GB2420050A - WLAN admission control - Google Patents

Abstract

The present invention relates to admission control for new traffic on a contention based wireless network in order to ensure quality of service on the network. In general terms in one aspect the present invention provides an admission control method for a contention based wireless network which receives requests from new calls or data flows to access the network. The method determines the current bandwidth availability of the network from the current network utilisation and the network bandwidth. In one embodiment, the network utilisation is determined from a network busy time parameter, which can be determined by utilising the carrier sensing function in IEEE802.11 wireless network nodes for example. The call can then be admitted if the current bandwidth availability is greater than the bandwidth requirement of the requesting call.

Description

M&C Folio: GBP90340 Document: 1036084

WLAN ADMISSION CONTROL

Field of the Invention

The present invention relates to admission control for new traffic on a contention based wireless network in order to ensure quality of service on the network.

Background of the Invention

Real-time wireless applications (e.g. voice, video streaming) often have stringent quality of service (Q0S) requirements. Admission control for new applications or calls on a resource constrained wireless network is therefore necessary in order to maintain a minimum Q0S on the network.

There is a growing demand for Q0S support in IEEE 802.1 1-based wireless LANs. As the number of stations or traffic streams increases to the point of channel overload, network performance such as throughput and medium access delay will degrade significantly. Therefore some form of admission control is needed to restrict the volume of traffic in order to maintain the service quality of existing traffic. The purpose of admission control is to ensure that acceptance of a new data flow into a resource-limited network does not degrade Q0S of admitted flows while optimising the network resource usage.

Admission control has been investigated extensively for wired networks in the past few years and it has been recognised that traditional traffic model-based algorithms (e.g. Markov, or even fractal) cannot accurately predict bursty traffic variations and often require complicated parameter settings. In contrast real-time, measurement-based approaches are able to track network conditions and make admission decisions more effectively. So far a number of measurement-based admission control algorithms have been proposed, e.g. G. Bianchi et al., Throughput analysis of end-to-end measurement- based admission control in IP, IEEE Jnfocom, 2000. However, this is focused on wired networks such as IP or ATM networks, which is not readily applicable to 802.11 Recently a new standard group IEEE 802.1 le has been formed to study the QoS enhancements to 802.11 MAC protocols IEEE 802.lle WG, Wireless medium access control (MAC) and physical layer (PHY) specifications: medium access control (MAC) enhancements for quality of service (QoS), IEEE P802.1 le/Draft 6.0, November 2003.

In 802.11 e the contention access scheme is called enhanced distributed channel access (EDCA). Although the 802.lle standard recognises the importance of admission control in regulating bandwidth resources and providing Q0S guarantees, it does not give any detailed algorithm for the contention- based admission control procedure. Instead, it is left as "admission control in general depends on vendors' implementations".

Therefore there is a need for an implementation for an admission control scheme.

Summary of the Invention

In general terms in one aspect the present invention provides an admission control method for a contention based wireless network which receives requests from new calls or data flows to access the network. The method determines the current bandwidth availability of the network from the current network utilisation and the network bandwidth. In one embodiment, the network utilisation is determined from a network busy time parameter, which can be determined by utilising the carrier sensing function in IEEE8O2. 11 wireless network nodes for example. The call can then be admitted if the current bandwidth availability is greater than the bandwidth requirement of the requesting call.

By only admitting the call if there are sufficient network bandwidth resources available, the Q0S of other already admitted calls on the network is maintained. By monitoring network utilisation no additional control traffic overhead is required. This can be implemented using the existing carrier sensing function for 802.11 networks, making the method easy to implement, and applicable to both infrastructure (centralised) and ad hoc (distributed) network configurations.

In particular there is provided a method of call admission control for a wireless network according to claim 1.

In particular in another aspect there is also provided an apparatus for call admission control for a contention based wireless network according to claim 11.

In general terms in another aspect the present invention provides an admission control method for a contention based wireless network which receives requests from new calls or data flows to access the network. The method monitors the proportion of time that packets are transmitted across the network, in order to determine a network utilisation parameter, such as network busy time for example. It then determines the current bandwidth availability of the network from the current network utilisation and the network bandwidth. The call can then be admitted if the current bandwidth availability is greater than the bandwidth requirement of the requesting call.

Brief description of the Drawings

Embodiments will now be described with reference to the following drawings, by way of example only and without intending to be limiting, in which: Figure 1 is a schematic of a centralised WLAN; Figure 2 illustrates the data transmission protocol on an IEEE 802.11 WLAN; Figure 3 a schematic of a centralised WLAN according to an embodiment; Figure 4 illustrates a charmel busy time parameter; Figure 5 shows a flow diagram illustrating the method of admission control used in the WLAN of figure 3, and Figure 6 shows a flow diagram illustrating the method of admission control used in the WLAN of figure 3 according to another embodiment.

Detailed Description

Referring initially to figure 1, a wireless local area network (WLAN) is shown. The network comprises an access point AP which is in wireless communication with a number of remote stations ST1 - ST3. The access point is also connected to external networks IP and provides a gateway between the stations ST and the external networks IP. The access point AP and remote stations ST communicate via a shared wireless channel W. As the channel W is shared, the different wireless nodes (AP and ST) must contend for access.

In IEEE 802.11 WLANs, a carrier sensing function is used by a node when it wants to access the channel W to check if this is free. If the channel W is in use, a random "back- off' time is allocated before the node can check the channel again. If the channel W is free, the node transmits a reservation packet which includes a duration field corresponding to its network allocation vector (NAY), which is the estimated time the sending node requires to complete its transmission. At the end of this time (NAy), other nodes can again sense the channel W and contend for its use.

The (simplified) data, stream or flow transmission protocol for an IEEE 802.11 WLAN is illustrated in figure 2. For example the access point AP may receive a flow or call from the external network IP intended for remote station ST2. The access point determines from the data flow a duration for its transmission based on the size of the data flow and the transmission rate within the WLAN as is known. When any NAY value in the access point has exhausted, the access point AP then senses the network channel W to determine if it is free. If not, the access point waits a random back-off time before trying again. When the carrier channel W is free, the access point transmits a reservation packet which includes a duration field, which other nodes use to set their NAV counters so that they will not attempt to access the channel W during transmission by the access point AP. The data is then transmitted to the intended remote station ST2

using an address field in the packets as is known.

It may be that there are multiple data flows or calls co-existing on the channel W, with the flows being characterised by bursts of data packets. For example ST1 may be supporting a voice call, ST2 instant messaging, and ST3 internet browsing. Because each flow is not using the channel W continuously, the channel W can accommodate a number of such flows utilising the above described contention mechanism for accessing the channel to transmit a burst of data packets. However as the number or size of these flows increases, the network will become congested and there will be greater delay in sending and receiving data packets on the various applications these flows are supporting. Some flows will be more sensitive to this, for example for real-time voice and video calls, and these delays may be noticeable and annoying for the users.

Therefore the number and size of the data flows admitted on the WLAN needs to be restricted in some way in order to maintain acceptable performance or quality of service (Q0S) for the admitted flow.

Figure 3 shows a WLAN similar to that of figure 1, but additionally including an admission controller or block AB associated with the access point AP. The QoS enabled access point (QAPAB+AP) periodically monitors the channel W and measures the network's utilisation (u). Based on this information, the available bandwidth can be determined. When a new flow requests for admission, the QoS enabled access point QAP compares the available bandwidth and the bandwidth requirement of the new flow and makes the admission decision.

Given a network utilisation (u) and a maximum network bandwidth (Bm), the available bandwidth is estimated by Bavaji = (1 - u) * Bm.

To measure the bandwidth utilisation u, a metric called the channel busy time is used, although other metrics such as the number of collisions could alternatively be used.

Referring to figure 4, during a fixed time interval T (e.g. the beacon interval in 802.11), the amount of time the channel is busy (Tb) is defined to be the total time within T that nodes are transmitting packets and receiving packets within the carrier sensing range.

packet transmissions. This measurement does not incur much extra overhead (and doesn't require any use of the channel resource W), since in 802.11 a station has to sense the channel and check the network allocation vector (NAY) to see if the medium is idle for transmission. The Q0S enabled access point Q0S determines the channel busy time Tb by simply sensing the channel every T/n seconds, where n is the number of times it measures channel busyness during time T, ie the total number of measurement slots during T. Hence the network utilisation u is given by u=Tb/T.

Various other ways of measuring the network utilisation u have been used in wired networks, however these are not suitable for wireless networks given their comparative extreme resource constraints. For example, buffer (queue) length or probing packets have been used in wired networks. However it has been shown in previous studies that queue length or buffer occupancy alone is often not an accurate measure of medium utilisation. Congestion or hot spot conditions can be created without any buffer occupancy. On the other hand, round trip time (RTT)-based probing techniques are quite popular in measurement studies for wired networks. The disadvantage of this method is that probing packets need to be sent into the network hence consuming extra wireless bandwidth.

Another parameter used in the available bandwidth calculation is the maximum useable network bandwidth Bm. This is a system parameter and can be tuned to achieve an optimal effect. There are a number of ways to determine Bm. Both theoretical analyses and previous experiments have shown that the usable bandwidth (or throughput) in a wireless network is often far less than the nominal channel capacity or maximum theoretical network bandwidth. For example, it can be easily derived that the theoretical maximum throughput (TMT) of the IEEE 802.11 MAC is given by: TMT(x) = 8x / (ax+b) Mbps, where x is the MSDU (MAC service data unit) size in bytes, a and b are parameters specific to different MAC schemes and spread spectrum technologies as described for example in: J. Jun et a!., "Theoretical Maximum Throughput of IEEE 802.11 and Its Applications", IEEE International Symposium on Network Computing and Applications, 2003. . For example, when the channel data rate of an OFDM 802.lla MAC is 54 Mbps, and the payload size is 1000 bytes, the theoretical maximum throughput is 25.97 Mbps. A simple policy would be choosing TMT for a typical packet size (e.g. 1000 bytes) as Bm. However, this TMT is defined under a number of idealized assumptions such as zero bit error rate, no losses due to collisions, etc. Therefore to be on the safe side and keep the channel use in an un-congested state, it is preferred to reserve a small portion (eg 20%) of the bandwidth. In this case, 80% of TMT is chosen as the maximum useable network or channel bandwidth Bm.

Other methods of choosing Bm are also possible. For example, analytical models have been proposed for 802.11 (and 802.11 e) contention-based channel access under saturation conditions in U. Bianchi, Performance analysis of the IEEE 802.11 distributed coordination function, IEEE JSAC, March 2000. Analytical formulas are used to determine the maximum achievable saturation throughput. This can also be used as the basis for computing Bm, for example by setting Bm as 80% of this figure.

Bm can also be obtained by simulation or measurement. One method based on MAC layer measurement uses the following formula: Bm S / (Tr - T) = S / D, where S is the packet size, T is the time the packet is ready at the MAC layer and T,. is the time the ACK is received. The time interval D = T T is the MAC delay of packet transmission at a station. Here Bm is measured using only successful link layer transmissions of an ongoing data flow. The measured maximum bandwidth can be normalized to be independent of packet size.

As is known, the maximum bandwidth parameter depends on admission policy and should be selected based on channel conditions, traffic profiles and network scenarios.

Selecting a large available bandwidth Bm means the policy is optimistic which allows for more connections but could result in Q0S violations when traffic is heavy. A small Bm is a conservative policy which should limit QoS violations but may lead to under- utilisation. As a rule of thumb, one can first select Bm using the analytical-model-based methods since they are simple and very often conservative. Then it can be fine-tuned using simulations or measurements to obtain an optimal value.

It is also possible to make Bm be adaptive to channel conditions and traffic load using metrics such as the collision ratio. In this case, the average collision ratio is calculated every update period 1: C = number of collisions occurred / total number of packets transmitted in the same period.

Exponential weighted average can be used to smooth out short-term fluctuations. In one embodiment there are two values of Bm, B 1 and B2, where B 1 > B2. Switching of the maximum bandwidth is dependent on two key parameters: collision ratio threshold Cth, and N, the number of times that the measured collision ratio exceeds a predetermined threshold Cth consecutively. Specifically, at the beginning the admission threshold is Bi. When the collision measurements exceed a predetermined threshold (i.e. , Cth) for more than N times consecutively, the channel is considered overloaded and the maximum bandwidth parameter in admission control is changed to a smaller one B2.

Hence when the wireless medium is busier the admission control policy becomes more conservative, i.e. accept fewer connections.

When a new flow with bandwidth requirement Brcq requests for admission, it sends a flow request message to the QAP which includes its traffic information (flow ID, access category, bandwidth, etc.). The QAP obtains this flow information and admits the flow when the available bandwidth is greater than the required bandwidth, i.e. Bavaii > Breq.

Then the QAP sends back a flow response message with its admission decision to the requesting station. In this way the channel will not get congested and the Q0S of admitted traffic is protected.

This admission control procedure applies to traffic flows requiring a minimum Q0S, and need not apply to best-effort flows such as email for example.

Thus the embodiments provide simple measurement-based admission control schemes for 802.11 WLAN, which allow high throughout and low delays for Q0S traffic by limiting the traffic load in the network. It does not incur excessive complexity as it makes use of the built-in carrier sensing capability of 802.11. It monitors the channel passively and hence avoids wasting scarce wireless bandwidth by sending control messages. It is therefore well suited to real-time applications (e.g. voice, video streaming) which have stringent QoS requirements. This admission control scheme could therefore be readily implemented for QoS enhancement in 802. 11 wireless LANs, such as recent IEEE 802.11 standards.

Referring to figure 5, a flow chart illustrating the method of an embodiment is shown.

The admittance or access controller or block AB monitors or senses the channel periodically to determine a busy time parameter as described above. Over a predetermined period T, the channel is sensed using the access point's sensing mechanism every T/n seconds to provide n readings. Whether or not the channel is busy is measured by a counter which at the end of the period T provides the channel busy time parameter as a percentage, eg 50%.

The theoretical maximum channel bandwidth TMT is determined by one of the methods described above. This may adapt to channel conditions. A safety margin is incorporated (eg 20%) and the maximum useable bandwidth Bm provided. The available channel bandwidth Bay is then determined from the channel busy time and the maximum useable bandwidth (eg Bay = 50%Bm) Data flow requests to the access point AP are routed to the admittance block AB for processing. These requests may be from an external network IP, or the stations forming part of the network itself. Admission block AB determines whether the bandwidth requested by the new data flow Breq is less than the available network bandwidth Bay. If so, the new call is admitted, and the process is the same as that described with respect to figure 2. If the requested bandwidth Breq is too large, a decline message is sent to the application requesting the data flow.

For example, in a 2 Mbps 802.11 network, the maximum bandwidth estimation methods described above all yield a similar result: 1.2 Mbps is a good choice for Bm. When a new constant bit rate (CBR) flow request with bandwidth requirement of 200 kbps arrives, the QAP measures the network utilisation using the busy time metric - u 0.8.

Then Bavaji = (1 - u) * Bm = 0.24 Mbps, which is greater than Breq of 200 kbps.

Therefore the new flow is accepted.

In 802.11 e traffic flows are classified into four levels of access categories (ACs). A smaller contention window (CW) is assigned to the AC with higher priority so that high priority ACs can transmit before the low priority ones in a statistical sense. This can be achieved by setting different transmission and medium access parameters for the different access categories as is known. For example lower access category traffic will have longer back off times so that the higher access category traffic can successfully contend for the channel first. Higher access category traffic may also require a higher network available bandwidth in order to ensure high QoS.

In a further embodiment a differentiated admission policy is employed to accommodate different AC's. This is illustrated in figure 6 which for simplicity shows only two access categories, a higher priority ACI and a lower priority AC2. However it will be understood by the skilled person that this could be readily expanded to incorporate additional AC's, for example the 4 AC's of the 802.lle standard. For a new flow intended as a high priority AC, if Breq > Bavaii and Breq < p * Bavaii (where p is a system parameter, say, 1.1), the flow can still be accepted, but it has to be downgraded to a lower priority traffic class. This approach benefits high priority ACs while optimising the network performance.

Referring to figure 6 in more detail, an incoming traffic flow is handled by the method on the right if it has a high access category (AC1), and the method on the left if it has a low access category (AC2). For low priority traffic (AC2), the admission block AB determines whether the requested bandwidth (Breq) is less than the available bandwidth Bav(AC2) for low access category (AC2) traffic, and if so, proceeds as described above with respect to figure 5 to sense the channel for access. If the channel W is in use, the method backs off for a random wait time, then tries again. Whilst the wait time is randomised with respect to other low access category traffic, it will be longer than back off wait times for high priority traffic flows (AC 1). As with figure 5, if the requested bandwidth Breq is greater than the available bandwidth, then a decline message is sent to the requesting entity attempting to forward the traffic flow.

For high priority traffic (AC 1), the admission block AB determines if the requested bandwidth (Breq) is less than the available bandwidth Bav(AC1) for high access category (AC 1) traffic, and if so, proceeds as described above with respect to figure 5 to sense the channel for access. If the channel W is in use, the method backs off for a random wait time, then tries again. Whilst the wait time is randomised with respect to other high access category traffic, it will be less than back off wait times for low priority traffic flows (AC 1). If the requested bandwidth Breq is greater than the available bandwidth Bav(AC1), then the system determines whether the requested bandwidth Breq is less than a predetermined multiple of the available bandwidth P.Bav(ACI). If this is the case, the traffic flow has its access category downgraded (from AC 1 to AC2), and the left hand method takes over as if a low priority traffic flow had requested admission. A notification is sent to the requesting entity that its traffic flow has been accepted at a lower access category, in order to enable this entity to cancel the traffic flow if this is not acceptable. If the requested bandwidth Breq is still greater than a predetermined multiple of the available bandwidth P.Bav(ACI), then a decline message is sent to the requesting entity attempting to forward the traffic flow.

Whilst the embodiments have been described with respect to an infrastructure-mode wireless LAN where the admission control unit is implemented at the AP, in fact the admission control algorithm can also be carried out in a distributed fashion in an ad hoc network. In that case, each node may need to monitor the channel, compute the network utilisation u and make admission control decisions locally. Alternatively, a central node can be selected via a voting protocol to act as an admission controller.

Also whilst the embodiments have been described with respect to IEEE 802. 11...

WLAN's, they are equally suitable for other wireless networks.

The skilled person will recognise that the above-described apparatus and methods may be embodied as processor control code, for example on a carrier medium such as a disk, CD- or DVD-ROM, programmed memory such as read only memory (Firmware), or on a data carrier such as an optical or electrical signal carrier. For many applications embodiments of the invention will be implemented on a DSP (Digital Signal Processor), ASIC (Application Specific Integrated Circuit) or FPGA (Field Programmable Gate Array). Thus the code may comprise conventional programme code or microcode or, for example code for setting up or controlling an ASIC or FPGA. The code may also comprise code for dynamically configuring re-configurable apparatus such as re- programmable logic gate arrays. Similarly the code may comprise code for a hardware description language such as Verilog TM or VHDL (Very high speed integrated circuit Hardware Description Language). As the skilled person will appreciate, the code may be distributed between a plurality of coupled components in communication with one another. Where appropriate, the embodiments may also be implemented using code running on a field-(re)programmable analogue array or similar device in order to configure analogue hardware.

The skilled person will also appreciate that the various embodiments and specific features described with respect to them could be freely combined with the other embodiments or their specifically described features in general accordance with the above teaching. The skilled person will also recognise that various alterations and modifications can be made to specific examples described without departing from the scope of the appended claims.

Claims (19)

1. An admission control method for a contention based wireless network having a number of nodes and a carrier channel for wirelessly coupling said nodes; the method comprising: receiving a request to access the network from a new call including its bandwidth requirement; determining the current bandwidth availability of the network from the current network utilisation and the network bandwidth; admitting the call if the current bandwidth availability is greater than the bandwidth requirement of the requesting call; wherein the network utilisation is determined from the proportion of time said nodes are transmitting packets across the channel.

2. A method according to claim I wherein the network utilisation is determined from a network busy time parameter.

3. A method according to any one preceding claim wherein the network utilisation is determined by periodically monitoring the channel in order to estimate the bandwidth usage of network traffic over a predetermined period.

4. A method according to any one preceding claim wherein the network bandwidth is less than the maximum possible network bandwidth.

5. A method according to any one preceding claim wherein the network is an IEEE8O2. 11 centralised or ad hoc network.

6. A method according to claim 5 wherein the network utilisation determination comprises carrier sensing.

7. A method according to any one preceding claim wherein the network bandwidth is adaptable to network conditions.

8. A method according to claim 7 wherein the network bandwidth is dependent on the packet collision rate.

9. A method according to any one preceding claim, the network further accommodating multiple access categories of data flow each having its own available bandwidth, and wherein a high priority access category flow is admitted as a lower priority access category flow if the current bandwidth availability for the higher priority access category flow is less than the bandwidth requirement of the requesting call and if the current bandwidth availability for the lower priority access category flow is greater than the bandwidth requirement of the requesting call.

10. A carrier medium carrying processor code which when implemented on a processor will carry out the method according to any one preceding claim.

11. An admission control apparatus for a contention based wireless network having a number of nodes and a carrier channel for wirelessly coupling said nodes; the apparatus comprising: means for receiving a request to access the network from a new call including its bandwidth requirement; means for determining the current bandwidth availability of the network from the current network utilisation and the network bandwidth; means for admitting the call if the current bandwidth availability is greater than the bandwidth requirement of the requesting call; means for determining the network utilisation from the proportion of time said nodes are transmitting packets across the channel.

12. An apparatus according to claim 11 wherein the network utilisation is determined from a network busy time parameter.

13. An apparatus according to claim 11 or 12 wherein the network utilisation determining means comprises means for periodically monitoring the channel in order to estimate the bandwidth usage of network traffic over a predetermined period.

14. An apparatus according to any one of claims 11 to 13 wherein the network bandwidth is less than the maximum possible network bandwidth.

15. An apparatus according to any one of claims 11 to 14 wherein the network is an IEEE8O2. 11 centralised or ad hoc network.

16. An apparatus according to claim 15 wherein the network utilisation determination means is implemented using carrier sensing.

17. An apparatus according to any one of claims 11 to 16 further comprising means for adapting the network bandwidth dependent in network conditions.

18. An apparatus according to claim 17 wherein the network bandwidth is dependent on the packet collision rate.

19. An apparatus according to any one of claims 11 to 18, the network further accommodating multiple access categories of data flow each having its own available bandwidth, and wherein the apparatus further comprises means for admitting a high priority access category flow as a lower priority access category flow if the current bandwidth availability for the higher priority access category flow is less than the bandwidth requirement of the requesting call and if the current bandwidth availability for the lower priority access category flow is greater than the bandwidth requirement of the requesting call.