Distributed Call Admission Control for VoIP over 802.11 WLANs Based on Channel Load Estimation

Distributed Call Admission Control for VoIP over 802.11 WLANs based on Channel Load Estimation

Paolo Dini, Nicola Baldo, Jaume Nin-Guerrero, Josep Mangues-Bafalluy

Sateesh Addepalli, Lillian L. Dai

IP Technologies Area Centre Tecnològic de Telecomunicacions de Catalunya

Av. C. F. Gauss 7 – 08860 Castelldefels, Barcelona, Spain {paolo.dini, nicola.baldo, jaume.nin, josep.mangues}@cttc.es

Cisco Research Centre 425 East Tasman Drive

San Jose , CA 95134 - USA {sateeshk, lildai}@cisco.com

Abstract —Call Admission Control (CAC) is recognized as one of the key strategies to achieve satisfactory QoS support for VoIP over IEEE 802.11 WLANs. However, most of the prior CAC solutions for VoIP over WLAN are centralized, and the few distributed CAC schemes proposed so far do not account for issues such as the loss of channel time due to medium contention and the coexistence of VoIP traffic with background traffic such as TCP data flows. In this paper, we describe a distributed CAC scheme which aims at addressing these issues by leveraging on channel monitoring techniques, thus being readily implementable in today's consumer devices. The proposed scheme is thoroughly evaluated by means of testbed experiments as well as network simulations in different scenarios, and it is shown to yield a better CAC performance with respect to prior solutions. Keywords: VoIP, WLAN, QoS, Distributed Call Admission Control, Access Point Selection

I. INTRODUCTION By offering the possibility of performing voice

communications at bargain prices using whatever access network available, Voice over IP (VoIP) services have reached an enormous popularity in the recent years. One of the most interesting use cases for VoIP is in combination with the IEEE 802.11 technology to provide wireless voice services to mobile devices such as laptops, smart phones and PDAs. The reason for this interest is the fact that over the last decade 802.11-based WLANs have been deployed almost everywhere, and provide wireless Internet connectivity in home and office environments as well as in many public places.

Unfortunately, achieving a satisfactory service quality for VoIP over WLANs is rather challenging due to the random access nature of the CSMA/CA protocol on which 802.11 is based. Even though the most recent version of the 802.11 standard [1] provides some means for traffic differentiation, Quality of Service (QoS) is not guaranteed.1 Therefore, a VoIP user might still experience a poor quality service due to the wireless channel being congested by other VoIP users [2] and/or other types of traffic such as web browsing, video streaming and peer-to-peer file sharing.

Call Admission Control (CAC), which is known as a very successful paradigm for cellular communications, has been recently proposed as a solution to provide satisfactory voice quality in WLANs. Most of the CAC solutions proposed so far for VoIP over WLAN are centralized [3],[4], i.e., the

1 We refer to the QoS support provided by the EDCA only. We note that

the HCCA effectively provides some QoS guarantees, but it is not commonly implemented in commercial devices.

choice of whether to accept a new call or not is taken by the access point (AP). However, a distributed CAC solution in which the decision is performed by the mobile station (STA) is more appealing for next generation networks. The reason is that, in a distributed CAC scheme, the STA is in charge of deciding whether a particular AP can offer a suitable service quality and, consequently, whether to send its traffic through that AP or not. Such a user-centric approach is advantageous for the end-user since the choice of the AP can be made also considering user preferences (e.g., cost) within the CAC algorithm, which runs at the STA. Besides, considering that different WLAN APs often belong to different operators or private citizens, a distributed solution has more chances to be deployed than a common strategy based on agreements among multiple network providers. One might argue that, in such a user-centric approach, a malicious STA could potentially disobey the CAC rules to selfishly enhance its performance. However, it is worth highlighting that in VoIP over WLAN scenarios there is no incentive for a STA to misbehave (e.g., by joining a network which is already at its maximum capacity), since this would lead to poor service quality for the misbehaving STA as well.

A limited number of distributed CAC schemes can be found in the literature. In [5] an algorithm is proposed, which relies on each STA doing an active probing of the wireless link to infer the achievable service quality. While able to make an effective CAC decision, this solution has the disadvantage of increasing the control traffic overhead of the network, potentially harming ongoing data communications by other users. In [6] the authors propose a perceptive admission control for 802.11 ad-hoc networks based on the busy-time ratio metric. A similar approach is investigated in [7], where the authors consider an infrastructure WLAN and introduce a new metric, the Time Between Idle Times (TBIT), which is shown to yield a very effective CAC algorithm. In fact, the use of the TBIT metric is more effective than the scheme in [6], and it can be considered the state of the art in distributed CAC for VoIP over WLAN. Nevertheless, we note that the method for measuring the TBIT metric does not consider the time consumed by erroneous transmissions (e.g., collisions), and due to this reason the measured number and duration of the idle time periods are not correct. Furthermore, in [7] the performance evaluation (i) does not consider the presence of heterogeneous voice codecs, (ii) does not consider the case in which STAs use different PHY rates, and (iii) does not show any numerical result for the case in which background traffic is present. All these conditions are commonly encountered in the vast majority of real WLAN scenarios, and could

978-1-4244-6404-3/10/$26.00 ©2010 IEEE

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE ICC 2010 proceedings

Authorized licensed use limited to: ACADEMIA SINICA COMPUTING CENTRE. Downloaded on July 08,2010 at 06:58:12 UTC from IEEE Xplore. Restrictions apply.

negatively affect the performance of a user-centric CAC scheme.

In this paper we propose a distributed CAC algorithm based on a new method for channel load estimation. This method is founded on the determination of the fraction of time in which the channel is occupied by successful and unsuccessful frame transmissions, respectively. Each STA willing to start a new voice call is expected to calculate this metric for a candidate WLAN AP to be selected, in order to understand whether the AP is able to support its new VoIP session with sufficient service quality, without degrading the quality of pre-existing VoIP sessions. Like the CAC scheme based on the TBIT metric, our proposed approach is distributed and passive (i.e., no probing packet is sent through the radio link), and furthermore it is easily implemented by developing an opportune channel monitoring software (e.g., using the PCAP library [8]). We claim that, with respect to the TBIT scheme, our solution has the advantage of providing a more accurate estimation of the channel load, due to the fact that it considers for this purpose not only the channel time spent in successful transmissions but also the time consumed by collisions and channel errors.

The proposed scheme has been subject to an extensive performance evaluation. Different scenarios have been tested both in a real testbed and by means of simulations. A comparison with the TBIT scheme has been performed for all scenarios. The results presented in this paper show that, thanks to its more accurate evaluation of the channel load, our proposed scheme yields a higher success rate for CAC decisions.

The remainder of the paper is organized as follows. In Section II we introduce the distributed CAC algorithm. Section III is devoted to the method for the estimation of the channel load which is used by the CAC algorithm. In section IV we present our performance evaluation setup and discuss the performance obtained by the proposed scheme in comparison with the TBIT scheme. Finally, in Section V the conclusions are drawn.

II. CAC ALGORITHM DEFINITION The CAC algorithm that we propose in this section is

designed for infrastructure IEEE 802.11 networks in which all STAs are within communication range (no hidden terminals). We also assume that the capture effect can be neglected, which is a reasonable assumption in a scenario with no hidden nodes.

Let us consider a STA monitoring the radio channel for a given time interval. The observed channel can be in two different states: busy, when one or more transmissions are being performed or idle when there are no ongoing transmissions. A busy channel can be occupied by either successful or unsuccessful frame transmissions, depending on whether the frame is correctly received or not. Unsuccessful transmissions can be due to packet collisions among active stations and/or channel errors. In practice, channel errors are smoothed by the use of rate adaptation scheme which instructs the physical layer to use a reliable modulation and coding scheme. On the other hand, the presence of collisions with probabilities on the order of a few tenths is part of the normal operation of a WLAN [10]. The link layer stop-and-wait ARQ procedure adopted by 802.11 is very effective in preventing network-layer packet losses due to collisions and channel errors. However, failed transmission attempts can consume a significant portion of busy channel time. The proposed CAC algorithm is based on

a model which aims at evaluating the channel busyness ratio of a WLAN by considering both successful and unsuccessful transmissions.

In a scenario with heterogeneous traffic (e.g., voice plus background data), let us define: as the ratio of time in which the channel is occupied by successful voice traffic transmissions; as the fraction of time dedicated to successful background traffic transmissions and as the ratio of time in which the channel is occupied by failed transmissions. The equation here below should be verified: 1

(1)

where is the fraction of time in which the channel is occupied by the back-off procedure.

The model described by equation (1) is used for the implementation of a distributed CAC scheme to support the AP selection process by a STA willing to start a new VoIP session. This is done according to the following algorithm:

• the STA monitors the wireless channel in which the AP is operating to determine if an AP is eligible;

• and are calculated based on the information gathered by channel monitoring;

• the forecast channel load expected after the introduction of the new VoIP session is determined as a function of the actual values of and . This calculation is described in Sections III.A and III.B;

• the forecast fraction of channel time consumed by the backoff procedure is determined according to the method described in Section III.C;

• the term is approximated by the pre-determined fixed values . The method we propose to determine these values is discussed in Section IV.B;

• if 1 , then the AP is said to be eligible for the new VoIP session.

The channel load model just proposed has been designed with the aim of supporting CAC for voice traffic. However, we note that it can be extended, with minor modifications, to other types of real-time traffic, such as video conferencing and video streaming.

III. CHANNEL LOAD ESTIMATION METHOD The method is based on the hypothesis that a STA can

monitor the radio link over a time window of duration , and gather the information contained within the MAC and PHY headers of the captured data frames. We note that with this procedure only successfully decoded frames will be observed. We define a frame exchange sequence as the sequence of a voice DATA frame plus its following ACK. Let the index denote the generic observed frame exchange sequence, and let denote its duration, where we account for the sum of the duration of the DATA and ACK frames as determined by the 802.11 standard [1], plus one DIFS (or AIFS if QoS support is being used) and one SIFS. Let be the total number of frame exchange sequences observed in the window.

A. Calculation of and The monitoring station can determine as: ∑

(2)

Its forecast value is then calculated as:



(3)

where is the cumulative number of packets per second of the two new VoIP flows which are to be introduced (from the STA to the correspondent node and vice versa), and is the duration of the frame exchange sequences for these flows. Both and are expected to be known by the STA, e.g., from SIP signaling and PHY rate adaptation information.

B. Calculation of and expresses the portion of time in which the channel is

occupied by collisions and channel errors. Unfortunately, these events cannot be observed directly by a monitoring STA. In fact, when capturing the wireless traffic using a device in monitor mode only successfully decoded frames are reported. Some devices actually allow gathering frames which are detected as erroneous due to the failure of the MAC CRC. However, we note that frames lost due to collisions will still not be collected using this method, since a collision is most likely to corrupt the PHY header of the frame. In theory, counting the number of PHY errors reported by the device could allow accounting for collision events as well, as proposed in [9]. However, in practice, more than one PHY error can be reported for the same frame, and furthermore PHY errors often happen for other causes, like transmissions by non-802.11 devices and spurious synchronization events. As a consequence, it is not practical to use the number of PHY errors for the calculation of . Finally, some analytical models (e.g., [10] and its derivates) provide methods for the estimation of the channel time consumed by erroneous transmissions. However, these methods rely on the hypothesis that every station always has a new packet to transmit, which clearly does not hold for the case of VoIP traffic. As a consequence, these methods are not suitable for the scenarios considered in this paper.

To overcome the above mentioned issues, we hereby propose a method for the estimation of based only on the observation of successful frame exchange sequences. We exploit the fact that the IEEE 802.11 MAC header has a retry flag which is set to 1 for all DATA frame which are being retransmitted. Let and be the number of successfully decoded DATA frames (both VoIP and background traffic) which have the retry flag set to 0 and 1, respectively. We assume that there is no hidden terminal problem, and furthermore that whenever a DATA frame is delivered also the subsequent ACK is successfully delivered (this is reasonable in the absence of hidden terminals). With these assumptions, we have:

(4)

where is the total number of MAC Service Data Units (MSDUs) which have been successfully delivered, i.e., the number of DATA packets at the networking layer who experienced a successful transmission attempt within the retransmission limit ( ). In other words, since only the last transmission attempt (the successful one) for every successfully delivered MSDU is seen, then a generic successfully delivered MSDU will be counted either in or in , depending on whether the first transmission attempt was successful or not. Let us now further assume that transmissions fail mostly due to collisions, and that consequently all frame transmission attempts in the

observation period have the same failure probability . Then, the first transmission attempt for a successfully delivered MSDU will also fail with the same probability . As a consequence, we can estimate the value of in the observation period as:

(5)

Since 802.11 adopts a stop-and-wait ARQ procedure, we can use to calculate the expected value of the number of failed transmission attempts per MSDU, conditioned on the event that the MSDU is delivered within the retransmission limit, as:

Pr failures | MSDU delivered 11

(6)

Then, is calculated according to the following formula:

(7)

where is the average channel busy time due to a failed transmission attempt and is the average number of stations having a contemporary collision. An accurate estimation of is derived in [11] under the assumption that all the STA are in saturation. Unfortunately this assumption, as previously discussed, is not valid in the VoIP scenarios we are considering in this work. Then we approximate this value as:

(8)

where is the measured average duration of all DATA frames which are observed by the monitoring STA.

Similarly, an accurate expression for cannot be derived without the saturation assumption. For this reason, we approximate with a value of 2. This is reasonable, since the collision probability is much lower in a VoIP scenario than in saturation conditions, and therefore the probability of having more than two STAs involved in a collision is negligible.

Finally, the forecast value ρ is calculated as

(9)

where has been defined in the previous sub-section, and is calculated as per equation (6) by substituting with the transmission attempt error probability that is expected after the new VoIP flows are introduced. We calculate as a function of which is determined by fitting experimental data, as will be discussed in Section IV.B.

C. Calculation of This parameter aims at considering the channel time that

the AP spends in back-off, i.e. the idle time necessary for the AP to decrement the backoff counter prior to starting a transmission. We focus on the AP since it is well known that



the downlink is the bottleneck in a VoIP over WLAN scenario [2].

We observe that when the wireless channel is close to the VoIP capacity limit, Immediate Access ([1], section 9.2.5.1) is very unlike to happen. Then, we express the fraction of time that the AP spends in backoff as:

(10)

where is the number of MSDUs transmitted in downlink during the time window , and is the channel time spent in backoff for each MSDU transmitted by the AP, defined as follows:

2 2 (11)

where is the duration of a time slot according to the PHY specification being used.

Finally, the forecast value is calculated as

(12)

IV. TEST ENVIRONMENT SETUP AND PERFORMANCE EVALUATION OF THE PROPOSED CAC ALGORITHM

The CAC algorithm proposed in this paper has been validated through an extensive test campaign, using both the EXTREME testbed [13] and the ns3 simulator [14]. The scenario is composed of one AP and several mobile nodes sending/receiving traffic to/from an external fixed node. The simulator was used to scale our performance evaluation to scenarios involving higher number of nodes than the one supported by the testbed. The parameters of the simulator have been opportunely tuned to match the characteristics of the testbed; in particular, we verified that the results obtained for the experiments that we ran on EXTREME matched those obtained with the simulator for the same scenarios. For a detailed description of this tuning process, the reader is referred to [15].

In the remainder of this section we describe the tested scenarios, we discuss the experimental method used to determine and , and finally we conclude with the evaluation of the performance of the proposed CAC algorithm and its comparison with the TBIT algorithm in [7].

A. Definition of the considered scenarios The following four scenarios are used to evaluate the

CAC algorithm and to experimentally obtain the values of and .

1) Homogeneous traffic scenario This is the type of scenario which is considered in the

most of prior works dealing with VoIP over WLAN [2]-[7]. For this scenario, noise-related channel errors are

minimized by using RF cables in EXTREME and locating wireless nodes close to the AP in the NS3 simulations. For each STA, two voice flows are started, one from the STA to the AP and the other vice versa. In every experiment, all the nodes use the same fixed PHY rate and the same voice codec. We repeated several experiments varying the number of STAs (from 1 to ), the voice codec (G.711, G.723, G.729) and using different PHY rates belonging to different

PHY specifications 2 . Note that always exceeds the maximum number of users that an AP can support with an adequate voice quality. We used the E-Model [16] as method to evaluate the perceived voice quality by the user; then it is assumed that a user is satisfied when 70.

The following table summarizes the settings used in this scenario. For every particular setting, 20 independent repetitions of the same experiment were run to get a satisfactory statistical confidence in the results.

Table 1. Homogeneous scenario settings voice codec PHY specification PHY rate

G.711 DSSS 2.4 GHz 1 Mbps 6G.711 DSSS 2.4 GHz 2 Mbps 8G.711 OFDM 5 GHz 6 Mbps 21G.711 OFDM 5 GHz 12 Mbps 36G.723 DSSS 2.4 GHz 1 Mbps 12G.723 DSSS 2.4 GHz 2 Mbps 19G.723 OFDM 5 GHz 6 Mbps 45G.723 OFDM 5 GHz 12 Mbps 95G.729 DSSS 2.4 GHz 1 Mbps 9G.729 DSSS 2.4 GHz 2 Mbps 12G.729 OFDM 5 GHz 6 Mbps 38G.729 OFDM 5 GHz 12 Mbps 55

2) Heterogeneous codec scenario This scenario is identical to the previous scenario, except

that different codecs are used within the same experiment. In particular, STAs number 1 to use a first type of codec, while STAs number 1 to use a second type of codec. Experiments are repeated for values of ranging from 0 to . All experiments for this scenario were run on the EXTREME testbed using the DSSS PHY only. The following table summarizes the experimented settings with their most relevant parameters.

Table 2. Heterogeneous scenario settings scenario 1st codec 2nd codec PHY rate

A G.711 G.729 1 Mbps 3 7B G.729 G.723 2 Mbps 8 14

3) Multi-rate scenario

The purpose of this scenario is to test the performance of the CAC schemes in a situation in which nodes use different PHY rates to adapt to the differences in signal propagation that they perceive to and from the AP due to their different position. We note that this situation is potentially critical for the CAC schemes, since the presence of data frames with different durations might potentially be detrimental for the estimation of the residual channel capacity.

For this scenario, we ran all experiments using the NS3 simulator with the OFDM PHY specification. In every experiment, STAs are randomly placed in a circle of radius 150m. We used a log distance propagation model with a propagation exponent of 3 and a reference loss of 46 dB at a distance of 1 m. These settings result in the value of the disc radius being slightly less than the maximum practical communication range for the 6 Mbps PHY rate. All STAs use an ideal rate adaptation algorithm which selects the highest PHY rate that provides communications with a Bit Error Rate less than 10 . Several experiments were

2 For the PHY specifications, we used the DSSS specification for the

2.4GHz ISM band (commonly referred to as 802.11b) and the OFDM specifications for the 5 GHz band (commonly referred to as 802.11a). We executed DSSS experiments both with the testbed and with the simulator, whereas 802.11a experiments were run on the simulator only. This was needed to cope with the high value of required to go beyond the VoIP capacity when using 802.11a.



repeated varying between 1 and 80. For every value of , we ran 30 independent replication of the experiment.

4) TCP background traffic scenario This scenario is similar to scenario 1, with the difference

that, in addition to the STAs performing voice communication, further STAs were added which perfomed a TCP file transfer in the downlink direction. For this scenario we used the EDCA access method (commonly referred to as 802.11e) for traffic differentiation. In particular, we used the AC_VO and AC_BE access categories defined by the standard [1] for VoIP and TCP traffic, respectively. The use of EDCA is motivated by the fact that only traffic differentiation can make VoIP perform satisfactorily when TCP flows are present. Experiments for this scenario were run repeating the same settings of the homogeneous scenario and adding 1, 2, 3 , with 15 independent replications for every resulting combination. All experiments for this scenario were run using the NS3 simulator.

B. Determination of and 1) Expected failure probability ( )

As we discussed in Section III.B, the calculation of the metric requires an estimate of the failure probability that is expected after the introduction of the new voice call. This estimation is to be based on the current failure probability determined from (5). To investigate possible ways to realize such estimation, we evaluated experimental data obtained from all the scenarios introduced in the above, which are shown in Figure 1. From the figure, a linear dependency between the current and the estimated failure probability is evident. Based on this consideration, we chose to implement the calculation by using linear regression over the experimental data obtained. As a final note, we point out that the resulting linear model is very close to the identity. This fact is due to the limited impact of the introduction of a new voice flow on the collision probability.

Figure 1. Future failure probability versus actual failure probability

2) Background traffic channel time ratio (

We recall that our CAC algorithm requires the determination of the fraction of channel time which is occupied by background (or non real-time) traffic. In IEEE 802.11 deployments where QoS support is enabled, such as the one we consider in Scenario 4, voice traffic is normally assigned the highest priority in medium contention. Therefore, as voice flows are added, lower priority traffic will eventually shrink its channel usage in favor of voice traffic, up to a minimum level of occupancy. After this

consideration, we define as the minimum fraction of time which is expected that background traffic will occupy as a consequence of its lower medium access priority. As a result from the evaluation of the experimental data obtained from Scenario 4, we determined that a value of 0.1 is representative for scenarios with TCP flows in the downlink direction (the number of TCP flows is not relevant since they all share the same device – i.e., the AP – for medium access).

C. CAC Algorithm Evaluation We ran both our proposed CAC algorithm and the TBIT

scheme in all the scenarios described in Section IV.A. The performed experiments consisted in evaluating the voice quality (measured using the R-factor metric defined within the E-Model [16]) obtained after the addition of a new voice call. As performance metric for the comparison, we introduce a parameter which describes the difference between the voice capacity, expressed as the maximum number of user which can be accepted with a good quality (i.e. R>70) in a given scenario and the maximum number of users accepted by the algorithm being considered. It is expressed as follows: Δ

(13) Δ assumes positive values when , i.e., the

algorithm overestimates the network congestion and admits fewer users than the voice capacity, and it assumes negative values when , i.e., the algorithm underestimates the network congestion and admits more users than the voice capacity. We note that the latter is much worse than the former, , since all the users will experience a bad voice quality when the voice capacity is exceeded.

Table 3 compares the performance of our proposal and of TBIT for all the considered test scenarios. We can appreciate that our solution performs better in every studied case,most notably in the homogeneous scenario with high PHY rates and in presence of TCP background traffic. Two examples of such cases are analysed in the following. They are namely, the homogeneous scenario with the G711 codec at 12 Mbps (Figure 2) and TCP background scenario with the G711 codec at 2 Mbps (Figure 3). Figure 2 and 3 represent the fraction of blocked voice calls averaged over all the repetitions of the same experiment versus the number of voice users. and of our solution are highlighted in the pictures.

In Figure 2, cannot be shown for TBIT, since in our test scenarios it always admits new flows regardless the number of voice users already present. This is due to the fact that TBIT does not consider the channel time spent due to collisions. In the same scenario, our algorithm performs very close to the real case (Δ 0.03), thus estimating properly the congestion and taking the decision accordingly. TBIT, on the contrary, underestimates the congestion, thus causing a quality degradation of all the active voice communication.

Also in the case described by Figure 3, our solution is close to the real case (Δ 0.17), whereas TBIT completely misinterprets the network status. It is well known that, when EDCA is used, TCP occupies the capacity left by voice traffic (i.e. UDP). TBIT recognizes it as congestion and therefore does not admit new voice calls. When most of the traffic in the network is voice, then TBIT underestimates the network congestion (as in Figure 2) and admits a high



Table 3. Measurement campaign results for every tested scenario

Scenario Codec Bitrate Our TBIT

homogeneous

G711

1M 0 1 2M 0 0.14 6M 0.05 -1 12M 0.03 -1

G723

1M 0 0.18 2M 0 0.13 6M 0.03 -1 12M 0.07 -1

G729

1M 0.13 0.13 2M 0 0.09 6M 0.03 -1 12M 0.02 -1

heterogeneous Scenario A 0 0.25 Scenario B 0 0.33

multirate G711 - 0.03 0.03

TCP background

G711

1M 0 1 2M 0.17 -1/1 6M 0.18 -1 12M 0.24 -1

G723

1M -0.29 1 2M 0.08 -1/1 6M 0.15 -1 12M 0.21 -1

G729

1M 0 -1/1 2M 0.13 -1/1 6M 0.18 1 12M 0.24 1

number of voice sessions, again degrading the perceived quality of all the active users.

Figure 2: Fraction of blocked users for G711 codec scenario at 12 Mbps

Figure 3: Fraction of blocked users for G711 codec + 1 TCP connection scenario at 2 Mbps

V. CONCLUSIONS In this paper we have proposed a distributed CAC

scheme which is based on channel monitoring and load estimation. The channel load estimation method accounts for both the fraction of time spent in successful and erroneous frame transmissions. An extensive measurement campaign has been conducted through both testbed experiments and simulations. The performance evaluation has showed that, in the considered scenarios, the proposed scheme is more robust and accurate in making CAC decisions than the TBIT scheme, which to our knowledge is the best among the CAC solutions for VoIP over WLAN previously appeared in the literature.

ACKNOWLEDGMENTS This project has been made possible through joint

collaboration with Cisco Advanced Architecture & Research Group and by Spanish Ministry of Science and Innovation under grant TEC2008-06826/TEC (project ARTICO).

REFERENCES [1] IEEE Std. 802.11-2007, “Wireless LAN Medium Access Control

(MAC) and Physical Layer (PHY) specifications”, 2007. [2] S. Garg, M. Kappes “Can I add a VoIP Call” in proc. IEEE ICC 2003,

Seattle, USA, May 2003 [3] S. Garg, M. Kappes “Admission Control for VoIP Traffic in IEEE

802.11 Networks” in Proc. of IEEE GLOBECOM, 2003

[4] H. Zhai, X. Chen, Y. Fang, “How Well Can the IEEE 802.11 Wireless LAN Support Quality of Service?”, IEEE Trans. On Wireless Comm., vol.4, n.6, November 2005.

[5] P. McGovern, S. Chung, S. Murphy, L. Murphy “Endpoint Admission Control for VoIPoWLAN” in Proc. ICT 2006, May 2006

[6] I.D. Chakeres, E. M. Belding-Royer “PAC: Perceptive Admission Control for Mobile Wireless Networks” in Proc. QSHINE 2004, Washingthon D.C., USA, 2004

[7] K. Yasukawa, A. G. Forte, H. Schulzrinne “Distributed Delay Estimation and Call Admission Control in IEEE 802.11 WLANs”, in Proc. of IEEE ICC, June 2009

[8] Various, Libpcap [online] available at: http://www.tcpdump.org [9] B. R. Tamma, N. Baldo, B.S. Manoj, R. Rao, “Multi-Channel

Wireless Traffic Sensing and Characterization for Cognitive Networking”, in Proc. of IEEE ICC, June 2009

[10] G. Bianchi “Performance Analysis of the IEEE 802.11 Distributed Coordination Function” IEEE Journal Selected Areas in Communications, 18(3), March 2000.

[11] N. Baldo, A. Zanella, “A Game Theoretic evaluation of Rate Adaptation strategies for IEEE 802.11 based Wireless LANs”, in Proc. of ICST GameComm, October 2009.

[12] ICT PERIMENTER Project webpage: http://www.ict-perimeter.eu/index.php

[13] M. Portolés, M. Requena, J. Mangues, M. Cardenete, “EXTREME: Combining the ease of management of multi-user experimental facilities and the flexibility of proof of concept testbeds”, in Proc. of IEEE TRIDENTCOM, 2006.

[14] The ns-3 network simulator [online] available at: http://www.nsnam.org

[15] N. Baldo, M. Requena, J. Nuñez, M. Portolès, J. Nin, P. Dini, J. Mangues, “Validation of the IEEE 802.11 MAC model in the ns3 simulator using the EXTREME testbed”, in Proc. of Simutools, March 2010.

[16] ITU-T Recommendation G.107, “The E-Model, a computational model for transmission planning”, December 1998.



Distributed Call Admission Control for VoIP over 802.11 WLANs Based on Channel Load Estimation

Documents

Transcript of Distributed Call Admission Control for VoIP over 802.11 WLANs Based on Channel Load Estimation