J. Nonnenmacher, M. Lacher, M. Jung, E. Biersack, and G. Carle. How bad is reliable multicast without local recovery? In Proceedings of IEEE INFOCOM '98, San Francisco, CA, Mar. 1998.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....packets are often of use only to a small subset of the receivers. More sophisticated solutions that address these limitations by using techniques such as local repair, polling, or the use of a hierarchy have been proposed [10] 15] 26] 21] 34] but these solutions as yet appear inadequate [24]. Moreover, whereas adaptive retransmission based solutions are at best unscalable and inecient on terrestrial networks, they are unworkable on satellite networks, where the back channel typically has high latency and limited capacity, if it is available at all. The problems with solutions based ....

J. Nonnenmacher, M. Lacher, M. Jung, G. Carl, and E.W. Biersack, \How Bad is Reliable Multicast Without Local Recovery?" In Proc. of INFOCOM '98, San Francisco, April 1998, pp. 972979.

....Council, Taiwan, R.O.C. under Contract NSC 89 2213E007 127 and the program for promoting academic excellence of universities 89 E FA04 1 4. I. Introduction Using FEC (Forward Error Correction) for multicasting in the Internet has received a lot of attention recently (see e.g. 15] 4] [16], 20] 18] 12] Several important observations were made in the literature based on numerical comparisons of various protocol schemes that use FEC. For instance, it is noted in [15] that multicasting protocols with FEC scale well for a broad range of receivers and that the amount of ....

....made in [15] On the other hand, if there is no spatial loss correlation, then the proof for Theorem 1 may still be applicable by recomputing the G arter Ellis limit in Lemma 7(i) However, there may not exist a close form solution for S. Another possible extension is the tree model in [16]. For instance, consider a two tier multicasting model in Figure 8. Instead of sending packets directly to the r receivers, the sender may send packets to a subset of receivers with size r 1 and have each of them resend packets to receivers beneath them. As now the numbers of receivers in ....

[Article contains additional citation context not shown here]

J. Nonnenmacher, M. Lacher, M. Jung, E.W. Biersack and G. Carle, \How bad is reliable multicast without local recovery?," Proc. IEEE INFOCOM'98, San Francisco, CA U.S.A., March 1998.

....67, 150] or local groups to facilitate local repair [44] However, the e#ectiveness of such schemes is undermined by the lack of explicit topological information under IP multicast. As a result, there is no explicit basis to determine how receivers should be organized into repair trees or groups [91]. Consequently, complex and inac 87 curate methods have been used to infer the network topology [44, 150] Alternatively, manual building of repair trees [97, 78] and manual configuration of networks [67] have been proposed, with the drawback of requiring a priori knowledge of the set of ....

J. Nonnenmacher, L. Martin, J. Matthias, E. Biersack, and G. Carle. How bad is reliable multicast without local recovery. In Proc. IEEE Infocom '98, volume 3, pp. 972--9, San Francisco, CA., March 1998.

....are perfectly synchronized and thus only one NACK is sent back to the sender in case of packet loss. While the analysis in [7] is independent of the network topology, in [8] a delay analysis of generic ACK and NACK based protocols operating over star and linear topologies was presented. In [9] the e#ect of local recovery and retransmission of parity packets on bandwidth and delay of NACK based protocols is examined. While the bandwidth analysis is made in detail, the delay analysis is rather brief and comparatively simple. For example, they do not consider queuing delay in detail and ....

....results with simulation studies. 3 Protocol Classification and Description In this section we briefly describe and classify the reliable multicast protocols analyzed and simulated in this paper. A more detailed and more general description for some of these classes can be found in [10] 11] and [9]. The first considered protocol class is a receiver initiated protocol. Receiverinitiated protocols return only negative acknowledgments (NACKs) from receivers to the sender instead of positive acknowledgements (ACKs) When a receiver detects an error, e.g. by a wrong checksum, a skip in the ....

Nonnenmacher, J., Lacher, M., Jung, M., Carl, G., Biersack, E.: How bad is reliable multicast without local recovery. In: Proceedings of IEEE INFOCOM Conference on Computer Communications, New York, USA, IEEE Press (1998) 972--979

....called blocks. For each block, h FEC encoded packets are generated for suitably chosen h. Receivers can recover the original block by receiving any B packets out of the B h ones. When combined with an appropriate automatic repeat request (ARQ) technique, FEC incurs very low network delay [7]. The FEC approach can be employed in any of the protocols we have described, such as SRM and tree based protocols. An example is the SHARQFEC protocol that combines hierarchical scoping and hybrid FEC ARQ. It breaks the entire multicast group into hierarchically nested scopes and designates a ....

J. Nonnenmacher et al., "How Bad Is Reliable Multicast Without Local Recovery," Proc. IEEE INFOCOM, Mar. 1998, pp. 972--79.

....1 to reliably deliver a single packet to the entire group, or for the proxy of Figure 2 to reliably deliver a packet to the entire subgroup. 3. 1 The E[M] Measure E[M] is an important measure, which first appeared in [1] and ever since has been used in most analytical work on Reliable Multicast [18, 20, 19, 22, 25, 12]. n l pn pl Figure 2. Inside a subgroup E[M] is defined to be the expected number of times that a packet should be multicast (original transmission retransmissions) by the source, until all group members receive it correctly, at least once. It depends on the shape of the tree and the link ....

....3. Calculation of E[M] Although these formulas are general enough to capture any tree topology and any loss probabilities, they don t give any intuition and they are considered computationally intense, 19] In practice, either simulation or approximations are used for the calculation of E[M] [20]. 3.1.2 Reduction Techniques In this section, we present a way to find the equivalent link of any subtree. By equivalent link we mean a link with only one receiver that has the same loss behavior in terms of E[M] as the subtree it substitutes for. In other words, if a subtree is replaced by its ....

[Article contains additional citation context not shown here]

J. Nonnenmacher, M. Lacher, M. Jung, E. Biersack, G. Carle, "How bad is Reliable Multicast without Local Recovery?", INFOCOM ' 98

....work at the source and bandwidth. Previous work on performance analysis of reliable multicast has focused on the work at the source. The E[M ] measure of subsection IV A was defined and analytically calculated in [1] and ever since it has been used in most performance analysis work on RM: 12] [20], 10] 19] 26] 6] 18] Its calculation is computationally intense, so approximations [18] or simulations [6] are often used. We extend the understanding of this measure and we apply it on a realistic example. The second measure E[T ] captures the bandwidth used to reach all receivers at ....

....calculated as explained in section IV A. 1) Star topology: Such special topologies are realistic in cases like those measured in [27] where loss happens mainly on the links connecting the source and the receivers to MBONE and are widely used in evaluating reliable multicast protocols, 10] 12] [20]. For star topologies, every transmission from the root crosses all L links and the upper bound is tight: L (1) 2) Chain topology: Consider the chain topology of Figure 5, where node 0 is the proxy and nodes 1, 2, L are the receivers. Let us first consider the same loss rate p on all links. ....

J. Nonnenmacher, M. Lacher, M. Jung, E. Biersack, G. Carle, "How bad is Reliable Multicast without Local Recovery?", in Proc. IEEE INFOCOM '98, pp. 972-979, San Francisco, CA, USA, March 1998.

....approach. To name a few examples, SRM[ 10] RMTP[ 11 ] TMTP[ 12] and PGM[ 13] are all receiver initiated protocols that build relaibility on top of a best effort service. Further classification of reliable multicasting schemes is based upon the method used to recover from packet losses [14]. Some protocols, for instance RAMP [25] use centralized error recovery, also known as source based recovery, in which missing data is recovered exclusively from the original source. Others, like RMTP and LBRM[26] use distributed error recovery and may provide retransmissions from nodes other ....

Jorg Nonnenmacher, Martin Lather, Matthias Jung, Ernst W. Biersack, George Carle, How Bad is Reliable Multicast without Local Recovery ? In IEEE InFOCOM'98.

....over the sender initiated approach. To name a few examples, SRM[10] RMTP, TMTP and PGM are all receiver initiated protocols that build relaibility on top of a best effort service. Further classification of reliable multicasting schemes is based upon the method used to recover from packet losses [14]. Some protocols, for instance RAMP [25] use centralized error recovery, also known as source based recovery, in which missing data is recovered exclusively from the original source. Others, like RMTP and LBRM, use distributed errorrecovery and may provide retransmissions from nodes other than ....

Jorg Nonnenmacher, Martin Lacher, Matthias Jung, Ernst W. Biersack, George Carle, How Bad is Reliable Multicast without Local Recovery ? In IEEE INFOCOM'98.

....(Acks) of correctly received data. Producing a reliable multicast protocol that scales well with the number of receivers, in terms of network traffic and the processing required of the source and receivers, has proven to be a challenge, as demonstrated by the number of approaches taken in the past [42, 12, 65, 27, 26, 18, 36, 64, 30, 38, 24, 31, 57]. Moreover, as protocols for multicast error control are developed, 11 mechanisms must also be developed for multicast congestion control [15, 34, 7] similar to those developed for such unicast reliable protocols as TCP [20] One technique that has been used in the past by reliable multicast ....

....loss at the expense of delay, or whether they provide low delay, deadline oriented delivery at the expense of complete recovery from packet loss. Both classes of error control protocols can benefit from forward error correction (FEC) schemes, and many such schemes have been proposed recently [35, 36, 24]. Tracer s organization of the receiver set is topology based, and therefore can be considered as a replacement component for any reliable multicast protocol that needs to organize receivers according to packet loss correlation. The rest of this section describes how Tracer can be used as a ....

J. Nonnenmacher et al. How bad is reliable multicast without local recovery? In Proc. IEEE INFOCOM'98, March 1998.

....sender and receiver initiated protocols. Levine et al. 1] have extended this analysis to the class of ring and tree based approaches and showed that tree based approaches are superior. Bandwidth analysis of generic reliable multicast protocols were done by Kasera et al. 8] Nonnenmacher et al. [9] and Poo et al. 10] In [8] local recovery techniques are analyzed and compared. In Nonnenmacher et al. 9] they studied the performance gain of protocols using parity packets to recover from transmission errors. In [10] non hierarchical protocols are compared. Our paper extends previous work ....

....and tree based approaches and showed that tree based approaches are superior. Bandwidth analysis of generic reliable multicast protocols were done by Kasera et al. 8] Nonnenmacher et al. 9] and Poo et al. 10] In [8] local recovery techniques are analyzed and compared. In Nonnenmacher et al. [9] they studied the performance gain of protocols using parity packets to recover from transmission errors. In [10] non hierarchical protocols are compared. Our paper extends previous work in the following ways. The major di erence is that we consider overlapping retransmission scopes, which signi ....

[Article contains additional citation context not shown here]

J. Nonnenmacher, M. Lacher, M. Jung, G. Carl, and E. Biersack, \How bad is reliable multicast without local recovery," in Proceedings of IEEE INFOCOM Conference on Computer Communications, New York, USA, Apr. 1998, pp. 972-979, IEEE Press.

....in case of message loss. Another simplification they made is that control packets are reliably delivered. While the analysis in [5] is independent of the network topology, in [6] a delay analysis of generic ACK and NAK based protocols operating over star and linear topologies was presented. In [7] the effect of local recovery and retransmission of parity packets on bandwidth and delay of NAK based protocols is examined. While the bandwidth analysis is made in detail, the delay analysis is rather brief and comparatively simple. For example, they do not consider queuing delay in detail and ....

J. Nonnenmacher, M. Lacher, M. Jung, G. Carl, and E. Biersack, "How bad is reliable multicast without local recovery," in Proceedings of IEEE INFOCOM'98, New York, Apr. 1998, pp. 972--979, IEEE.

....of scalability. In [7] a more realistic system model including loss of control packets was analyzed and further protocol classes were introduced. Besides processing requirements, bandwidth efficiency was subject to several analytical studies for example by Kasera et al. 11] Nonnenmacher et al. [12] and Maihofer et al. 8] Regarding delay analysis, the first comparative delay analysis of sender and receiver initiated approaches was presented by Yamamoto et al. 6] and DeCleene [13] Yamamoto et al. have analyzed the expected average delivery delay and showed that receiver initiated ....

....are perfectly synchronized and thus only one NAK is sent back to the sender in case of message loss. While the analysis in [6] is independent of the network topology, in [13] a delay analysis of generic ACK and NAK based protocols operating over star and linear topologies was presented. In [12] the effect of local recovery and retransmission of parity packets on bandwidth and delay of NAK based protocols is examined. While the bandwidth analysis is made in detail, the delay analysis is rather brief. They concluded that local recovery techniques and parity packets outperforms other ....

J. Nonnenmacher, M. Lacher, M. Jung, G. Carl, and E. Biersack, "How bad is reliable multicast without local recovery," in Proceedings of IEEE INFOCOM'98, New York, Apr. 1998, pp. 972--979, IEEE.

....packets are often of use only to a small subset of the clients. More sophisticated solutions that address these limitations by using techniques such as local repair, polling, or the use of a hierarchy have been proposed [5] 10] 15] 16] but these solutions as yet appear inadequate [19]. Moreover, whereas adaptive retransmission based solutions are at best unscalable and inefficient on terrestrial networks, they are unworkable on satellite networks, where the back channel typically has high latency and UC Berkeley and International Computer Science Institute, Berkeley, ....

J. Nonnenmacher, M. Lacher, M. Jung, G. Carl, and E.W. Biersack, "How Bad is Reliable Multicast Without Local Recovery?" In Proc. of INFOCOM '98, San Francisco, April 1998.

....sender and receiver initiated protocols. Levine et al. 1] have extended this analysis to the class of ring and tree based approaches and showed that tree based approaches are superior. Bandwidth analysis of generic reliable multicast protocols were done by Kasera et al. 6] Nonnenmacher et al. [7] and Poo et al. 8] In [6] local recovery techniques are analyzed and compared. The system model is based on a special topology structure consisting of a source link from the sender to the backbone, backbone links and finally tail links from the backbone to the receivers. In Nonnenmacher et al. ....

....and Poo et al. 8] In [6] local recovery techniques are analyzed and compared. The system model is based on a special topology structure consisting of a source link from the sender to the backbone, backbone links and finally tail links from the backbone to the receivers. In Nonnenmacher et al. [7] a similar topology structure is used. They studied the performance gain of protocols using parity packets to recover from transmission errors. The protocols use receiver based loss detection with multicasted NAKs and NAK suppression. In [8] non hierarchical protocols are compared. In contrast to ....

J. Nonnenmacher, M. Lacher, M. Jung, G. Carl, and E. Biersack, "How bad is reliable multicast without local recovery," in Proceedings of IEEE INFOCOM'98, New York, Apr. 1998, pp. 972--979, IEEE.

....support very large multicast groups and to solve the scalability problem by reducing the number of acknowledgements computed by the sender. Moreover, the local loss recovery mechanism allows to reduce the retransmission time, what is essential for multimedia and cooperative applications. In fact, [NON98] showed that multicast transport protocols have improved performances if a local loss recovery mechanism is used. An architecture example respecting the described approach is depicted on Figure 1. 3 RG : Responsible of Group Figure 1. Example of sub groups (depending on the RTT and QoS) QoS1 ....

J. Nonnenmacher, M. Lacher, M. Jung, E. Biersack, G. Carle, "How bad is reliable multicast without local recovery ?", Proceedings of INFOCOM'98, San Francisco, USA, March 29 - April 2, 1998

No context found.

J. Nonnenmacher, M. Lacher, M. Jung, E. Biersack, and G. Carle. How bad is reliable multicast without local recovery? In Proceedings of IEEE INFOCOM '98, San Francisco, CA, Mar. 1998.

No context found.

J. Nonnenmacher, L. Martin, J. Matthias, E. Biersack, G. Carle, "How bad is Reliable Multicast without Local Recovery?," Proc. I IEEE INFOCOM'98, volume 3, pp. 972-9, March 1998.

No context found.

J. Nonnenmacher et al, "How bad is reliable multicast without local recovery?,"Proc of IEEE INFOCOM, March 1998.

No context found.

Nonnenmacher, J., M. Jung, M. Lacher, E. W. Biersack, G. Carle. How bad is reliable multicast without local recovery? in IEEE INFOCOM'98. 1998. San Francisco, U.S.A.

No context found.

J. Nonnenmacher, M. Lacher, M. Jung, E. Biersack, and G. Carle. How Bad is Reliable Multicast without Local Recovery. In Proceedings of INFOCOM'98, San Francisco, CA, March 1998.

No context found.

J. Nonnenmacher, M. Lacher, M. Jung, E. Biersack, and G. Carle. How bad is reliable multicast without local recovery? In Proceedings of IEEE INFOCOM '98, San Francisco, CA, Mar. 1998.

No context found.

J. Nonnenmacher, M. Lacher, M. Jung, E.W. Biersack, G. Carle. How bad is reliable multicast without local recovery? In Proc. IEEE INFOCOM'98, March 1998.

No context found.

J. Nonnenmacher, M. Lacher, M. Jung, G. Carl, and E. W. Biersack, "How Bad is Reliable Multicast Without Local Recovery ?", Proc. IEEE Infocorn '98, San Francisco, CA, April 1998.

No context found.

J. Nonnenmacher, M. Lacher, M. Jung, G. Carl, and E. Biersack, \How bad is reliable multicast without local recovery," in Proceedings of IEEE INFOCOM'98, pp. 972-979, IEEE, (New York), Apr. 1998.

First 50 documents

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC