P O Z N A N U N I V E R S I T Y O F T E C H N O L O G Y A C A D E M I C J O U R N A L S
No 52 Electrical Engineering 2007
__________________________________________
Dariusz KOŚCIELNIK*
Jacek STĘPIEŃ*
INFLUENCE OF THE HIDDEN STATIONS FOR THE QoS
OF ISOCHRONOUS DATA STREAM IN IEEE 802.11e WLAN
The new IEEE 802.11 standard version, denoted by 802.11e, was developed in order to introduce mechanisms enabling handling of isochronous data stream in the network. The proper work of the EDCF protocol is influenced by the presence of hidden stations in the network. These nodes may produce a significant increase of the number of collisions.The presented results of simulation tests make possible to determine the influence of the presence of a hidden station for the performances of the entire system and the efficiency of the EDCF mechanism. In the summary we have presented the most important conclusions from the experiments that has been performed.
Keywords: wireless network, EDCF mechanism, hidden station.
1. INTRODUCTION
The range of transmission of a WLAN network terminal not always includes all the other nodes of the same basic service set (BSS), placed in a certain basic ser-vice area (BSA). However, the out-of-range nodes may significantly disturb the transmission of data to the neighboring terminals. The simplest example of such a situation is presented in Fig. 1.
Fig. 1. Example of a system with a hidden station; a) network configuration, b) transmission diagram
2007
Poznańskie Warsztaty Telekomunikacyjne Poznań 6 - 7 grudnia 2007 POZNAN UNIVERSITY OF TECHNOLOGY ACADEMIC JOURNALS
The QSTA1 station can send data to the QSTA2 terminal. The transmission is not “heard” by the QSTA3 terminal – for the QSTA1 node it is a hidden station. But the QSTA2 node stays inside the range of QSTA3. If the QSTA1 terminal sends data to QSTA2, the hidden station doesn’t know anything about the transac-tion in progress. After receiving a transmission request, this node shall check the channel availability and start the transmission. Independently of the physical local-ization of the receiver of this transmission, the QSTA2 terminal will be included in the transmission range. In the area occupied by QSTA2, a collision of both packets will take place. At least one (the first) of the described transactions will fail.
The collisions caused by hidden terminals present a much more important risk for the system that those produced by rivalry concerning access to radio channels. An important time of transmission of a single packet significantly increases the probability of generating a new request in the hidden terminal [9]. Its immediate realization will disturb the transaction of the remote terminal.
The IEEE 802.11 standard has been equipped with a mechanism that partially eliminates the consequences of the presence of a hidden station. The mechanism reserves a radio channel around the target node, before the data transmission. Two short frames: RTS (request to send) and CTS (clear to send) are used for reserving the channel. The RTS packet is sent by the data sender and informs the receiver of the starting transaction and its duration. The receiver confirms its acceptance of the transmission, sending a CTS frame (Fig. 2). The frames informs simultaneously all the receiver’s neighbors, that the channel shall be busy during a defined time. Dur-ing the reserved time, any other terminal (ie. also a hidden station) can not start its own transaction, even if the receiver informs it, that the channel is empty.
Fig. 2. Process of channel reservation around the transaction receiver
2. INFLUENCE OF A HIDDEN STATION FOR THE EFFICIENCY
OF THE EDCF MECHANISM
The results of studies concerning the efficiency of the EDCF mechanism, ob-tained by the authors of this work has been published in: [7]. The obob-tained results confirm the effectiveness of the described solution as it concerns the differentiation of the quality of services related to different data classes. The performed simulation tests has also revealed some disadvantages of the EDCF mechanism, resulting from its non-deterministic character.
The main purpose of the experiments described in the following parts of this paper is to determine the influence of the presence of a hidden terminal for the functioning of the EDCF mechanism. The most important aspects of the realized studies concern the verification of the method of service of real time applications by the network as well as the guarantee of a permanent width of available transmis-sion band for the isochronous traffic streams. The most interesting cases are: the situation, in which the system is overload by requests with limit intensity value as well as the work of the network in a high-traffic state.
The NetSim program was used in order to perform the simulations. NetSim software is written in C++. The original purpose of the NetSim was to execute tests of multimedia networks: FDDI II (fiber distributed data interface) and DQDB (dis-tributed queue dual bus), in which it was necessary to reproduce, with high fidelity, complex transmission protocol mechanisms. From the very beginning, NetSim de-velopers wanted to make it as modular as possible. With such an approach, it was later achievable to use it to simulate operation of other local and metropolitan net-works. Each time, only those software blocks had to be replaced that described be-haviour of nodes or other active system elements. Other modules, used e.g. to in-troduce simulation parameters, generate request streams, handle the events queue, test achievement of the stable condition, as well as collect, process and visualize the results, constituted the invariable core of the simulator.
NetSim makes use of the event planning technique (commonly referred to as the events queue). Its mechanisms ensure that mutual time relationships between mul-tiple concurrent processes can be correctly simulated. Both the scale of the simu-lated time lapse and the number of stages which different operations of the exam-ined protocol should be split to can be dynamically adjusted to: character of ob-served activities, temporary offered traffic intensity, size of modeled system and requested accuracy of received results (unfortunately affecting the duration of simulation).
2.1. Model of examined system
Four types of request streams were used in the tests, The first one contained asynchronous requests, corresponding to typical computer data transfer (e.g. FTP). Frames transporting this type of information were assigned the lowest priority (class 0). Request intensity, depending on the character of tests, could have infi-nitely large value (restricted by available system throughput) or predefined maxi-mum value.
Streams of the other three types were generated by isochronous data. Each of them was assigned a different level of priority: 1, 2, or 3, respectively. Data por-tions, generated in equal intervals (resulting from predefined throughput of differ-ent streams), were placed in packets of the same length. All the results presdiffer-ented further were obtained given the assumption that each of Data frames transports a 1024-octet data field.
For streams of different priorities, respective parameter sets were defined to be used by the transmission channel access procedure. The most important of them, determining the contention process and effects, are presented in Tab. 1.
Tab. 1. Contention window parameters for different priority levels
Type of data AIFS CWmin CWmax
asynchronous (class 0) DIFS+6TSL 63 1023
isochronous (class 1) DIFS+4TSL 31 127
isochronous (class 2) DIFS+2TSL 15 63
isochronous (class 3) DIFS 7 15
The other network operation parameters are consistent with the existing stan-dards (Tab. 2). During the simulations, the number of utilized priority levels was reduced to four. This operation was supposed to facilitate data analysis, and did not affect by any means the accuracy or generality of the results.
Tab. 2. Transmission system parameters
Parameter Value
Transmission rate 11 Mb/s
Time-slot – TSL 20 µs
Short inter-frame space – SIFS 10 µs Number of priority levels 4
Data field length – DFL 1024 octets
2.2. Fair distribution of system resources
The first ones of the presented results make possible to evaluate the influence of the presence of hidden stations for the fair distribution of the available bandwidth between the system nodes. The tests has been performed with a network consisting of six terminals. The first of these terminals – QSTA0 - occupies the central place in a ring created by the remaining nodes (Fig. 3). QSTA0 is the only node not be-ing itself a hidden station and for which no hidden stations exist. Every other node is a hidden station for two remote terminals, that perform the same functions to-wards it (Fig. 3). The transmission in the tested network is realized only between the central node and its neighbors. All the packets has the same lowest priority.
The results of simulation tests are presented in Fig. 4. The family of features concerning the intensity of traffic for every terminal with reference to the offered traffic (Fig. 4.a) shows an important disproportion existing in the case of system overload, between the band used by the central nodes and its neighbors. The QSTA0 terminal, receiving all the RTS packets, avoids all the certain collisions (appearing after occupation of the channel by another node). The remaining sta-tions doesn’t „hear” certain RTS frames (starting new transacsta-tions) and - trying to start transmissions in progress – they decrease not only their own efficiency, but also the efficiency of the remote terminals. Although the length of RTS packets isn’t large, the time of waiting for the CTS confirmation as well as the time of competition for the channel access that is not used may cause a difference in the utilization of the common band, amounting even to 30%.
Fig. 4. System loaded by streams of requests with the same priority; a) intensity of serviced traffic, b) average delay of transmission
The average delay values (Fig. 4.b) concerning the transmitted packets (the de-scribed results doesn’t include frames not entered to overloaded transmitters’ queues) are increasing for the external terminals significantly earlier and faster than for the central node. This situation causes a potential risk for the functioning of the EDCF mechanism, because the order of servicing different requests may be de-pendent not only of their priorities.
2.3. The influence of the number of hidden stations for the node efficiency
To study in a more detailed way the interrelation between the bandwidth avail-able to the system node and the number of hidden stations connected with it, a ma-trix structure system (Fig. 5) has been tested. The distances between successive node lines and columns are chosen in such a way, that all the system nodes are lo-cated in the range of the central terminal. The terminals placed in the middle of system borders are exchanging data with their five closest neighbors (everyone having 3 hidden stations). The corner nodes can exchange data only with three neighbors (everyone having 5 hidden stations).
Fig. 5. Matrix structure system
The families of features describing the performance of the tested system are presented in Fig. 6. The first of them indicates the dependence of the effective bandwidth available for every terminal (Fig. 6.a). The obtained results show a sig-nificant dependence between the bandwidth available for a given system node and the number of hidden stations connected with it. In some non-uniform systems, the effect of hidden stations may dominate the planned data priorities.
Fig. 6. A matrix system, loaded by streams of requests with the same priority; a) intensity of serviced traffic, b) average delay of transmission
The features of the delays of packets waiting for transmission (Fig. 6.b) confirm an important influence not only of the presence, but also of the number of hidden terminals as it concerns the quality of service of transmissions from different ter-minals. The valuation of the importance of such a situation for the functioning of a system of priorities demands a comparison of the above presented results with con-sequences of differentiation of the classes of transmitted data – this is done in the following part of this work.
2.4. The influence of priorities for the efficiency of their service
To study the influence of priorities of different requests for the quality of their service [8], the network model shown in Fig. 3 has been used. Therefore, this time the data transmitters are only external terminals. The central point is the receiver of all the transactions. The work conditions of the transmitting terminals are identical. The transmitting terminals constitute a uniform system. Towards different nodes of this network is directed an offered traffic with identical intensity, but different pri-orities, varying between 0 and 3. Two terminals (QSTA2 and QSTA3) are servic-ing the traffic with the same priority equalservic-ing 1 (Fig. 7). Such a solution makes possible to evaluate the credibility of obtained results, that should be very similar for all the abovementioned nodes.
Fig. 7. System loaded by streams of requests with the different priority; a) intensity of serviced traffic, b) mean delay of transmission
The family of curves presented in Fig. 7.a indicates the effective bandwidth for each terminal of the system. The value of this parameter in the case of an over-loaded system is completely different for terminals servicing traffic with 0 and 1 priority and those generating data streams with priorities 2 and 3. Between priori-ties 1 and 2 exists a limit (border) value, for which the effects of the hidden station is comparable to this of the EDCF protocol. The modification of the priority makes that the decisive factor for the quality of service belongs to one of the abovemen-tioned mechanisms. Unfortunately, the subsequent tests indicate that the modifica-tion of the number of hidden stamodifica-tions displaces the limit (border) value. Comparing the charts presented in Fig. 7.a and. 6.a we can also notice, that the difference of the quality of service of requests, caused by changing their priority from 0 to 1 (2 to 3) may result less important, than the effect of differentiation of the number of hidden terminals.
The diagrams showing the delays for packets waiting for transmission confirm an important influence of the presence of hidden stations for the efficiency of the system. A modification of their number, for instance produced by the movement of
the nodes themselves or of the obstacles existing between them can disturb a proper service of different traffic categories.
2.5. Quality of service of isochronous data with low priority
A very important aspect of the quality of service of a given network is the level of guaranteeing the same available bandwidth for all the isochronous data streams. The performed tests define the interrelation of this parameter according to the level of loading of the systems by requests with higher priority.
The tests performed in a system with structure presented in Fig. 3 suppose that only external terminals are active. Four of these nodes are loaded with an isochro-nous traffic with priority 1 and a constant intensity. Its total bandwidth is signifi-cantly lower that the bandwidth of the transmission channel used. A stream with priority equaling to 3 and an intensity that is progressively increased in successive tests is directed towards the fifth terminal. The family of obtained features is pre-sented in Fig. 8.
Fig. 8. The influence of traffic with higher priority for the service of streams of requests with lower priority; a) intensity of serviced traffic, b) average delay of transmission
Analyzing the curves presented in Fig. 8.a we can notice that the modification of the intensity of traffic with higher priority does not have any importance for the quality of service of the tested stream of requests, if the total intensity of offered traffic does not exceed the bandwidth of the transmission channel. As it concerns this kind of system traffic, the quality of service related to data with lower priority does not raise any problems.
However, a further increase of the intensity of requests of priority equaling 3 produces a significant decrease of the quality of servicing less important data. The bandwidth available for these packets is progressively decreasing and the effect concerns in lower degree the terminals being direct neighbors of the node no. 5, than those that constitute hidden stations for this node (Fig. 8.a). The appearing dif-ferences are produced by a stronger negative influence of the collisions of RTS frames (the collision window in hidden stations is doubled), that of clearly lost
ri-valries (increasing the probability of success in the successive process of rivalry concerning access to a shared transmission channel). The described effect confirms again the influence of the presence of hidden stations for the fair service of differ-ent nodes of the system.
The restriction of the traffic with lower priority produces an increase of the time of delay for servicing of different packets. The family of features presented in Fig. 8.b indicates a clear influence of the presence of a hidden station generating a higher priority traffic for the quality of servicing of requests originated by different nodes.
2.6. Quality of service of isochronous data with high priority
The last examined results concern an opposite situation that the one presented in the previous paragraph. This time, four terminals of the system, loaded with a traf-fic with permanent intensity (with a total intensity lower than the transmission channel bandwidth), transmit packets with a high priority (equaling 3). Towards the node no. 1 is directed a stream of less important requests (priority 1), which in-tensity is progressively increased in successive tests. The features presented in Fig. 9.a show that the loading of the system with low-priority traffic hasn't any in-fluence for the way of servicing requests with high priority. This conclusion is true also in the situation of overloading, resulting only in a restriction of the stream of less important requests.
Fig. 9. The influence of traffic with lower priority for the service of streams of requests with higher priority; a) intensity of serviced traffic, b) average delay of transmission
It is worth of note that as well the intensity of serviced traffic with high priority (Fig. 9.a), as the average time of waiting for the start of the transmission are abso-lutely independent of the fact if a given node is directly neighboring with the rival node with lower priority or is a hidden station for this node. So, the presence of a hidden station has a more important influence for the work of a given node if the priority of requests serviced by this node is lower.
3. SUMMARY
The performed tests indicate an important influence of the presence of hidden stations for the efficiency of the IEEE 802.11e network. The observed decrease of the quality of service of different nodes is connected with the increase of the num-ber of collisions of RTS packets. The probability of collision is proportional to the number of hidden stations connected to a given node. This relation has been con-firmed by the performed tests, indicating also an important influence of the de-scribed mechanism for the delays concerning packets waiting for transmission.
The presence of hidden stations may be important for the quality of service of individual requests with a same level than the mechanism of data prioritetization, used by the EDCF protocol. Because of this, the band allocation between different applications done by the network administrator may be significantly disturbed. The control of this parameter will be particularly difficult in systems that permit the movement of terminals or a dynamic appearance between terminals of obstacles that make impossible the transmission of data between them. In these systems, the number of hidden stations connected with a given node continuously changes and such a situation influences in a significant way the quality of service of requests directed to the system.
4. REFERENCES
[1] Annese A., Boggia G., Camarda P., Grieco A., Mascolo S., Providing Delay Guaran-tees in IEEE 802.11e Networks, Computer Communications, 2004.
[2] Banchs A., Perez-Costa X., Qiao D., Providing Throughput Guarantees in IEEE 802.11e Wireless LANs, 18th International Teletraffic Congress, Berlin 2003.
[3] Bejerano Y., Bhatia R. S., A Framework for Fairness and QoS Assurance in Current IEEE 802.11 Network with Multiple Access Points, The Conference of Computer Communication, 2004.
[4] Choi S., Prado J., Mongold S., Shankar S., IEEE 802.11 Contention-Based Channel Access (EDCF) Performance Evaluation, IEEE ICC, Anchorage, 2003.
[5] Dajiang H., Schen Ch. Q., Simulation Study of IEEE 802.11e EDCF, National Uni-versity Singapore, IEEE VTC-Spring, Korea 2004.
[6] Dhanakoti N., Gopalan S., Perfectly Periodic Scheduling for Fault Avoidance in IEEE 802.11e in the Context of Home Networks, 14th IEEE International Symposium on Software Reliability Engineering, Denver, 2003.
[7] Kościelnik D., Analysis of IEEE 802.11e Standard in Terms of Real Time Application Requirements, International Conference on Wireless and Mobile Communication, Bu-charest, Romania 2006.
[8] Kościelnik D., Analysis of the Quality of Service of Isochronous Data Streams in IEEE 802.11E WLAN, Poznańskie Warsztaty Telekomunikacyjne, Poznań, 2007, un-published.
[9] Zhu H., Chlamtac I., An Analitic Model for IEEE 802.11e EDCF Differential Ser-vices, ICCCN03, Dallas 2003
