802.11e for QoS
- the new standard to provide Quality of Service, QoS for 802.11 Wi-Fi applications
Wi-Fi technology based on the 802.11 standard is now widespread in its use. Not only is it used to provide real wireless LAN (WLAN) functionality, but it is also widely used to provide localised mobile connectivity in terms of "hotspots". A variety of flavours of the IEEE 802.11 are available: 802.11a, 802.11b, 802.11g, and these different standards provide different data throughput speeds and operate on different bands.
One of the major shortfalls for the developing applications for Wi-Fi is that it is not possible to allocate a required quality of service for the particular application. Now with IEEE 802.11e the Quality of Service or QoS problem is being addressed.
The need for QoS
The issue of Quality of Service, QoS on 802.11 Wi-Fi is of particular importance in some applications, and accordingly 802.11e is addressing it. For surfing applications such as internet web browsing of sending emails, delays in receiving responses or sending data does not have a major impact. It results in slow downloads, or small delays in emails being sent. While it may have a small annoyance to the user, there is no real operational impact on the service being provided. However for applications such as voice or video transmission such as Voice over IP, VoIP, there is a far greater impact and this creates a much greater need for 802.11e. Delays, jitter and missing packets result in the system loosing the data and the service quality becoming poor. Accordingly for these time sensitive applications it is necessary to be able to prioritise the traffic. This can only be done by allocating a service priority level to the packets being sent, and this is now all being addressed by IEEE standard 802.11e.
The way in which data is transmitted and controlled has a major impact on the way that QoS is achieved. This is largely determined by the way the Medium Access Control (MAC) layer operates. Within 802.11 there are two options for the MAC layer. The first is a centralised control scheme that is referred to as the Point Coordination Function (PCF), and the second is a contention based approach called Distributed Coordination Function (DCF). Of these few manufacturers of chips and equipment have implemented PCF and the industry seems to have adopted the DCF approach.
The PCF mode supports time sensitive traffic flows to some degree. Wireless Access Points periodically send beacon frames to communicate network management and identification which is specific to that WLAN. Between the sending of these frames, PCF splits the time frame into a contention free period and a contention period. If PCF is enabled on the remote station, it can transmit data during the contention free polling periods. However the main reason why this approach has not been widely adopted is because the transmission times are not predictable.
The other scheme, DCF uses a scheme called Carrier Sense Multiple Access with Collision Avoidance (CSMA/CA). Within this scheme the MAC layer sends instructions for the receiver to look for other carriers transmitting. If it sees none then it sends its packet after a given interval and awaits an acknowledgement. If one is not received it then it knows its packet was not successfully received. It then waits for a given time interval and also checks the channel before retrying to send its data packet.
In more exact terms the transmitter uses a variety of methods to determine whether the channel is in use, monitoring the activity looking for real signals and also determining whether any signals may be expected. This can be achieved because every packet that is transmitted includes a value indicating the length of time that transmitting station expects to occupy the channel. This is noted by any stations that receive the signal, and only when this time has expired may they consider transmitting.
Once the channel appears to be idle the prospective transmitting station must wait for a period equal to the DCF Inter-Frame Space (DIFS). If the channel has been active it must first wait for a time consisting of the DIFS plus a random number of back off slot times. This is to ensure that if two stations are waiting to transmit, then they do not both transmit together, and then repeatedly transmit together.
A time known as a Contention Window (CW) is used for this. This is a random number of back-off slots. If a transmitter intending to transmit senses that the channel becomes active, it must wait until the channel comes free, waiting a random period for the channel to come free, but this time allowing a longer CW.
While the system works well in preventing stations transmitting together, the result of using this access system is that if the network usage level is high, then the time that it takes for data to be successfully transferred increases. This results in the system appearing to become slower for the users. In view of this WLANs may not provide a suitable QoS in their current form for systems where real time data transfer is required.
The problem can be addressed by introducing a Quality of Service, QoS identifier into the system. In this way those applications where a high quality of service is required can tag their transmissions and take priority over the transmissions carrying data that does not require immediate transmission and response. In this way the level of delay and jitter on data such as that used for VoIP and video may be reduced.
To introduce the QoS identifier, it has been necessary to develop a new MAC layer and this has been undertaken under the standard IEEE 802.11e. In this the traffic is assigned a priority level prior to transmission. These are termed User Priority (UP) levels and there are eight in total. Having done this, the transmitter then prioritises all the data it has to waiting to be sent by assigning it one of four Access Categories (AC).
In order to achieve the required functions, the re-developed MAC layer takes on aspects of both the DCF and PCF from the previous MAC layer alternatives and is termed the Hybrid Coordination Function (HCF). In this the modified elements of the DCF are termed the Enhanced Distributed Channel Access (EDCA), while the elements of the PCF are termed the HCF Controlled Channel Access (HCCA).
Of these the EDCA provides a mechanism whereby traffic can be prioritised but it remains a contention based system and therefore it cannot guarantee a give QoS. In view of this it is still possible that transmitters with data of a lower importance could still pre-empt data from another transmitter with data of a higher importance.
When using EDCA, a new class of interframe space called an Arbitration Inter Frame Space (AIFS) has been introduced. This is chosen such that the higher the priority the message, the shorter the AIFS and associated with this there is also a shorter contention window. The transmitter then gains access to the channel in the normal way, but in view of the shorter AIFS and shorter contention window, this means that the higher the chance of it gaining access to the channel. Although, statistically a higher priority message will usually gain the channel, this will not always be the case.
The HCCA adopts a different technique, using a polling mechanism. Accordingly it can provide guarantees about the level of service it can provide, and thereby providing a true Quality of Service level. Using this the transmitter is able to gain access to a radio channel for a given number of packets, and only after these have been sent is the channel released.
The control station which is normally the Access Point is known as the Hybrid Coordinator (HC). It takes control of the channel. Although it has an IFS, it has what is termed a Point Coordination IFS. As this is shorter than the DIFS mentioned earlier, it will always gain control of the channel. Once it has taken control it polls all the stations or transmitters in the network. To do this it broadcasts as particular frame indicating the start of polling, and it will poll each station in turn to determine the highest priority. It will then enable the transmitter with the highest priority data to transmit, although it will result in longer delays for traffic that has a lower priority.
By Ian Poole
Want more like this? Register for our newsletter