| Public wireless local area network (WLAN) system based on IEEE802.11 is exponentially becoming popular, because it can provide high-speedInternet connectivity up to 54Mbps or even more. At the same time, most ofthe applications concentrate on data services, but the performance ofmultimedia services are unsatisfying which are normal applications on wiredLAN such as VoIP,audio frequency,video frequency etc. One import factoris the handoff delay. Because WLAN is composed of small cells and thehandoff is due to happen between adjacent cells when users are moving. Thusthe communication will be interrupted and the transferring of packets will bedelayed or even lost. So the QoS of multimedia applications in the mobilehost will be impacted significantly. This paper is aiming at reducing thehandoff delay, especially the probe delay which is dominant in the wholedelay. By doing so, we can alleviate the impacts resulted from handoff andimprove the QoS of realtime applications over WLAN.Firstly, the paper analyzed the handoff procedure in WLAN. Theprocedure can be divided into four phases as follows:1. The trigger of handoff, which means the conditions under which handoffswill happen. There are certain criteria to be used to judge whether thehandoff is needed, such as the signal strength judgment. When the signalstrength is less than some threshold (for example, -100dBm), the handoffwill be taken.2. The scanning phase. It is used by mobile host to find available AP toassociate with. Two kinds of scannings are defined by IEEE 802.11, theactive scanning and the passive scanning. If the passive scanning is taken,MH has to wait and listen on every channel in turn to sense the beaconframe sent by APs and records the results it received. In the condition of acrive scanning, MH sends probe request frames to urge APs to sendresponse frame as soon as possible. So the active scanning can reducehandoff delay compared with the passive scanning. When all channels arescanned, the MH can choose the suitable AP to associate with according tothe signal strength or link quality.3. The authentication phase. After finding the AP suitable to associate with,the MH must experience the authentication procedure of the target AP. TheMH must pass the authentication before it is allowed to associate with thetarget AP.4.The association and reassociation phase. After the MH passes theauthentication of the candidate AP, it needs to associate or reassociate withthe AP in order to get necessary resource allocated by the AP. Theprocedure is triggered by the association or reassociation request frames sentby the MH.Handoff delay lies in the last three phases and the scanning delayaccounts for more than 90% of the overall delay. In order to reduce handoffdelay, studies focus on reducing the three delays mentioned above. Becausethe scanning delay is dominant in handoff delay, it is effective for reducinghandoff delay to study to reduce the scanning delay.Secondly, the paper reviewed existing fast handoff schemes which havebeen proposed in the literature and analyze their advantages anddisadvantages respectively. Most schemes are trying to reduce theauthentication delay and reassociation delay and there are few schemesaiming at the scanning delay.At last, in the view of the dominant position of the scanning delay in theoverall delay, the paper proposed a prediction-based channel scanningscheme to reduce scanning delay. The basic idea of the scheme is that themost suitable target AP should not be judged only by the signal strength and only the AP which is consistent with the MH's mobility pattern is the bestone. So, at first the scheme put forward a handoff prediction algorithm basedon WDAG and then proposed a simple but effective scanning algorithm.The concept of WDAG has been mentioned in other papers. The weightof every edge represents the probability that the MH handoff from the startpoint of the edge to the end. So by maintaining the WDAG it is feasible topredict the mobility pattern of the MH. But the scalability of the existingweight maintaining methods are poor and not practical. To solve the problem,this paper put forward a new weights updating method, it takes the WDAG asan OutStar network in neural networks and then updates the weightsaccording to Hebb rules iteratively. This method is scalable and can maintainthe weights effectively.The fundamental idea of the scanning algorithm proposed by the paperis that, facilitates the mobility pattern of mobile hosts to improve the channelscanning procedure by organizing optimized channel scanning order so as tofind the target AP as soon as possible to reduce the channel scanning delay.The paper sorted the channels by the prediction information input asparameters firstly and then scanned the sorted channel vector. Thus when theprediction is accurate, the target AP can be found only by scanning fewchannels. So it is not necessary to scan the left channels and the channelscanning delay can be shortened dramatically.The scanning algorithm proposed above provides a clear interface forprediction information. So, the prediction information, which can be used bythe scanning algorithm, is not limited to the HWDNG put forward by thepaper. Thus, if the system can provide prediction information as the interfacedescribes, it can use the scanning algorithm to reduce the scanning delay.In order to evaluate the performance of the scanning algorithm, thepaper provides prediction accuracy metric. And then based on the metric itanalyzes the theoretical performance of the scanning algorithm and then uses NS-2 to evaluate the algorithm by simulation. It is proved that the scheme putforward by the paper can reduce the channel scanning delay effectively ifcertain prediction accuracy can be guaranteed. The more accurate theprediction, the more effective the algorithm is. And its performance is sameas the active scanning at worst.So, the channel scanning algorithm can facilitate the improvements inhandoff prediction techniques to improve its performance and then to reducethe handoff delay. It is valuable to improve the QoS inWLAN. |
PDF Full Text Request