Empirical Analysis of Video Multicast over WiFi - MWNL

Document Sample
Empirical Analysis of Video Multicast over WiFi - MWNL Powered By Docstoc
					   Empirical Analysis of Video Multicast over WiFi
    Yeonchul Shin† , Munhwan Choi† , Jonghoe Koo† , Young-Doo Kim‡ , Jong-Tae Ihm‡ , and Sunghyun Choi†
                          Department of EECS and INMC, Seoul National University, Seoul, Korea
                                     Institute of Network Technology, SK Telecom, Korea
                Email: {ycshin, mhchoi, jhkoo}, {, jtihm},

   Abstract—Video multicast service is becoming one of the            On the other hand, a STA in AM always operates in the
most important applications over WiFi, due to the increasing          awake state. The access point (AP) buffers all the frames
popularity of WiFi for multimedia communication. On the one           destined to STAs in PSM and announces the existence of
hand, energy efficient operation is required to WiFi technology
due to the limited battery power of most WiFi-equipped devices.       such buffered frames through a traffic indication map (TIM)
In this paper, we empirically study the video multicast operation,    element in beacon frames. STAs in PSM wake up (i.e., switch
especially, along with power management operation in two              from doze state to awake state) periodically in order to receive
aspects: (1) whether commercial WiFi devices correctly operate        beacons. Then, the STAs operate following the information
as defined in IEEE 802.11 standard and (2) what problem the            specified in the TIM element of the beacon frame. Therefore,
standard-compliant operation can induce. From our experimental
results, we first figure out that some of commercial WiFi devices       IEEE 802.11 standard describes how access points (APs) set a
do not follow the standard with respect to the power saving           TIM and how STAs respond to the TIM in different situations.
operation, and this noncompliance worsens interoperability. We        Broadcast and multicast (B/M) frames are also buffered when
also find that the standard-compliant operation may cause              at least one station operates in PSM. Buffered B/M frames are
significant delay of voice over IP (VoIP) traffic, when video           transmitted only after delivery TIM (DTIM) beacons, special
multicast coexists with VoIP. Through this experimental study, we
provide the guidelines for energy-efficient video multicast service.   periodic beacons, so every station can receive B/M frames
                                                                      after when they receive DTIM beacons.
  Index Terms—Video multicast, WiFi, IEEE 802.11, power
saving mode, empirical study                                             In this paper, we empirically analyze the video multicast
                                                                      operation, especially, along with power management opera-
                       I. I NTRODUCTION                               tion in two aspects: (1) whether commercial WiFi devices
                                                                      correctly operate as defined in the standard and (2) what
   Today, the interest in WiFi [1] system, based on IEEE              problem the standard-compliant operation can induce. In order
802.11 standard [2], is increasing rapidly due mainly to the          to investigate these two issues, commercial WiFi devices
prevalence of smartphones. The Internet access and voice              are tested for both cases, i.e., when the power management
over IP (VoIP) communications have been the most popular              option of a STA is enabled and disabled. With respect to
applications over WiFi so far. As both the data rate of WiFi and      the standard-compliance issue, our experimental results show
the computational capability of mobile devices increase, video        that all the devices which have been tested in the experiments
streaming service is also becoming one of main applications           are not compliant with the standard when the power manage-
over WiFi today. Regarding the video streaming service, the           ment functionality is enabled. Moreover, such noncompliance
multicast transmission which transmits data simultaneously to         disturbs the communication and causes significant losses of
a group of users is more bandwidth-efficient than the unicast          multicast packets.1 On the other hand, we claim that the PSM
transmission which sends data to one user at a time. Therefore,       operation defined in the standard delays unicast traffic so that
video multicast, i.e., the video streaming service via multicast,     real-time unicast services, such as VoIP, suffer. We verify this
is being widely investigated in WiFi.                                 problem with the experiment and analyze the degradation of
   Recently, most of portable devices including laptops, smart-       quality-of-service (QoS) of the VoIP service. Based on the
phones, and tablet PCs are equipped with WiFi, so an energy           results, we discuss how to guarantee the energy-efficient video
efficient operation is required to WiFi technology due to the          multicast service without interfering real-time unicast services.
limited battery power of such devices. To this end, IEEE
802.11 standard defines a power management protocol, which                The rest of this paper is organized as follows. The operation
distinguishes the operations according to the transmission            of IEEE 802.11 power management is explained as back-
method, i.e., unicast and multicast.                                  ground in Section II. We present the video multicast operation
   For the efficient power management of WiFi stations                 of commercial devices in Section III. Section IV analyzes the
(STAs), IEEE 802.11 standard defines two power management              effect of video multicast on VoIP. Related work is introduced
modes, namely, power saving mode (PSM) and active mode                in Section V, and finally Section VI concludes the paper.
(AM). A STA in PSM toggles between the two different power
states, i.e., (1) awake state in which a STA can transmit or
receive signals and (2) doze state in which a STA turns off             1 We use the terms packet and (MAC) frame interchangeably across this
the radio components so that the STA cannot sense signals.            paper.
                               N-th bit for AID (N-1)
                                                                                                                    DTIM interval
                                 (1 N 2008)
          0   DTIM Count     (DTIM Period-1)                                                TIM       DTIM         TIM           TIM         DTIM
          Element            DTIM     DTIM     Bitmap     Partial Virtual                                                                             ...
                  Length                                                             M0           U   U   M1 M1 M1      M1 M0 U        U         M1
             ID              Count    Period   Control      Bitmap
Octets:       1      1         1         1       1           variable                                                                                 AP
                                                            (1 251)                       Buffering
                             1 bit for AID 0   Bitmap
                                                                                      Beacon frame whose bit for AID 0 is one
                         Fig. 1.   TIM element format.
                                                                                      Beacon frame whose bit for AID 0 is zero

                                                                                     M1 Multicast frame whose More Data bit is one
                            II. BACKGROUND                                           M0 Multicast frame whose More Data bit is zero
A. PSM operation for unicast service                                                 U    Unicast frame for other STA
                                                                                      Null data frame whose PM bit is one
   As mentioned earlier, there are two power management
modes, i.e., PSM and AM. Fig. 1 shows the TIM element                                     Awake state of STA             Doze state of STA
conveyed in beacon frames. An AP buffers frames destined
to a STA in PSM, and informs the existence of the buffered                        Fig. 2. PSM operation for multicast service defined in IEEE 802.11 standard.
frame via the partial virtual bitmap field in a TIM element
by setting a bit, corresponding to the association identifier
(AID)2 of the STA, to one. The AP periodically transmits                          frames. If the AP is unable to transmit all the buffered B/M
beacon frames, including a TIM element, with a predefined                          frames within a beacon interval following the DTIM beacon,
beacon interval, and STAs in PSM wake up periodically in                          the AP indicates that it will continue to transmit the remaining
order to receive the beacons.3 When a STA recognizes the                          buffered B/M frames by setting the bit for AID 0 of the bitmap
existence of the buffered frame addressed to itself, the STA                      control field of the TIM element in the next beacon frame to
transmits a power save (PS)-poll frame in order to request the                    one. STAs which receive a buffered B/M frame should remain
delivery of that frame. Then, AP transmits the buffered frame                     to be awake until receiving a B/M frame whose More Data
in response to the PS-Poll frame. If more frames destined to                      bit is set to zero or the beacon frame in which the bit for AID
the STA remain in the AP, the AP sets More Data bit in the                        0 of bitmap control field is set to zero.
medium access control (MAC) header. If the STA receives                              Fig. 2 shows an example of the operation for multicast
frames, whose More Data bit is set to one, it continues to be                     service when there is a STA in PSM and the DTIM period is
in the awake state until it receives a frame whose More Data                      three. The subsequent transmission of multicast frames after a
bit is zero.                                                                      DTIM beacon frame is depicted in the upper line. The lower
B. PSM operation for multicast service                                            line describes a power state transition of the STA. In this
                                                                                  figure, a null data frame whose power management (PM) bit
   An AP should buffer B/M frames if any STA associated                           is one is sent to notify AP that the STA switches to PSM.
with it operates in PSM. In order to announce the presence
of buffered B/M frames, the AP sets the bit for AID 0 of the
bitmap control field in a TIM element to one as shown in                             III. V ERIFICATION OF V IDEO M ULTICAST O PERATION
Fig. 1. The Delivery TIM (DTIM) period is determined as an
                                                                                     In this section, we experimentally verify the video multicast
integer multiple of the beacon interval, and it is announced via
                                                                                  operation of commercial WiFi devices in conjunction with
the DTIM period field in a TIM element of beacon frame. The
                                                                                  the power management. It is important to make sure that the
DTIM count field in a TIM element is decreased by one for
                                                                                  devices behave following the standard, because the standard-
every beacon frame transmission, and is reset to the value of
                                                                                  compliant operations ensure the devices to communicate each
(DT IM period−1) in the next beacon frame when it becomes
                                                                                  other. In Section III-A, we introduce our experimental setup
zero. A beacon frame, with the DTIM count field of zero, is
                                                                                  including the system topology and the experimental equip-
referred to as a DTIM beacon frame.
                                                                                  ments, and Section III-B shows the experimental results and
   Immediately after a DTIM beacon, the AP sends out the
                                                                                  the analysis of each device.
buffered B/M frames before transmitting any unicast frames.
Similar to the unicast transmission rule, the More Data bit of
a B/M frame is set to one if there are more buffered B/M                          A. Experimental Setup
  2 An AID is assigned to a STA by an AP, when the STA gets associated              We present our experimental setup including wireless net-
with the AP.                                                                      work environments and hardware/software configurations. We
  3 According to the standard, a STA in PSM does not need to wake
                                                                                  construct a simple network topology to analyze the operations
up at every beacon transmission time. When to wake up is actually an
implementation issue, and there is normally a tradeoff relationship between the   more clearly and experiment with off-the-shelf devices widely
energy saving and the delay performance depending on the wake-up period.          used in our daily lives.
                                                                                                             TABLE I
                        Video Multicast Streaming                                                     E XPERIMENTAL R ESULTS

                                                                                                            Apple AP             Cisco AP
                                                                                   PM function             On        Off       On        Off
                                             Wireless                               Nserver               4810     4810       4810      4810
   Video server                                            Video client
                              WiFi AP                                               Ncapturer             3688     4807       4711      4784
                                                                                     Nclient              2312     4557       2800      4734
                                                                                                       ( # of packets captured at each laptop)
                                                                                 Nclient /Nserver        48.1 % 94.7 % 58.2 % 98.4 %
                                                                                Ncapturer /Nserver       76.7 % 99.9 % 97.9 % 99.5 %
                                             WiFi                               Nclient /Ncapturer       62.7 % 94.8 % 59.4 % 99.0 %
                                        Packet capturer

Fig. 3.   System topology for the analysis of the video multicast operation.

   1) Wireless Network Environment: Fig. 3 shows our system
topology, composed of three laptop computers working as a
video streaming server, a video client, and a wireless packet
capturer, respectively. The video streaming server, connected
to the WiFi AP with a Ethernet cable, multicasts video traffic.
The client, a receiver of multicasted video traffic, receives the
video traffic from the AP, and the packet capturer reports
the list of packets observed from the wireless channel. In
this experiment, to focus on the measurement of the protocol
operations, we minimize losses due to channel errors and
                                                                                            Fig. 4.   A snapshot of Apple AP operation.
interferences by placing the client about one meter apart from
the AP and performing the experiments at night.
   2) Hardware and Software Setup: Two different AP devices                    capturer, and the number of packets received by the video
and one STA device, which are widely used, are tested in this                  client, respectively. When the client laptop disables the power
experiment:                                                                    management (PM) functionality (Off), the end-to-end delivery
   • Apple: AirPort Extreme Base Station [3] (Apple AP)                        ratio (i.e., Nclient /Nserver ) is over 94 %, thus video traffic is
   • Cisco: AIR-AP1232AG-A-K9 [4] (Cisco AP)                                   well delivered. Note that end-to-end delivery ratio, when the
   • Intel: Intel Pro/Wireless 3945 ABG Network Connec-                        packets are transmitted through multicast, is lower than that of
      tion [5] (Intel STA, embedded in Lenovo ThinkPad X60s                    the unicast case in general, due to the lack of retransmission
      laptop)                                                                  on group addressed frames. However, when the STA enables
   Every device utilizes IEEE 802.11g [2] Physical Layer                       the PM functionality (On), the measured end-to-end delivery
(PHY) in this experiment. The DTIM period and PHY rate for                     ratio is lower than 60 % for both cases. We present the
multicast frames are configured as 3 and 2 Mbps, respectively.                  detailed analysis for each device in the following. From this
The packet capturer uses AirPcap Nx USB Wireless Capture                       observation, we conjecture that this significant loss comes
Adapter [6] as a packet capturing network interface card                       from the misbehavior of the devices.
(NIC) and Wireshark Network Protocol Analyzer 1.4.1 [7]                           1) Operation of Apple AP: Fig. 4 shows a snapshot of the
as a software tool for analysis. The video is streamed and                     packet list captured at the packet capturer, when the Apple AP
played with Video LAN Connector Media Player 1.0.5 [8],                        is used. Packets with blue color are multicast video packets,
and the video (416x240 resolution, MPEG-4 codec, 800 kbps,                     and white-colored packets are 802.11 beacon frames. Fig. 4
29 fps) is encapsulated with MPEG-2 [9] transport stream (TS)                  presents that the AP stops transmitting video packets when the
and transferred through User Datagram Protocol and Internet                    beacon interval following a DTIM beacon ends, even though
Protocol (UDP/IP) network stacks.                                              the More Data bit of the last packet within the beacon interval
                                                                               is one. In other words, the Apple AP transmits multicast video
B. Measurement Results and Analysis                                            packets only after DTIM beacons even when the AP cannot
   The experimental results are summarized in Table I. The                     transmit all the buffered multicast packets during the beacon
numbers of packets recorded in each laptop during 60 sec-                      interval following a DTIM beacon, not continuing to send the
onds are presented. That is, Nserver , Ncapturer , and Nclient                 buffered packets as defined in IEEE 802.11 standard. Since
represent the number of packets transmitted from the video                     such an operation utilizes only 1/(DT IM period) of wireless
server to the AP through the Ethernet cable, the number of                     resources, it is inappropriate to support video multicast which
packets sent into wireless channel as measured by the packet                   requires more resources. In this experiment, because the AP
                                          Fig. 5.   Snapshots of Cisco AP and Intel STA operations.

is not able to deliver the 800 kbps video traffic successfully                                    Video multicast streaming
by using 2 Mbps PHY rate, the ratio of Ncapturer /Nserver is
76.7 % when PM function is On in Table I.
   Moreover, the Apple AP continues to transmit the video                                    Ethernet
multicast packets until the next beacon transmission when the                                                         Wireless
                                                                              Video server                                          Video client
STA changes its power management mode from AM to PSM.                                                     WiFi AP
Thus, the AP sends the video packets into the wireless channel                                 Internet
                                                                                                                          le   ss
while the STA sleeps in the doze state, causing additional
losses. In Table I, the number of packets measured at the client                          PSTN                                       VoIP phone
is smaller than that of the capturer due to this additional effect.          Telephone
   2) Operations of Cisco AP and Intel STA: Snapshots of                                                       Packet capturer
the packet lists measured at the packet capturer and the client,
when the Cisco AP runs, are shown in Fig. 5. The left and               Fig. 6. System topology for analyzing the effect of video multicast on VoIP.
right snapshots in Fig. 5 show a part of packets captured at
the packet capturer and the client, respectively. Decoding time
stamp (DTS) and presentation time stamp (PTS) are the timing            the standard. Such an operation ignoring the standard leads to
information for synchronized decoding and playing defined                the packet losses from number 4530 to number 4536 in the
in MPEG-2 TS [9]. From this information and time report                 snapshot of the packet capturer. Furthermore, like the Apple
in Wireshark, we detect the same packets in the packet lists            AP, the Cisco AP also does not halt the transmission when
measured at the packet capturer and the client. For example,            the STA switches to PSM, which causes some losses. These
the packets which are connected with a line in Fig. 5 are               two malfunctions, each of the STA and the AP, decrease the
the same packets measured by different devices. In the left             number of received packets at the client in Table I.
snapshot of Fig. 5, packets from number 4526 to number
4528 have been transmitted after the previous DTIM beacon.                       IV. E FFECT OF V IDEO M ULTICAST ON VO IP
Until that time, the AP has not been able to transmit all the
buffered packets, so the AP sets the bit for AID 0 of the                  Thus far, we have examined the video multicast operations
bitmap control field in the TIM element at number 4529. Then,            to see whether the devices’ operations follow the standard.
the AP keeps transmitting the multicast packets until all the           Basically, a standard-compliant behavior guarantees the suc-
buffered packets are transmitted at number 4536. After the next         cessful communication, but the standard may cause a side
DTIM beacon at packet number 4541, other buffered video                 effect, so called anomaly. In this section, we raises a side effect
packets are transmitted again. This operation of the Cisco AP           that may occur when video multicast and VoIP traffic coexist.
is compliant with the IEEE 802.11 standard as described in              As mentioned earlier, after a DTIM beacon, every buffered
Section II.                                                             multicast frame is successively transmitted, while any unicast
   On the other hand, upon not receiving packets after number           traffic is held. This strict priority of the buffered multicast
4529 beacon, the Intel STA goes to the doze state. Note that if         frames degrades QoS of the VoIP, which is supposed to get
a STA wakes up to receive B/M frames, it should stay in the             the highest priority in general. Since video applications require
awake state until receiving a B/M frame whose More Data                 lots of wireless resources, the problem becomes even worse
bit is set to zero or the beacon frame in which the bit for             when video multicast is served. In this section, we confirm
AID 0 of bitmap control field is set to zero. In that sense, this        this phenomenon with experiments and analyze the effect of
operation of the Intel STA can be said not to be compliant with         the video multicast on VoIP.


                                                                                                                                    Uplink (PM=Off)
                                                                                  0.2                                               Downlink (PM=Off)

                                                                                                                                    Uplink (PM=On)
                                                                                                                                    Downlink (PM=On)
                                                                                        0              20          40             60         80         100
                    0                 1                   2
                                                                                                                Inter-packet time (ms)
                                  time (sec)

                            (a) PM function is Off                                          Fig. 8.   CDF of inter-packet time for VoIP packets.

                                                                                                                  TABLE II

                                                                                                            Q O S P ERFORMANCE

                                                                                      Traffics                         Jitter (ms)        Loss ratio (%)

                                                                               U plink(P M = Of f )                      2.891                1.34
                                                                              Downlink(P M = Of f )                      5.146                2.60
                                                                               U plink(P M = On)                         7.509                2.27

                                                                              Downlink(P M = On)                        26.679                9.45
                        0              1                      2
                                   time (sec)
                                                                           packets are generated every 20 msec in our VoIP devices, about
                             (b) PM function is On                         fifty lines are drawn within one second. Some irregular points
Fig. 7. Packet transmission times of voice uplink traffic, voice downlink   are interpreted as errors and retransmissions. On the contrary,
traffic, and video multicast traffic.                                        different tendencies are observed from Fig. 7(b), when the
                                                                           PM function of the video client is On. In the case of voice
                                                                           uplink, the periodic pattern is maintained, because the VoIP
A. Experimental Setup                                                      phone contends with AP irrespectively of the video client. On
   Fig. 6 shows our system topology where the video multicast              the other hand, since the AP subsequently transmits buffered
and VoIP services co-exist. The additional components from                 multicast frames after a DTIM, the video traffic shows such
Fig. 3 are a VoIP phone and a telephone, attached to the public            a bursty traffic pattern. In this figure, DTIM beacons, after
switched telephone network (PSTN). In this experiment, we                  which the multicast transmission starts, are shown every about
make a call between the VoIP phone and the telephone, while                300 msec, because the DTIM period is set to three. Now, we
streaming video traffic via multicast. The Cisco AP (AIR-                   see that the voice downlink traffic also shows a bursty pattern
AP1232AG-A-K9) and the Intel STA (Intel Pro/Wireless 3945                  like the video. This pattern indicates that the VoIP frames
ABG Network Connection) are also used as the AP and the                    are delayed due to the video multicast transmission. Based on
video client, respectively. The model of the VoIP phone is                 this result, we confirm that the standard-compliant multicast
WPU-7800 made by Unidata Communication Systems. [10]                       delivery rule disturbs real-time unicast services when any STA
The VoIP traffic is generated using ITU-T G.711 [11] codec,                 works in PSM.
and VoIP packets are transmitted every 20 msec via realtime                   The cumulative distribution fuction (CDF) of the inter-
transport protocol (RTP). The VoIP phone does not enter the                packet time of VoIP packets is presented in Fig. 8 for four
PSM during a call. Other configurations are the same as those               different cases. Regardless of the power management of the
in Section III.                                                            video client, the distributions of inter-packet time are the
                                                                           same for the uplink cases. We observe that most of inter-
B. Measurement Results and Analysis                                        packet times are distributed between 15 and 25 msec. For
   Fig. 7 shows that the packet transmission times of VoIP                 the downlink (PM=Off) case, the distribution is similar to
uplink (VoIP-UL, i.e., VoIP phone to AP), downlink (VoIP-                  that of the uplink cases but more gradual, since the downlink
DL, i.e., AP to VoIP phone), and video multicast traffic within             traffic internally contends with the video multicast traffic.
a two-second time interval for both power management modes                 Unlike the other three cases, the inter-packet times of the
of the video client. A vertical line represents a point of a packet        downlink (PM=On) case are widely distributed, and about 5 %
transmission, and hence, periodic packet transmissions should              of the total samples experience even over 100 msec. Moreover,
show an even distribution of the lines across time. In Fig. 7(a),          as the VoIP downlink packets which have been delayed are
when the PM function of the video client is Off, the pattern               sent out successively, there are about 50 % of distribution
of every traffic is similar to the periodic traffic. Since VoIP              near 0 msec.
   We summarize the QoS performance of the VoIP traffic              IEEE 802.11e enhanced distributed channel access (EDCA),
in Table II, which indicates that the quality of VoIP suffers       and restrict the priority of buffered multicast frames according
from video multicast when the video client runs in PSM.             to the QoS characteristics of the traffic. In our future work, we
To calculate jitter and packet loss, successful packets, which      will experiment with more WiFi devices from other vendors.
are followed by an acknowledgement frame, are extracted in          Moreover, based on our observations, we are planning to
the Wireshark packet list. Comparing the number of packets          develop a scheme for energy-efficient video multicast.
successfully received with the RTP sequence number gap be-
tween the first and last packets, we can acquire the packet loss
ratio. Also using the RTP timestamp, jitter can be calculated         This work was supported by SK Telecom R&D program
according to the algorithm described in [12]. Especially, in        [A technology for enhancing Wi-Fi Video Multicast Perfor-
the downlink (PM=On) case, there occur significant jitter and        mance].
loss, thus the degradation of call quality is even perceivable                                   R EFERENCES
in practice.                                                        [1] WiFi Alliance,
                                                                    [2] IEEE 802.11, “Wireless LAN Medium Access Control (MAC) and
                     V. R ELATED W ORK                                  Physical Layer (PHY) Specifications (2007 revision), IEEE-SA, June
   There are several researches on the video multicast services.    [3] AirPort Extreme Base Station,
Closely related to this work, measurement studies on IEEE           [4] AIR-AP1232AG-A-K9,
802.11 power saving mode for multicast have been presented          [5] Intel     Pro/Wireless     3945      ABG       Network      Connection,
in [13, 14]. In [13], He et al. investigate the impact of           [6] AirPcap Nx USB,
beacon/DTIM period (internal factors) and background traffic         [7] Wireshark,
(external factors) in terms of the power consumption. Using         [8] Video LAN Connector Media Player,
                                                                    [9] Generic coding of moving pictures and associated audio information:
personal digital assistant (PDA), the authors examine the               Systems, ISO/IEC 13818-1 3rd edition, 2007.
power consumption in various environments. Moreover, the            [10] WPU-7800,
authors provide a theoretical framework of power saving proto-      [11] ITU-T Recommendation G.711: Pulse code modulation (PCM) of voice
                                                                        frequencies, 1988.
cols for multicast services in [15]. Hiraguri et al. also analyze   [12] H. Schulzrinne, S. Casner, R. Frederick, and V. Jacobson, “RTP: A
the influence of the multicast traffic on power consumption               Transport Protocol for Real-Time Applications,”RFC 1889, Jan. 1996.
in [14]. In addition to the analysis, the authors propose a         [13] Y. He, R. Yuan, X. Ma, and J. Li, “The IEEE 802.11 Power Saving
                                                                        Mechanism: An Experimental Study,” in Proc. IEEE Wireless Comm.
scheme that reduces the power consumption by distinguishing             and Networking Conf. (WCNC) 2008, Las Vegas, Nevada, USA, 2008.
broadcast and multicast transmissions instead of using one          [14] T. Hiraguri, M. Ogawa, M. Umeuchi, and T. Sakata, “Study of Power
bit for B/M frame in the existing TIM element. Though                   Saving Scheme Suitable for Wireless LAN in Multimedia Communica-
                                                                        tion,” in Proc. IEEE Wireless Comm. and Networking Conf. (WCNC)
these practical studies give beneficial observations on the              2009, Budapest, Hungary, 2009.
power saving operation for multicast, they neither consider         [15] Y. He, R. Yuan, and W. Gong, “Modeling Power Saving Protocols
the standard-compliant implementation of WiFi devices nor               for Multicast Service in 802.11 Wireless LANs,” IEEE Trans. Mobile
                                                                        Computing, vol. 9, no. 5, pp. 657-671, May 2010.
the effect on VoIP performance.
                VI. C ONCLUDING R EMARKS
   In this paper, we have evaluated the video multicast opera-
tions based on the experimental analysis. We first examine the
standard-compliance of commercial WiFi devices with respect
to the power management. Our experimental results demon-
strate that off-the-shelf devices operate without fully following
the standard, and such operations ignoring the standard disturb
the interoperability and decrease the system performance. We
also find that the standard-compliant operation has a side effect
that delays real-time unicast traffic. We verifiy the existence
of the problem empirically with VoIP traffic and analyze the
effect in detail.
   WiFi Alliance in charge of the certification of IEEE 802.11
are not testing the broadcast and multicast operations, so off-
the-shelf WiFi devices cannot guarantee the stable broadcast
and multicast services. Now, we insist that WiFi devices should
operate in a standard-compliant manner for the successful
video multicast service. Additionally, for the harmonic coex-
istence of video multicast and real-time unicast, IEEE 802.11
standard needs to be revised. It can be a possible solution
to give a higher priority to real-time unicast frames as in

Shared By: