Research Of IEEE 802.11 DCF Protocol Performance And Improvement Of Backoff Algorithm

With the recent development of wireless technology and Internet, therearise increasing demands for portable computers or workstations. AndWLAN (Wireless Local Area Network) is used widely to meet these demandsbecause of its high flexibility and robustness. There are many existingstandards, but IEEE802.11 is most widely used. The media in a wirelessnetwork is a shared and scarce resource, thus one of the key questions is howaccess to this shared media. The problem needs to be solved that the limitedchannel is distributed among the multi-users;MAC (Medium Access Control)layer is exactly used for solving how to distribute the shared channelresources fairly and efficiently. The performance of networks is mainly up tothe access protocol of MAC layer, so the important subject of wirelessnetworks is to select appropriate MAC protocol, dispose channel resourceseffectively according to the business characteristic of the network, improvethe service efficiency of wireless resources and enhance systematic capacityand transmission quality. IEEE802.11 Wireless LAN standard defines a distributed random accessMAC protocol: DCF (Distributed Coordination Function). DCF is one of themost important access mechanisms in MAC layer. DCF adopts the techniqueof Carrier Sense Multiple Access with Collision Avoidance which is based ona kind of competitive random access mechanism. Each station uses CSMAdistributed algorithm to obtain the right of sending through contending withevery station in the channel. For stations compete the channel randomly, it isimportant that how to resolve the problem of collision, therefore, DCFintroduces mechanism of carrier sense, inter-frame space and rules of randomback-off. IEEE 802.11 DCF adopts the binary exponential backoff (BEB)algorithm. The principle of algorithm is that each competing station sets up abackoff timer according to a randomly selected backoff time period. Thisbackoff time period is selected uniformly between 0 and the ContentionWindow (CW). The value of CW is dynamically altered with packet collisiontimes. In the BEB algorithm, the CW is doubled when a station experiences apacket collision. If a station is successful in its packet transmission, its CW isreset to the minimum value. This paper focus on DCF random back-offalgorithm and quotes the Markov chain model to study DCF access schemeand analyze the key performance parameters of DCF.From the model we can draw the conclusion: we can explicitly computethe optimal transmission probability (τ ) that each station should adopt inorder to achieve maximum throughput performance within a considerednetwork scenario, i.e., the number of stations (n) and the length of collisions( Tc? ). However, the two parameters are not directly controllable variables.The model also shows that transmission probability (τ ) depends on thesystem parameters: maximum backoff stage (m) and initial contentionwindow sizes (W). Fortunately these two parameters are controllable. To acertain specific network, the only way to achieve optimal performance is toemploy adaptive techniques to tune the values m and W (and consequentlyτ ).In the real wireless network, the hosts in the network change in real time, sooptimum value of τ also changes in real time, however, DCF's maximumbackoff stage (m) and initial contention window sizes (W) are constant, that isto say τ which is determined by W and m is immobile, so it unable tofollow the optimum value of τ self-adaptively, thus the network performanceis influenced. It is just the defect.According to above-mentioned analysis, this paper has put forward anew improvement scheme, its concrete thought is: A station has the ability toestimate the congested situation of the channel through accumulating thecounter of frozen times before sending the data. The "busy" indicates that atleast one station has transmitted a packet, no matter that the transmission issuccessful. If "busy" is often detected, it shows that the channel is morecongested, otherwise, we can think the channel idler on the contrary. So eachstation can tune the value of W according to the congested degree of channel.In this scheme, no matter how the number of stations (n) changes, theback-off window can always makes corresponding adjustment exactly.Consequently the saturation throughput can be improved a lot.This paper introduces a counter (Coccupied) to accumulate the times of"busy" during the back-off period. The parameter Cbackoff represents theback-off slots which have been chosen at random. And a parameter λ isdefined as Coccupied/(Cbackoff+1) to estimate the congested degree ofchannel. The λ is between 0 and 1. The channel is considered congested whileλ is close to 1, and it expresses that the channel is not congested that its valueapproaches 0. The new CW will be adjusted between CWmin and CW valueaccording to λ after each successful transmission, so it has achieved the goalof adjusting the connection window in self-adaptation.Afterwards, we employed NS to simulate the new back-off algorithm, theemulation experimental result shows: Comparing with the original standard,the performance, such as throughput and drop of packets, etc. has obviouslyimprovement, so the improvement algorithm that this paper puts forwardabout DCF is effectual. In the next work we should pay more attention to theanalysis of multiple scenarios, the selection of parameters and their effect onthe performance.