HLBP: A Hybrid Leader Based Protocol for MAC ... - Semantic Scholar

HLBP: A Hybrid Leader Based Protocol for MAC Layer Multicast Error Control in Wireless LANs Zhao Li*, Student Member, IEEE, Thorsten Herfet†, Senior Member, IEEE *Department of Computer Science, University of Science and Technology of China, China & Telecommunications Lab, Saarland University, Germany †Telecommunications Lab, Saarland University, Germany {li, herfet}@nt.uni-saarland.de Abstract—In IEEE 802.11 Wireless LANs current standard MAC layer protocols do not provide any error correction scheme for broadcast/multicast. In our previous work, we enhanced a Leader Based Protocol (LBP) and proposed a Beacon-driven Leader Based Protocol (BLBP) for the MAC layer multicast error control. In this paper, we combine BLBP and packet level Forward Error Correction (FEC) and propose a Hybrid Leader Based Protocol (HLBP). HLBP transmits the original data packets using raw broadcast and retransmits parity packets using an improved BLBP which is based on block feedback. To guarantee the required Packet Loss Ratio (PLR) under strict delay constraints, we analyze HLBP, develop formulas to optimize its performance and evaluate its performance via simulation experiments. Both theoretical analysis and simulation results show that HLBP is much more efficient than LBP and BLBP especially for large multicast groups and is even more efficient than the best application layer multicast error correction scheme. Index Terms—HLBP, BLBP, Multicast Error Control, Wireless LANs

I. INTRODUCTION IEEE 802.11 Wireless LANs are one of the fastest growing network technologies in the field of wireless communication. Today it is becoming possible to supply wireless terminal users not only with high rate data connections, but also with real-time communication services. The emerging real-time multicast applications in Wireless LANs include the local distribution of High Definition TV (HDTV) and Digital Video Broadcasting (DVB) [1], video-on-demand, video conferencing, gaming, local VoIP, IPTV, Internet-Radio distribution, P2P broadcasting, etc. Most of these applications allow for a residual Packet Loss Ratio (PLR) but are bounded by strict delay constraints. However, the IEEE 802.11 standards do not comply with multicast data requirements. In particular, the current MAC layer sends multicast packets in open-loop as broadcast packets, i.e., without any acknowledgements. Currently the multicast errors in Wireless LANs are controlled in the application layer. The existing application layer multicast error control schemes include automatic repeat request (ARQ) [2], [3], forward error correction (FEC) [4], [5] and hybrid error correction (HEC) [6]-[8]. Unfortunately the total multicast delays of the application layer schemes are always high and sometimes do not satisfy the strict application delay constraints, or these schemes are not efficient when the delay constraints are tight. Compared with application layer schemes, MAC layer

multicast error control schemes lead to shorter delays. Kuri [10] proposed a Leader Based Protocol (LBP) which elects one of the multicast group receivers as the leader and allows acknowledgement (ACK) and negative acknowledgements (NACKs) from the leader and non-leader receivers respectively. However LBP is not reliable for the non-leader receivers and has poor performance at high error rates. In our previous work, we proposed a Beacon-driven Leader Based Protocol (BLBP) [11] which enhances LBP using a beacon frame before the data frame to lead the non-leader receivers to set timers and to announce the sequence number of the following data frame. BLBP can correct all the errors for all receivers in the MAC layer, it is more efficient than LBP and the multicast delay is very short. However, like other pure ARQ based schemes, BLBP is still not efficient at high error rates and with large numbers of receivers due to the limitation to scale. In this paper, we combine BLBP and packet level FEC and propose a Hybrid Leader Based Protocol (HLBP) for the MAC layer multicast error control. HLBP uses block erasure codes denoted as (n, k ) . The first original k − 1 data packets are transmitted using raw MAC layer broadcast and the original k th data packet is transmitted using an improved BLBP in which different parity packets are retransmitted based on block feedback if necessary. To guarantee the required PLR under strict delay constraints, we analyze HLBP and develop formulas to optimize its performance. We also evaluate the performance of HLBP via simulation experiments and compare HLBP with BLBP and a nearly optimal application layer multicast error control scheme called Hybrid ARQ (HARQ) Type I [7], [8]. The remainder of this paper is organized as follows. Section II presents the related work and motivation. In section III, we first introduce LBP and BLBP and then describe the main scheme of HLBP. We theoretically analyze the performance of HLBP in section IV. And in section V, we evaluate the performance of HLBP via simulation experiments and compare HLBP with BLBP and HARQ Type I. Finally we conclude this paper in section VI.

II. RELATED WORK AND MOTIVATION

For the application layer multicast error control, many authors [2], [3] studied ARQ schemes and concluded that when combined with feedback suppression and other accessorial techniques, ARQ schemes are effective to repair multicast

978-1-4244-2324-8/08/$25.00 © 2008 IEEE. This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE "GLOBECOM" 2008 proceedings.

packet losses for small groups with low error rates. However, these schemes always take long multicast delays due to application layer protocol waiting, MAC layer queuing, hardware handling, etc. Moreover, they are not efficient at high error rates and with large numbers of receivers due to feedback implosion and the limitation to scale. Another technique commonly used to handle losses for multicast in the application layer is FEC, whereby redundant information in the data stream enables the receiver to correct losses without contacting the sender. Rizzo [4] studied the feasibility of software encoding/decoding for packet level FEC. A (n, k ) block erasure code (such as Reed-Solomon (RS) code [5]) converts k source packets into a group of n coded packets, such that any k of the encoded packets can be used to reconstruct the k source packets. Usually, the first k packets in each group are identical to the original k data packets; the remaining n − k packets are referred to as parity packets. The advantage of using block erasure codes for wireless multicasting is that a single parity packet can be used to correct independent single-packet losses among different receivers. The integrated FEC/ARQ schemes are referred to as HEC schemes in this paper. Previous works [6]-[8] indicate that HEC schemes are much more efficient for recovering data packets than the schemes with either FEC or ARQ alone. We consider HARQ Type I from [7], [8], in which the sender sends a certain amount of parity packets using FEC following the original k data transmissions. If the loss rate obtained after reconstruction at the receiver is still too high, ARQ is used to retransmit more parity packets. Tan [8] developed formulas to optimize the performance of HARQ Type I while guaranteeing the required PLR under strict delay constraints. HARQ Type I is a nearly optimal application layer multicast error control scheme. However, these application layer multicast schemes always take long multicast delays or are not efficient when the delay constraints are tight. And the FEC based ones are not adaptive to the heterogeneity of receivers because the code has to be set based on the receiver with the worst channel condition. Compared with application layer schemes, MAC layer multicast error control schemes correct multicast errors locally, are more efficient and take a shorter time due to the faster feedback and retransmission. Moreover, for multi-hop multicast in wireless networks or with wired network and Wireless LAN as the last hop, the need for additional transmissions due to errors in the Wireless LANs puts unnecessary processing burden on the original remote sender and the entire network. If the access points (AP) (or base stations) were to take the responsibility of supplying retransmissions rather than the original sender, then the load of supplying retransmission gets distributed across access points and takes a shorter time. About the MAC layer multicast error control, besides LBP and BLBP which we will discuss in detail in the following section, Tourrihes [9] proposed a robust broadcast using a collision detector to inform the sender whether the broadcast packet is successful or not. However, this scheme can not guarantee the reliability of multicast transmissions because the

feedbacks are only from the detector instead of all receivers themselves. Unfortunately, as based on pure ARQ, all these MAC layer multicast error control schemes are not efficient for large multicast groups due to the limitation to scale. To overcome this problem, we combine BLBP and packet level FEC and develop HLBP.

III. MAIN SCHEME The MAC layer reliable multicast LBP, BLBP and HLBP require a slight modification to the IEEE 802.11 MAC layer protocols. As mentioned earlier, 802.11 DCF (Distributed Coordination Function) unicast – assumed Request-To-Send (RTS) and Clear-To-Send (CTS) is switched on to solve the hidden terminal problem – is more reliable than broadcast/multicast, because unicast uses RTS/CTS signaling and ACK/retransmission scheme in the MAC layer and broadcast/multicast does not. A. LBP and BLBP As shown in Fig 3.1, LBP [10] works as follows. A receiver is selected as the leader for the multicast group. The AP first sends a RTS frame to all receivers, and only the leader receiver transmits a CTS frame in reply to the AP. The AP is then assured that the channel is granted and starts the transmission of the data frame. The leader receiver sends an ACK in reply if the data is received correctly, or does nothing otherwise. If any non-leader receiver detects a transmission error, a NACK is sent. This NACK frame will collide with the ACK, if any, sent by the leader receiver. We refer to this ACK/NACK jam as JACK in this paper. And if the AP receives an ACK, this transmission is done. Otherwise, the AP repeats the whole procedure and retransmits the data until the number of times reaches the retransmission limit. Intuitively LBP has two main problems. First, when the entire data frame is lost, the non-leader receivers can not reply NACKs because they do not know when or how to send them, as the destination is unknown especially when RTS/CTS is not used for small data frames. As a result, LBP is not reliable for the non-leader receivers and in practice application layer multicast error control schemes have to be used to correct the rest errors in LBP. Second, LBP has poor performance when the channel error rates are high. The non-leader receivers send NACKs whenever the received frame is in error, regardless of whether this erroneous frame has been received correctly before or not. This is because the receivers in LBP can not access the data frame sequence number before the frame is received, as there is no such field in the structure of RTS/CTS frames for multicast. So the AP has to retransmit until all receivers receive the data frame correctly at the same time. There are a lot of unnecessary transmissions, hence LBP is not efficient particularly for lossy channels. Our previous work BLBP [11] enhances LBP with a MAC control frame called beacon shown in Fig 3.2. Besides the same fields in RTS/CTS frames, such as frame control header, transmission duration, receiver address (RA), transmitter address (TA) and frame check sequence (FCS), the beacon


Fig 3.4. The format of the beacon frame in HLBP

Fig 3.5. The format of the data frame in HLBP

Fig 3.1. Main scheme of LBP (Ri denotes receiver i)

Fig 3.2. The format of the beacon frame in BLBP

Fig 3.6. Main scheme of HLBP

Fig 3.3. Main scheme of BLBP

frame also includes the sequence number of the following data frame. The use of the beacon frame is to lead the non-leader receivers to set timers and to announce the sequence number of the following data frame. The main scheme of BLBP is shown in Fig 3.3. Unlike in LBP, the AP broadcasts a beacon frame before the data frame. On receipt of the beacon frame, each of the non-leader receivers sets a timer according to the beacon frame. After receiving the data frame, the leader receiver replies an ACK frame if the data is correct or it has already got the data based on sequence check, or does nothing otherwise. When the timer times out, each non-leader receiver replies a NACK if the data is erroneous and it has not received it correctly yet based on sequence check, or does nothing otherwise. For example, in the retransmission phase in Fig 3.3, although this time the data frame is lost, the leader receiver still replies an ACK because it knows this data frame has been received correctly already in the first transmission, thanks to the beacon frame. BLBP solves the problems of LBP well. All the non-leader receivers can send feedbacks when the timers time out. Both the leader receiver and non-leader receivers send ACK and NACK respectively based on sequence check, thanks to the beacon frame, hence it avoids the unnecessary transmissions in LBP. BLBP can correct all the errors for all receivers in the MAC layer and is more efficient than LBP. However, like other pure ARQ based schemes, BLBP is still not efficient at high error rates and with large numbers of receivers due to the limitation to scale. B. HLBP To overcome the problem of pure ARQ schemes, we

combine BLBP and packet level FEC and develop HLBP. The format of the beacon frame in HLBP is shown in Fig 3.4, which is similar to the beacon frame in BLBP. Instead of the sequence control field of the beacon frame in BLBP, the beacon frame in HLBP includes the block number and the packet index (in the block) of the following data frame. The use of the beacon frame is to lead the non-leader receivers to set timers and to announce the block number and the packet index of the following data frame. The format of the data frame in HLBP is shown in Fig 3.5, which has a block number field and a packet index field instead of the sequence number field in the original data frame in 802.11 DCF. The main scheme of HLBP is shown in Fig 3.6. HLBP uses a packet level FEC code ( n, k ) in the MAC layer. The AP transmits the first k − 1 data packets using raw broadcast and then transmits the k th data packet using an improved BLBP as follows. Before transmitting the k th data frame, the AP broadcasts a beacon frame to lead the non-leader receivers to set timers and to announce the block number and packet index of the data frame. After receiving the k th data frame, the leader receiver replies an ACK frame if it has already got at least k correct packets for the current block, or does nothing otherwise. When the timer times out, each non-leader receiver replies NACK if it has got less than k correct packets for the current block, or does nothing otherwise. Then if the AP receives an ACK, this transmission is done. Otherwise, the AP repeats the whole BLBP procedure and retransmits different parity packets until the number of times reaches the retransmission limit. Both BLBP and HLBP use a leader election scheme from LBP [10], which we do not discuss in detail here due to the space limit. As described in the above subsection, we note that BLBP is just a special case of HLBP when k =1. Clearly, HLBP can correct all the errors for all receivers due to the block feedback and retransmission in the MAC layer. HLBP achieves


complete feedback suppression thanks to the JACK scheme. However, the loss of beacon frames will decrease the reception rate of the non-leader receivers. Fortunately, the beacon frames are much more reliable (nearly error free) than data frames because they are much smaller and are transmitted using the lowest data rate, like other control frames in 802.11 (RTS/CTS/ACK). Moreover, due to RTS/CTS signaling, the beacon frames also avoid collision loss. Please also note that the data frames in HLBP can be interpreted by the stations that do not use HLBP, because the combination of block number and packet index fields just equals the original sequence control field. And the one octet fields for block number (0-255) and packet index (0-255) are just big enough for RS code in Wireless LANs, whose block length is always no more than 255. Moreover, LBP BLBP and HLBP can run without RTS/CTS exchanges for small data frames just like IEEE 802.11 DCF unicast. Although our discussion is in the context of IEEE 802.11 DCF, LBP BLBP and HLBP are actually applicable to all ACK/retransmission based MAC protocols, such as 802.11 PCF (Point Coordination Function) etc.

receivers get at least k correct packets is shown as follows:

IV. PERFORMANCE ANALYSIS

E[T ] = ( ( k − 1) TBST + ( E[ N ] − k + 1) TBLBP ) k ,

In this section, we analyze the performance of HLBP. As in most referenced papers, it is assumed that the MAC layer control frames (RTS/CTS/Beacon/ACK) are error free. This makes sense in practice because the control frames are very small and are sent at the lowest data rate, hence they are more reliable than data frames. It is also assumed that the error events at different receivers are independent. First, we calculate the final PLR for all R receivers, given the original packet error rate pr for receiver r , 1 ≤ r ≤ R and retransmission limit m . Let I denote the number of data packets lost for a block (k + m, k ) at receiver r . The Probability Distribution Function (PDF) of I can be calculated as formula (1). The first part denotes the probability that there are i original data packets lost, while the rest part is the probability that the correct parity packets are not enough to recover the lost data packets. min( i −1, m ) ⎛⎛k ⎞ ⎛⎛ m⎞ k −i ⎞ j ⎞ P r [ I = i ] = ⎜ ⎜ ⎟ pr i (1 − pr ) ⎟ ∑ ⎜ ⎜ ⎟ pr m − j (1 − pr ) ⎟ (1) ⎝⎝i ⎠ ⎠ j =0 ⎝ ⎝ j ⎠ ⎠

So we get the expected value of I shown in formula (2). k

E r [ I ] = ∑ ( iP r [ I = i ]) .

(2)

i =1

And the final PLR for receiver r is: 1 k PLR ( r ) = E r [ I ] k = ∑ ( iP r [ I = i ]) k i =1 =

min( i −1, m ) ⎛ ⎛ m ⎞ m− j 1 k ⎛ ⎛⎛k ⎞ i k −i ⎞ j ⎞⎞ ⎜⎜ i ⎜ ⎜ ⎟ pr (1 − pr ) ⎟ ∑ ⎜ ⎜ ⎟ pr (1 − pr ) ⎟ ⎟⎟ . (3) ∑ k i =1 ⎝ ⎝ ⎝ i ⎠ ⎠ j =0 ⎝ ⎝ j ⎠ ⎠⎠

Then we calculate the expected number of transmissions for all receivers to receive a data packet correctly. The probability that the AP needs to transmit at least k + i packets to let all

R

P[ N ≥ k + i ] = 1 − ∏ P r [ k + i − 1, k ] .

(4)

r =1

where P r [ k + i, k ] is the probability that receiver r has

received at least k correct packets after the AP sends k + i packets. The value is shown in formula (5). k +i ⎛ k + i ⎛ ⎞ j (k +i− j ) ⎞ P r [ k + i, k ] = ∑ ⎜ ⎜ (5) ⎟ ⎟ (1 − pr ) pr j j =k ⎝ ⎝ ⎠ ⎠ So we get the expected number of transmissions for a block, E[ N ] , shown in formula (6). And the expected number of transmissions for a packet is E[ N ] k . m

E[ N ] = k + ∑ P[ N ≥ k + i ] i =1

R k + i −1 m ⎛ ⎞ ⎛ ⎛ k + i − 1⎞ j ( k + i −1− j ) ⎞ = k + ∑ ⎜1 − ∏ ∑ ⎜ ⎜ ⎟ ⎟⎟ (6) ⎟ (1 − pr ) pr ⎜ i =1 ⎝ r =1 j = k ⎝ ⎝ j ⎠ ⎠⎠ Next we consider the expected channel holding time E[T ] for a packet, which is a natural criterion to use because the reciprocal of the average channel holding time provides a measure of throughput. The value is shown in formula (7).

(7)

where TBST = TDIFS + TDATA + TPLCP and TBLBP = TRTS + TCTS + TBEACON + TDATA + TACK + TDIFS + 4TSIFS + 5TPLCP . TRTS , TCTS , TBEACON , TDATA and TACK are the transmission time of frames RTS CTS BEACON DATA and ACK respectively. TDIFS denotes the Distributed Inter Frame Space time while TSIFS is the Short Inter Frame Space time. TPLCP denotes the PLCP (Physical Layer Convergence Protocol) preamble and header transmission time. Then we consider the determination of FEC code k and retransmission limit m , which are calculated by optimizing the performance of HLBP under both PLR constraint Pt arg et and delay constraint Dt arg et , shown in formula (8). min ( E[T ]) s.t.

PLR(r ) ≤ Pt arg et , 1 ≤ r ≤ R

(8)

kTload + ( m + 1)(Tcc + TBLBP ) ≤ Dt arg et And Tload is the multicast load interval (assumed the load interval is much larger than the transmission time of one packet). TCC denotes the channel contention time which can be calculated theoretically following [12] or by measurements in practice. Formula (8) is quite simple for calculation, the bigger k the better performance. Given a k , the retransmission limit m for a high error rate is also appropriate for lower error rates. In other words, given a delay constraint, HLBP is adaptive to the channel conditions and the heterogeneity of receivers. The determination of FEC code n is shown in formula (9), which is just like in HARQ Type I. In practice, the value of n can be set big enough to satisfy most channel conditions.


Moreover, if n is less than k + m , the parity packets can be used repeatedly in loop or at random if necessary. n ≥ k +m. (9) Please note that, given the FEC code k and a big enough n (no less than k + m ), HLBP needs the minimum number of transmissions for all receivers to get at least k correct packets for a block. This is because the number of transmissions for a block in HLBP is just equal to the maximum number of transmissions required by all receivers. There is no useless transmission due to the JACK scheme and block feedback. Moreover, given a delay constraint, HLBP can use a nearly maximum FEC code k due to the fast MAC layer block feedback and retransmission. As a result, HLBP should be very efficient and even more efficient than the application layer HEC schemes. We will explore this further via simulation experiments.

Fig 5.1 Average channel holding time of HLBP (delay constraint 100ms)

V. PERFORMANCE EVALUATION In this section we evaluate the performance of HLBP via simulation experiments and compare HLBP with BLBP and HARQ Type I. To calculate the average channel holding time of HARQ Type I, the application layer loss feedbacks are also included. As in most referenced papers, the multicast management overhead is not considered for all the schemes. We conduct our simulation study using ns-2 and implement LBP BLBP and HLBP based on the IEEE 802.11e simulation model from [13]. We use IEEE 802.11a parameters to model the physical layer. The data rate we choose is 24Mbps. The first receiver that joins the multicast group acts as the leader. The application targets and parameters are presented in Table 5.1. Note that the PLR target (1e-6) is very strict. The total payload length in the MAC layer is 1356 bytes, and there is no fragmentation in the MAC layer or the network layer. The application layer multicast error control scheme HARQ Type I is implemented based on the real-time transport protocol (RTP) [14]. In simulation, the packet error rates at all receivers are the same (denoted as p ) and the error events at different receivers are independent. Moreover, the MAC control frames (RTS/ CTS/Beacon/ACK) are error free from the error model. (The control frames also may be lost because they might collide with the background traffic.) First we validate our theoretical analysis. Fig 5.1 shows the average channel holding time of HLBP with different numbers of receivers. We can see that the simulation result and the theoretical analysis match very well. The average channel holding time increases slowly even when the numbers of receivers are very large, thanks to the FEC coding and block Table 5.1 Application targets and parameters PLR Requirement 1e-6 Delay Constraint 20-100ms RTP Payload Length 1316Bytes Multicast load interval 2.5ms RTT ≈ 3.5ms Original Error Rate ≤ 10% Packet sent 40-100e6

Fig 5.2 Average channel holding time with different error rates (delay constraint 100ms)

feedback. HLBP is very good for large multicast groups. Next we compare HLBP with LBP and BLBP. Fig 5.2 shows the average channel holding time with different error rates and different numbers of receivers. (An application layer ARQ scheme is used to correct the rest errors in LBP.) The results show that HLBP is always more efficient than BLBP and LBP, especially for large multicast groups. Note that LBP’s performance is very poor at high error rates and with large numbers of receivers. BLBP improves LBP well thanks to the beacon frame. We also note that BLBP is less efficient than LBP at low error rates and with small numbers of receivers. This is because in these cases the overhead of the beacon frame counteracts the benefit the beacon frame creates. Then we compare HLBP with BLBP and HARQ Type I. Fig 5.3 shows the average channel holding time with different numbers of receivers and different error rates. We can see that both HLBP and HARQ Type I are much more efficient than BLBP especially when the numbers of receivers are large. This is because BLBP is a pure ARQ based scheme and hence it is bad to scale. Moreover, the results show that HLBP is always more efficient than HARQ Type I. This is because the MAC layer JACK based block feedback and retransmission of HLBP are much faster than the application layer feedback and retransmission in HARQ Type I and HLBP can use a larger FEC code k , hence it is more efficient.


Fig 5.3 Average channel holding time with different numbers of receivers (delay constraint 100ms)

Finally, Fig 5.4 shows the average channel holding time with different delay constraints. HARQ Type I is not efficient (It is even less efficient than BLBP.) when the delay constraints are very tight. This is due to the fact that the application layer feedback and retransmission in HARQ Type I always take a long time and HARQ Type I has to switch to pure FEC scheme (no ARQ) under strict delay constraints, hence it is not efficient. However, due to FEC coding and the fast JACK based feedback in the MAC layer, HLBP is always very efficient even when the delay constraints are very tight.

Fig 5.4 Average channel holding time with different delay constraints (error rate 0.10)

REFERENCES [1] [2] [3] [4] [5]

VI. CONCLUSION

[6]

In this work, we combine BLBP and packet level FEC and propose HLBP for the MAC layer multicast error control. HLBP uses a block erasure code ( n, k ) . The first original k − 1 data packets are transmitted using raw MAC layer broadcast and the k th data packet is transmitted using an improved BLBP in which different parity packets are retransmitted based on block feedback if necessary. To guarantee the required PLR under strict delay constraints, we analyze HLBP, develop formulas to optimize its performance and evaluate its performance via simulation experiments. Both the theoretical analysis and simulation results show that HLBP is much more efficient than LBP and BLBP especially for large multicast groups. HLBP needs nearly the minimum number of redundancy transmissions among all schemes. HLBP is more efficient than the best application layer multicast error control scheme HARQ Type I, especially under strict delay constraints, even in one-hop Wireless LANs, should be much more efficient in multi-hop wireless networks or with wired network and Wireless LANs as the last hop. HLBP is very good for multicast error control in Wireless LANs, especially for large multicast groups. In the future, we will evaluate the performances of BLBP and HLBP in multi-hop wireless networks such as wireless mesh networks and ad hoc networks. We also plan to consider automatic data rate selection and other subtopics for BLBP and HLBP, and improve the QoS of multicast cross the application layer and MAC layer.

[7] [8] [9]

[10] [11]

[12] [13] [14]

U.H.Reimers, “DVB-The Family of International standards for Digital Video Broadcasting,” Proc. IEEE, vol.94, no.1, pp.173-182, Jan. 2006. J. Nonnenmacher and E. W. Biersack, "Optimal multicast feedback," in IEEE Infocom , 1998, p. 964, March/April 1998. G. Tan and Th. Herfet, “The Optimization of an RTP Level Hybrid Error Correction Scheme for DVB Systems in Wireless Home Networks under Strict Delay Constraint”, IEEE Trans. On Broadcasting, Nov. 2006. L. Rizzo, “Effective erasure codes for reliable computer communication protocols,” ACM Computer Communication Review, April 1997. J. Nonnenmacher, E.W. Biersack and et al., “Parity-Based Loss Recovery for Reliable Multicast Transmission,” IEEE/ACM Trans. Networking, Vol. 6, Aug. 1998. D. Qiao and K. G. Shin, “A two-step adaptive error recovery scheme for video transmission over wireless networks,” in Proceedings of IEEE Infocom, 2000, March 2000. G. Carle and E.W. Biersack, “Survey of error recovery techniques for IP-based audio-visual multicast applications,” IEEE Network, Vol.11, Issue 6, Nov.-Dec. 1997, pp.24 – 36. G. Tan and Th. Herfet, "Application Layer Hybrid Error Correction with Reed-Solomon Code for DVB Services over Wireless LANs," WiCOM2007, Shanghai, China, Sep. 2007. J.Tourrilhes, “Robust broadcast: improving the reliability of broadcast transmission on CSMA/CA”, in Proceedings of the Ninth IEEE International Symposium on. Personal, Indoor and Mobile Radio Communications, 1998. J Kuri, SK Kasera. “Reliable multicast in multi-access wireless LANs”, in Proceedings of Infocom, 1999. Z. Li and Th. Herfet. “Beacon-driven Leader Based Protocol over a GE Channel for MAC Layer Multicast Error Control”, International Journal of Communications, Network and System Science (IJCNS), 2008, to be published. Giuseppe Bianchi, “Performance analysis of the IEEE 802.11 distributed coordination function”, IEEE Journal on Selected Areas in Communications, Vol. 18, Issue 3, Mar 2000. Sven Wiethölter, Christian Hoene, “An IEEE 802.11e EDCA and CFB Simulation Model for ns-2”, http://www.tkn.tuberlin.de/research/802.11e_ns2/. J. Ott, S. Wenger, N. Sato, C. Burmeister, and J. Rey, “Extended RTP profile for RTCP-based feedback,” draft-ietf-avt-rtcp-feedback-11.txt, August 2004.
