IEEE n Aggregation Performance Study for the Multicast

Transcription

1 IEEE n Aggregaton Performance Study for the Multcast Yousr Daldoul, Toufk Ahmed, Djamal-Eddne Meddour To cte ths verson: Yousr Daldoul, Toufk Ahmed, Djamal-Eddne Meddour. IEEE n Aggregaton Performance Study for the Multcast. IFIP Wreless Days (WD 2011), Oct 2011, Canada. pp.1-6, 2011, < /WD >. <hal > HAL Id: hal Submtted on 9 Jan 2012 HAL s a mult-dscplnary open access archve for the depost and dssemnaton of scentfc research documents, whether they are publshed or not. The documents may come from teachng and research nsttutons n France or abroad, or from publc or prvate research centers. L archve ouverte plurdscplnare HAL, est destnée au dépôt et à la dffuson de documents scentfques de nveau recherche, publés ou non, émanant des établssements d ensegnement et de recherche franças ou étrangers, des laboratores publcs ou prvés.

2 IEEE n Aggregaton Performance Study for the Multcast Yousr Daldoul 1, 2, Toufk Ahmed 2, Djamal-Eddne Meddour 1 1 France Telecom - Orange Labs, France 2 LaBRI, Unversty of Bordeaux 1, France {yousr.daldoul, djamal.meddour}@orange-ftgroup.com, tad@labr.fr Abstract- One of the major features proposed n the IEEE n amendment s the use of frame aggregaton. Thus, two types of frame aggregatons are defned: MAC Servce Data Unt Aggregaton (A-MSDU) and MAC Protocol Data Unt Aggregaton (A-MPDU). As frame aggregaton s manly defned to ncrease the effcency of the MAC layer, ths feature s very needed to support multmeda traffc such as Hgh Defnton TV (HDTV). The multcast transport s another way to ncrease the MAC layer effcency by transmttng the same nformaton only once to many recevers. However, the standard does not defne a relable multcast and consequently the n s unable to provde a relable transport for multcast streams. The aa remans the only draft to handle the unrelablty ssue of the multcast, but the proposed soluton s defned only for the legacy standard and does not consder the aggregaton feature of the n. In ths paper we desgn a new extenson for the aa n order to take a full advantage of the n capabltes and we provde a new analytcal model to evaluate the performance of the MAC layer. We partcularly study the mpact of the frame aggregaton type on the MAC throughput for the multcast transport. Our results show that n an deal channel (no losses), A-MSDU slghtly outperforms A-MPDU when usng the data rate of 6.5Mbps and both schemes have the same performance wth rates of 130 and 300 Mbps. However the use of A-MPDU aggregaton becomes more effcent n a nosy channel. I. INTRODUCTION The IEEE n [1] s known to be the next generaton technology for the hgh throughput wreless networks. It allows data rates up to 600 Mbps. Wth ts hgh capacty, ths soluton becomes commonly used wthn new communcaton devces. The reduced effcency of the legacy MAC layer was frstly resolved by e usng the Block frame and the Block Acknowledgement transfers, and then enhanced by n usng frame aggregaton. Thus two aggregaton schemes are defned: MAC Servce Data Unt Aggregaton (A- MSDU) and MAC Protocol Data Unt Aggregaton (A-MPDU). Although the data rate of the PHY layer has consderably ncreased wthn the n compared to the legacy standard, the unrelablty problem of the multcast remans unresolved. Lke the legacy , the n does not use any feedback polcy when a frame s transmtted to a group address (broadcast or multcast). Hence the relablty of the multcast s hugely reduced compared to the uncast. In order to provde a relable multcast transport, the aa [3] defnes a new Acknowledgement (ACK) polcy for the However ths soluton does not consder the frame aggregaton feature of the n. In ths paper we desgn a new extenson to the aa amendment n order to take a full advantage of the n capabltes. Then, we evaluate through an analytcal model the performance of A-MSDU and A-MPDU aggregaton schemes to transport a relable multcast traffc. We partcularly show how the encapsulated MSDUs sze and the bt error rate affect the effcency of the MAC layer. The remander of ths paper s organzed as follows. Secton II ntroduces related work of some proposed multcast protocols we have studed, dscussng ther compatblty wth n and hghlghtng ther problems and drawbacks. Changes brought to the aa are descrbed n secton III. In secton IV we provde our analytcal model. We devote secton V to show the performance of the two aggregaton schemes. Fnally, n secton VI, we conclude. II. RELATED WORKS II-A. State of the art of relable multcast proposals Many protocols were desgned to address the unrelablty of the multcast n the WLAN [3-7]. They can be classfed nto two categores: ACK-based [3,4,5] and Negatve Acknowledgement (NAK) based [6,7] protocols. The ACKbased protocols use a smlar concept to that of the uncast ACK and they requre each multcast member to send an explct ACK. Most of them can mgrate easly to n snce the sender performs the MAC layer recovery based on a trustworthy reply. However ther performance should be studed agan when used wth the new aggregaton feature. In the NAK-based protocols, the recever wll reply wth a NAK only f the frame s receved wth errors. In some proposals, the mplementaton of the protocol s joned wth the selecton of a leader for the multcast group, where the leader s the only responder wth an ACK n case of recepton success. The mplementaton of a NAK based protocol remans a challengng task as t requres resolvng two constrants. 1) Elmnatng cases where the leader s ACK sgnal strength may hde a multcast member s NAK, 2) Decdng whether or not a NAK should be sent, snce the receved nformaton wthn a /11/$ IEEE

3 faulty frame s not coherent and s not enough to buld the approprate feedback n a relable way. The former constrant may be resolved by selectng a leader on the bass of the lowest lnk qualty crteron [7]. The latter s resolved by usng addtonal control frames (RTS/CTS, CTS-To-Self) to carry the trustworthy nformaton whch wll be used to buld a feedback. Most of NAK-based protocols are napproprate for the A- MPDU aggregaton scheme snce a NAK s used to acknowledge only one MPDU. II-B. Aggregaton performance studes In the lterature, many works have studed the aggregaton performance of n [8-10]. In [8] authors evaluated the A- MSDU, the A-MPDU and the two-level aggregaton schemes,.e. when one or more of the aggregated MPDUs encapsulate an A-MSDU. They proved that the two-level scheme s the most effcent one. In [9], authors proposed a novel analytc model to compare A-MSDU and A-MPDU schemes wth UDP and TCP. However they lmted ther evaluaton of the A- MSDU scheme to the maxmum supported length of a sngle A-MSDU wthout consderng the two-level scheme. The second weakness of the model s the use of an nfnte retry lmt whch s not a realstc value n a real network. In [10], authors focused on the overhead s mpact on the used scheme. The major weakness of the paper s gnorng the overhead generated by channel contenton and feedbacks. All mentoned studes have nvestgated only the uncast transport. To our knowledge, the multcast mode has not been thoroughly studed, and the few papers whch have studed t, have manly focused on ncreasng ts relablty. Therefore, analyzng the mpact of the aggregaton scheme on the MAC effcency has been gnored for the multcast transport. In ths paper we fll ths gap and we ntroduce our analytcal model. To acheve our evaluaton of the MAC aggregaton schemes, we need a relable multcast transport allowng the source to retransmt when a recepton error occurs. However, no ACK polcy was defned for the multcast n the n. Hence we defne a new extenson for aa complant wth the n aggregaton schemes. We use ths extenson to develop our analytcal model. III. OVERVIEW OF THE N AND AA AMENDMENTS III-A. Presentaton of the n amendment The n s desgned to defne the new verson of Hgh Throughput (HT) Wreless Local Area Networks (WLAN). Its theoretcal PHY data rate may reach 600Mbps. Hence, many features were defned to enhance the performance of ths amendment compared to the standard. In ths paper we focus on the frame aggregaton feature only and we present both of the aggregaton schemes: A-MSDU and A-MPDU. The A-MSDU scheme allows the aggregaton of several MSDUs nto the same MPDU. The destnaton address (DA) and sender address (SA) of each aggregated MSDU must match wth the recever address (RA) and transmtter address (TA) of the MPDU MAC header. All MSDUs wthn an A- MSDU should belong to the same traffc dentfer (TID). The maxmum length of an A-MSDU s 7935 bytes. The A-MSDU aggregaton reduces the transmsson overhead snce only one MAC header s generated for all the aggregated MSDUs. However t becomes neffcent n nosy channels: f only one MSDU s corrupted, the entre A-MSDU s rejected and all the aggregated MSDUs are lost. The A-MPDU scheme allows the aggregaton of several MPDUs nto the same PHY layer frame (PPDU). The maxmum supported length for an A-MPDU s bytes. In an A-MPDU each MPDU s encapsulated wth ts MAC header and CRC. Hence the corrupton of an MPDU s separately detected and does not affect the entre A-MPDU. However, the A-MPDU aggregaton requres more overhead than A-MSDU whch becomes heavy for small MSDUs. In order to provde a relable uncast transport, both and n defne an Acknowledgement (ACK) polcy. Moreover, n allows the use of the mmedate Block ACK (BACK) polcy wth sngle PPDUs carryng an A-MPDU when the ACK polcy of the A-MPDU s set to Implct Block ACK (.e. the ACK polcy subfeld of at least one aggregated MPDU s set to Normal ACK ). The legacy ACK polcy s used wth frames encapsulatng a sngle MPDU (.e. MSDU or A-MSDU) and requrng an mmedate feedback. Unlke the uncast, the n multcast transport does not use any feedback polcy. Besdes, the n forbds non-access Pont (AP) statons (STAs) from transmttng group addressed frames and only the AP s allowed to transmt such frames. III-B. Presentaton of the aa amendment The aa amendment [3] defnes the More Relable Groupcast (MRG) servce n order to support a robust and relable multcast audo/vdeo streamng over the WLAN. Ths servce provdes three ACK polces: MRG- Unsolcted-Retry, MRG-Drected Multcast Servce (DMS) and MRG-Block-Ack. When usng the MRG-Unsolcted-Retry ACK polcy, the sender defnes a retry lmt, say N, and then transmts each multcast frame N tmes wthout watng for any feedback after each transmsson. Ths polcy generates a sgnfcant overhead partcularly when the value of N s hgh, and becomes useless when the channel qualty s good. The MRG-DMS sends each multcast frame n uncast mode to each multcast group member. Therefore, ths ACK polcy guarantees the same uncast relablty level to multcast traffc at the cost of the bandwdth. DMS can be used to stream a standard multcast traffc to a few recevers number. However t does not scale well for hgh throughput traffc lke Hgh Defnton (HD) TV streams. The MRG-Block-Ack feedback polcy s desgned to operate wth the block frame transfer. For such transfer, the recever s allowed to send a BACK only when t receves an explct Block ACK Request (BAR) frame. The MRG-Block-Ack feedback procedure s llustrated n Fg. 1.

4 AP Multcast member 1 Multcast member 2 Multcast member 3 Data Data Data BAR Snce MRG servce s defned for multcast transmssons, the MRG BAR s slghtly changed compared to that of the uncast n order to add the followng nformaton: the multcast members and the acknowledgement order. When a BAR s used as part of MRG-Block-Ack, a new feld wth a varable length called MRG BAR nformaton feld s added to the BAR. Ths feld allows the AP to ask for a feedback from up to 2008 assocated STAs. Thus, the feld s length may reach 253 bytes. Ths length may ntroduce a useless overhead snce multcast groups n a WLAN have lmted members typcally fewer than 8, and becomes annoyng when each multcast frame needs to be ndvdually acknowledged. Although MRG-Block-Ack provdes the best relablty to bandwdth rato among the three proposed ACK polces, t does not allow an optmzed utlzaton of the n frame aggregaton feature. Snce an A-MPDU has a sgnfcant supported length allowng the aggregaton of many MPDUs nto the same PPDU, t becomes more convenent to request a feedback per frame partcularly for real-tme traffc subject to low latency constrants. Concernng the multcast group ntalzaton, the aa amendment does not defne a partcular procedure, but proposes the use of IGMP [11] snoopng to acheve group membershp detecton. IV. THE MRG-IMPLICIT-BLOCK-ACK FEEDBACK POLICY In ths secton we ntroduce a new extenson to the aa called the MRG-Implct-Block-Ack (IBA) feedback polcy. As for A-MPDUs transmtted n uncast wth mplct ACK, the MRG-IBA s used to allow each A-MPDU to be acknowledged ndvdually and mmedately. In order to allow MRG-IBA to operate properly, 2 functons are requred: membershp management and ACK notfcaton. The membershp management functon allows the AP to attrbute a Membershp Identfer (MID) ndvdually to each new detected member. In ths paper we do not defne the frame format used to notfy the MID attrbuton for space constrants. The AP mantans a table called Group Membershp Table (GMT) per multcast sesson. Ths table contans all the attrbuted MIDs for a specfc sesson. An example of the GMT table wth 4 members s llustrated n Table 1. Table 1. GMT of a multcast sesson wth address 01:00:5E:00:00:01 Member address 00:20:4E:38:A1:9B 00:20:7F:74:4E:25 00:20:92:27:8A:75 00:20:A6:61:1F:2B BACK MID BACK BACK Fgure 1. Typcal frame exchange scenaro wth MRG-Block-Ack polcy When a multcast group contans 16 members or less, the hghest attrbuted MID value should not exceed 16. Gaps may exst between attrbuted MIBs, and ths may occur when a MIB s recovered after a member departure. If the members number s hgher than 16, all MIDs between 1 and 16 should be attrbuted. Thus, the AP performs MID reassgnment when a MID between 1 and 16 becomes free. The choce of the value of 16 wll be motvated n the remander of ths secton. The ACK notfcaton functon s mplemented to assgn an order for the feedbacks after each frame. In order to allow ths functon to operate properly we defne a new-two-bytes-nlength feld n the frame called the MRG-IBA Btmap feld. Ths feld s present n the MPDU header of a group addressed frame when both of the followng condtons are satsfed: the MRG-IBA polcy s used and the ACK polcy of the frame s set to Normal ACK. The new feld s nserted between the HT control feld and the data feld of the MPDU. Each bt of the MRG-IBA Btmap feld corresponds to a MID. Therefore, the AP requests that the member wth a MID equal to N to respond f the bt number N s set to 1 and to forbear otherwse. As the btmap feld s 16 bts n length, up to 16 members may reply. We beleve that a typcal multcast group sze n a WLAN does not exceed 16 members. However, when the sze s hgher than 16, the AP may choose between allowng only the frst 16 members to reply tll some partcpants leave the group, and swtchng to the MRG-Block- Ack polcy. When the MRG-IBA agreement s establshed, multcast PPDUs encapsulatng a sngle MPDU (.e MSDU or A-MSDU) are also acknowledged usng the MRG-IBA procedure. However, requested members reply wth ACK frames nstead of BACKs. V. MODEL DESCRIPTION In ths secton we evaluate the n aggregaton schemes n the nfrastructure mode. We buld a Basc Servce Set (BSS) of one AP and 8 assocated STAs. All nodes are n complant STAs. The AP delvers only one stream belongng to one multcast sesson and no traffc other than the multcast one s transmtted on the medum. All hgh level packets are delvered usng UDP. The MRG-IBA agreement s establshed at the begnnng of the sesson. We defne a retry lmt of 7 for each MPDU and we suppose that MPDUs falures wthn an A-MPDU are not correlated. All encapsulated MSDUs have 1500 bytes length (L). Hence an A-MPDU can aggregate at most 42 MPDUs encapsulatng a sngle MSDU, and an A-MSDU can carry no more than 5 MSDUs. To evaluate the A-MSDU scheme wth more than 5 aggregated MSDUs, we use the two-level aggregaton scheme where several A-MSDUs are aggregated wthn an A-MPDU. We assume that the PHY frame header s always receved correctly snce t has a small sze compared to the data part of the frame and s always transmtted at the lowest but the most robust avalable data rate. All feedbacks (BACK and ACK

5 frames) are transmtted at the same data rate used to delver the multcast frame. We also assume that an ACK/BACK frame s always receved correctly gven that t has a small sze compared to that of the data frame. Moreover, we consder that all PPDUs are transmtted durng transmsson opportuntes (TXOP) whch are contenton free perods. Hence, the backoff procedure s not performed before a frame transmsson. Results are obtaned for the followng 2 mandatory data rates: 6.5 and 130Mbps, and for the optonal data rate of 300Mbps. We consder that only the HT-greenfeld format s used for all transmtted PPDUs. V-A. Mathematcal tools: Let p() be the bt error rate for Member (M ), for =1..K, where K s the multcast group sze. We defne X as the number of transmsson attempts. The probablty for a gven member M to receve correctly a packet n any of the N frst transmssons s gven n Equaton (1): P X N = 1 (1 Ps()) N (1) Then we derve the probablty to serve all the K recevers n any of the N frst transmssons n Equaton (2): K P X N = =1 (1 (1 Ps()) N ) (2) where Ps() s the packet success rate for M. Therefore, Ps() s a functon of the data length n and p(), and s llustrated n Equaton (3): Ps = (1 p()) n (3) It s obvous that the probablty to receve correctly a packet n X = 0 attempts s null, hence P(X=0) = P (X=0) = 0. We also express the MAC layer effcency as follows: E = (data_length/phy_rate)/transmsson_tme (4) As the MAC effcency n the multcast transmsson mode may depend on the recever s channel qualty, we evaluate the AP MAC effcency as perceved by each multcast member. Hence, ths effcency as perceved by a gven member M s llustrated n Equaton (5): E = (total_receved_data /PHY_rate)/transmsson_tme (5) V-B. A-MPDU aggregaton performance: As we assumed that the PHY frame header s always receved correctly, MPDU losses wthn an A-MPDU are ndependent from each other. Consequently we calculate the overhead per aggregated MPDU as the sum of the MPDU s overhead and the A-MPDU s transmsson overhead dvded by the number of aggregated MPDUs. Ths assumpton s vald snce we are transmttng MPDUs of the same length. We obtan the average A-MPDU transmsson tme n Equaton (6) for a number of A MPDUs aggregated MPDUs and a retry lmt of 7: T A MPDU = A MPDUs. where: 7 N=1 T transmsson A MPDUs _over head + T MPDU. (1 P X N 1 ) (6) T transmsson _over head = SIFS + T PHY _header and: +K SIFS + T BACK (7) T MPDU = (MPDU_overhead + L)/R (8) We defne, n Equaton (9), the mean receved MPDUs for a gven member M, A MPDUs, n order to evaluate the AP MAC effcency as perceved by M. A MPDUs = A MPDUs. P (X 7) (9) Hence, we derve n Equaton (10), the effcency equaton for the A-MPDU scheme as perceved by M : E A MPDU = A MPDUs.L/R (10) T A MPDU V-C. A-MSDU aggregaton performance: We consder the two-level aggregaton scheme n order to evaluate the MAC effcency usng the A-MSDU scheme when the length of aggregated MSDUs exceeds the maxmum supported length of 7935 bytes of the A-MSDU. Thus, we consder two cases: A MSDUs 5 and A MSDUs 5, where A MSDUs s the number of aggregated MSDUs. In the frst case only one A- MSDU, s transmtted. When A MSDUs 5, the A-MPDU aggregaton scheme s used, and each MPDU encapsulates an A-MSDU. Therefore the maxmum supported length of an A- MSDU s lmted to 3839 bytes and no more than 2 MSDUs can be aggregated wthn an A-MSDU. For calculaton smplcty, for A MSDUs 5, we consder only even values of A MSDUs n order to obtan MPDUs wth equal szes. Case 1: A MSDUs 5 In the A-MSDU scheme, f an aggregated MSDU s receved wth errors, the entre A-MSDU s rejected. Besdes when the PHY frame carres only one A-MSDU, recevers wll reply wth ACK frames nstead of BACKs. We derve the average A-MSDU transmsson tme n Equaton (11) for A MSDUs aggregated MSDUs and a retry lmt of 7: T A MSDU = where: 7 N=1 T transmsson _over head + A MSDUs. T MSDU. (1 P(X N 1)) (11) T transmsson _over head = SIFS + T PHY _header and: +T MPDU _over head + K SIFS + T ACK (12) T MSDU = (MSDU_overhead + L)/R (13) The mean receved MSDUs for a gven member M, A MSDUs, s expressed n Equaton (14). Ths equaton s smlar to Equaton (9), however the value of P(X=N), N=1..7, wll be affected by the ncreased sze of the MPDU. A MSDUs = A MSDUs. P (X 7) (14)

6 In Equaton (15), we derve the effcency equaton for the A- MSDU scheme as perceved by M : E A MSDU = A MSDUs.L/R (15) T A MSDU Case 2: A MSDUs 5 A-MSDU evaluaton for A MSDUs 5, follows equatons defned for the A-MPDU scheme wth few changes. Thus the average A-MPDU transmsson tme s gven n Equaton (16): T A MSDU = A MSDUs 2 where:. 7 N=1 T transmsson _over h ead A MSDUs /2 + T MPDU. (1 P(X N 1)) (16) T MPDU = (T MPDU _over head + 2. T MSDU _over head + 2. L)/R (17) The expresson of T transmsson_overhead remans unchanged lke n Equaton (7). The values of A MSDUs and E A-MSDU follow Equaton (14) and Equaton (15) respectvely. V-D. Analytcal results: In Fg. 2 we evaluate the MAC layer effcency of the AP for dfferent szes of A-MSDU and A-MPDU wth three data rates: 6.5, 130 and 300Mbps. In Fg. 2 (a), (b) and (c) we consder an deal channel case (.e. bt error rate p=0). Except n Fg. 2 (a) where A-MSDU slghtly outperforms A-MPDU, both aggregaton schemes have almost the same effcency. Ths s explaned by the fact that the overhead dfference between the two schemes has no effect when usng hgh data rates. In a real channel however, the A-MPDU scheme has the hghest performance snce t carres MPDUs wth smaller szes than those carred usng the A-MSDU scheme. In Fg. 2 (g), (h) and (), when the bt error rate ncreases, ncreasng the sze of an MPDU ncreases the packet error rate and hence ncreases the average transmsson tme. Thus usng A-MSDU wth up to 5 MSDUs does not provde a real enhancement to the MAC effcency compared to the legacy mode (.e. one MSDU per frame). In Fg. 3 we show the evoluton of the MAC effcency when varyng the bt error rate. As depcted n Fg. 3 (a), when no aggregaton s used (only one MSDU per frame), the MAC layer has a reduced effcency when usng hgh data rates (130 and 300Mbps). Ths s explaned by the fact that the requred tme for transmttng the frame overhead and recevers feedbacks s mportant compared to that requred to delver the encapsulated data. Fg. 3 (b) shows that an A-MPDU behaves better than an A-MSDU wth 5 aggregated MSDUs n a bad channel condton. As depcted n Fg. 3 (d), the use of the aggregaton feature mproves the MAC effcency wth hgh data rates. Fgure 2. The AP MAC layer effcency vs. the number of aggregated frames for a multcast group of 8 members

7 Fgure 3. The AP MAC layer effcency vs. the channel bt error rate for a multcast group of 8 members VI. CONCLUSION In ths paper we ntroduced the unrelablty problem of the multcast wthn the IEEE standard and we manly focused on the aa soluton and ts extensons. We also presented both of the aggregaton schemes proposed by the n amendment, namely A-MSDU and A-MPDU. Then we hghlghted the weakness of aa extensons to provde a relable multcast delvery of aggregated frames. In order to enhance the aa amendment, we presented a new extenson for the MRG servce called MRG-IBA polcy. Ths extenson allows the amendment to take a full advantage of the n aggregaton feature. The second contrbuton of ths paper s defnng a new analytcal model to evaluate the performance of A-MSDU and A-MPDU aggregaton schemes. Ths model s defned for a relable multcast transport usng the MRG-IBA recovery polcy and allows the estmaton of the MAC layer effcency. Our model shows that both aggregaton schemes have almost the same performance. However, A-MPDU behaves better than A-MSDU n a bad channel condton. In ths study we lmted our evaluaton on the use of analytcal results. Nevertheless, valdatng our model through smulatons remans an mportant task whch s mssng n ths paper but wll be scheduled n our future works. REFERENCES [1] Wreless LAN Medum Access Control (MAC) and Physcal Layer (PHY) Specfcatons, IEEE std , [2] Wreless LAN Medum Access Control (MAC) and Physcal Layer (PHY) specfcatons: Enhancements for Hgher Throughput, IEEE std n, [3] Wreless LAN Medum Access Control (MAC) and Physcal Layer (PHY) specfcatons: MAC Enhancements for Robust Audo Vdeo Streamng, IEEE P802.11aa/D1.01, June [4] Xaol Wang, Lan Wang, Relable Multcast Mechansm n WLAN wth Extended Implct MAC Acknowledgment, IEEE VTC, [5] N. Cho, Y. Seok, T. Kwon, Y. Cho, Transparent Uncast Translaton to Improve Qualty of Multcast over Wreless LAN, IEEE CCNC, [6] N. Cho, Y. Seok, T. Kwon, Y. Cho, T. Kwon. Multcastng multmeda streams n IEEE networks: a focus on relablty and rate adaptaton, Sprnger Scence+Busness Meda, LLC [7] J. Mroll, Z. L and Th. Herfet, Wreless Feedback Cancellaton for Leader-Based MAC Layer Multcast Protocols, IEEE ICSE, [8] D. Skordouls, Q. N, H. Chen, A. Stephens, C. Lu, and A. Jamalpour, IEEE n MAC frame aggregaton mechansms for nextgeneraton hgh-throughput WLANs, IEEE Wreless Communcatons, [9] B. Gnzburg and A. Kesselman, Performance analyss of A-MSDU and A-MPDU aggregaton n IEEE n, IEEE SARNOFF, [10] A. Saf, M. Othman, S. Subramanam and N. AbdulHamd, Impact of aggregaton headers on aggregatng small MSDUs n n WLANs, ICCAIE, [11] B. Can et al., Internet Group Management Protocol, Verson 3, RFC 3376, 2002.