the estimated polling time of traffic stream according to the value specified in the extended QoS control field of the data frame.

Moreover, if the HC’s MAC receives a QoS-Null frame from stations, it shall send an event message to inform the SH module. The SH module is responsible for identifying the

Moreover, if the HC’s MAC receives a QoS-Null frame from stations, it shall send an event message to inform the SH module. The SH module is responsible for identifying the