Rajagopalan, B. Reliability and scaling issues in multicast communication. In Proc. ACM SIGCOMM (1992), pp. 188--197.  Home/Search   Document Not in Database   Summary   ACM   TOC   Related Articles   Check

....a multicast session without an error and in order. As the scale of networks like the Internet increases in terms of both the number of users and geographic span, reliable multicast protocols have faced two intrinsic scalability challenges: acknowledgement implosion and exposure to retransmissions [1], 2] 3] 4] Implosion refers to the situation where the source and the network are overwhelmed as receivers send the source negative or positive acknowledgements for notification of loss detection or successful delivery. It consequently reduces the overall performance of a session. Exposure ....

B. Rajagopalan, "Reliability and Scaling Issues in Multicast Communication," ACM SIGCOMM 92, August 1992

....are driving researchers to consider distributed transport protocol design for large scale, wide area multicast. 2.4.1 Distributed Approach at the Network Layer One distributed technique is to push error control and state feedback services into the network layer. For example, Rajapolan [85] presents the reliable multicast (RM) protocol which distributes protocol processing into the network layer to achieve reliable, K reliable, and subgroup reliable multicast. Error recovery is local between router pairs, thereby incurring low network overhead and recovery delay. Each channel ....

B. Rajagopalan. Reliability and Scaling Issues in Multicast Communication. In Proc. Sigcomm '92, pages 188--198, Baltimore, Maryland, August 1992.

....of link cost, and considering single group decision making when all the users subscribe simultaneously. In [BFC93] different (TCP IP packet switched) network environments is assumed, without explicit real time guarantees. More results about statistical models can be found in [BFC93, CESZ91, Raj92] An alternative measure of network performance is the amortized throughput defined as the average over time of the number of bits transmitted by the accepted connections. In this setting, the network s bandwidth is assumed to be insufficient to satisfy all the requests so some of the requests ....

Bala Rajagopalan. Reliability and scaling issues in multicast communication. In Proc. of the Annual ACM SIGCOMM Symposium on Communication Architectures and Protocols, Baltimore, MD, pages 188--197, August 1992.

....on reliable multicasting only captures some of the properties of our work. For example, in [18] protocols for reliable one to many multicasting are discussed. In [1] protocols are designed for a distributed environment and do not address issues such as ACK NACK, timeout and retransmission. In [19] an interesting window based scheme for inter gateway protocol is presented. The protocol is based on creating and maintaining multicast trees at the gateways. 1 In window based flow control protocol the size of the window determines the number of outstanding packets, i.e. the packets which have ....

B. Rajagopalan. Reliability and scaling issues in multicast communication. Proc. of ACM SIGCOM'92, pages 188--198. 22

....as MH FIFO) The modifications to be made to the multicast protocol to accommodate each of the three models of user location is also described. 1 Introduction In distributed systems, efficient mechanisms are required that enable a process to multicast a message to a group of other processes [5][13]. In a Personal Communication System (PCS) this problem of multicasting a message from a mobile host to other mobile hosts (henceforth referred to as mobiles) becomes more difficult because of the constant movements of the mobiles. A mobile that initiates a multicast message to a group of other ....

B. Rajagopalan, "Reliability and scaling issues in multicast communication", in Proc. ACMSIGCOMM, pp. 188-197, 1992.

....send the ACK by piggybacking only in the case when the node itself has a message to multicast [4, 5] Therefore, messages should be generated from each node at rather equal rate so that all nodes send the ACKs in turn. Other multicast protocols, for example, Acknowledgement gathering protocol [6] or MS protocol [7] are not examined in this paper, since they deal with the store and forward networks, not the broadcasting environments. The multicast protocol proposed in this paper reduces the number of control messages for ACKs by each receiver node sending the ACK in round robin, and ....

B. Rajagopalan, "Reliability and scaling issues in multicast communication," ACM SIGCOMM, vol. 22, no. 4, pp. 188--198, August 1992.

....the set of receivers. Finally, the GCS can be seen as a way to provide different quality of service to outgoing data streams, providing filtering capabilities. Although targeting other problems, a similar approach was developed in the network layer while introducing gathering filtering functions [Raj92, Zha93]. An important parameter is the cost of such functions that might be more efficient to provide at the network layer than at upper layers. They have to be combined with other approaches such as the one we proposed at the transport layer. Another important issue is the one related to the impact of ....

Rajagopalan B., "Reliability and Scaling Issues in Multicast Communications", Sigcomm'92, pp188-198, 1992.

....of a given packet. 62 There are many important issues concerning multicast what the nature of the tree is (source rooted [27] centered [2] how the tree is formed, scaling, managing resources, distributing addresses, controlling group membership, achieving reliable packet delivery, and so on [8, 26, 90, 112, 2]. In keeping with the scope of this thesis, however, the discussion here is largely limited to the form of the multicast address and related information required in the packet header. A fundamental distinction among tree types is source rooted trees versus non source rooted trees. Source rooted ....

B. Rajagopalan. Reliability and Scaling Issues in Multicast Communication. Proceedings of ACM SIGCOMM 92, pages 188--198, August 1992. 185

....as distributed data processing [1] broadcast information systems [2] and teleconferencing, among others, which represent the driving forces behind the development of high speed networks. It is, therefore, important that next generation networks employ efficient broadcast multicast mechanisms [3]. Lightwave technology has emerged as a promising candidate for implementing this next generation of multiuser high speed communication networks. A wide variety of novel design approaches and network architectures has been proposed in the literature to exploit the unique properties of the optical ....

B. Rajagopalan. Reliability and scaling issues in multicast communication. In Proceedings of SIGCOMM '92, pages 188--198. ACM, 1992.

....only the lost packets are retransmitted. For multicast communication we assume that each receiver sends acknowledgements to the sender. As a result there is an implosion effect at the sender as the size of the receiver set increases. While strategies have been proposed to aggregate these acks [15] or for a single receiver to multicast acks [1] the feasibility of these ideas in a WAN environment is unclear. We believe that it is appropriate to reduce the amount of reverse traffic by using either a block based ACK scheme [9] or a NACK scheme [11] While NACK based schemes are likely to ....

Bala Rajagopalan. Reliability and Scaling Issues in Multicast Communication. In Proceeding of ACM SIGCOMM, pages 188--198, August 1992.

.... if a voice packet(s) gets lost, it makes no sense to re transmit the lost packet since it only has relevance if preceeded and succeeded by the other packets generated in the first instance. Secondly, acknowledgements converging on a data source from any number of receivers (so called concast [81]) are a considerable burden to a sender. For large groups this could also cause serious congestion problems at or near a sender. A heterogeneity of network types may pervade any layer of the Internet hierarchy, for example, X.25 [17] Frame Relay [13] SMDS [62] ATM [61] Of these, only SMDS ....

B. Rajagopalan. Reliability and Scaling Issues in Multicast Communication. In Conference Proceedings of ACM Sigcomm'92, pages 188--198. ACM SIGCOMM, August 1992.

....to destination failures and is not appropriate if the destination set is dynamic. 4. 3 Experimental Results A reliable multicast service can be used by many applications besides parallel processing, such as file replication, software distribution maintenance, and medical image broadcasting [20]. Evaluation criteria depend on the application. In some cases, throughput is paramount, while in others, turnaround time is most critical. Since we are focusing on cluster based scientific computing, the transfer time for data from the memory of the source to the memories of the destinations is ....

B. Rajagopalan, "Reliability and Scaling Issues in Multicast Communication," in Proceedings of ACM SIGCOMM '93, 1993.

....problem with this approach is that it does not capture interaction between different commodities. In [BFC93] broadcast is considered for environments that do not reserve bandwidth and hence cannot provide quality of service guarantees. A comprehensive survey of efficiency issues can be found in [Raj92] Other work on statistical approachs to multi cast can be found in [BFC93, CESZ91, Raj92] The existing work in the competitive on line setting [AAP93] applies only to a special case where there is a single request for each broadcast group or all requests for the broadcast group arrive at once ....

....In [BFC93] broadcast is considered for environments that do not reserve bandwidth and hence cannot provide quality of service guarantees. A comprehensive survey of efficiency issues can be found in [Raj92] Other work on statistical approachs to multi cast can be found in [BFC93, CESZ91, Raj92] The existing work in the competitive on line setting [AAP93] applies only to a special case where there is a single request for each broadcast group or all requests for the broadcast group arrive at once with the option of accepting all of the requests or none of the requests. 2 Problem ....

Bala Rajagopalan. Reliability and scaling issues in multicast communication. In Proc. of the Annual ACM SIGCOMM Symposium on Communication Architectures and Protocols, Baltimore, MD, pages 188--197, August 1992.

....with this approach is that it does not capture the interaction between different broadcast groups. In [BFC93] broadcast is considered for environments that do not reserve bandwidth and hence cannot provide quality of service guarantees. A comprehensive survey of efficiency issues can be found in [Raj92] Other work on statistical approachs to multi cast can be found in [BFC93, CESZ91, Raj92] The existing work in the competitive on line setting [AAP93] applies only to a special case where there is a single request for each broadcast group or all requests for the broadcast group arrive at once ....

....groups. In [BFC93] broadcast is considered for environments that do not reserve bandwidth and hence cannot provide quality of service guarantees. A comprehensive survey of efficiency issues can be found in [Raj92] Other work on statistical approachs to multi cast can be found in [BFC93, CESZ91, Raj92] The existing work in the competitive on line setting [AAP93] applies only to a special case where there is a single request for each broadcast group or all requests for the broadcast group arrive at once with the option of accepting all of the requests or none of the requests. Structure of ....

Bala Rajagopalan. Reliability and scaling issues in multicast communication. In Proc. of the Annual ACM SIGCOMM Symposium on Communication Architectures and Protocols, Baltimore, MD, pages 188--197, August 1992.

....performance of selected key aspects of such a mechanism, and to provide a first set of guidelines for the sizing of main parameters of the protocol in order to achieve a target Quality of Service (QoS) The concept of multicast communications has received much recent attention. Service, protocols [Cha84, Bir91, Gar91, Ngo91, Raj92] and mechanisms [Agh94,Pin94] have been addressed for various network environments [Car94a, Dee94, Rob95] One can point the work carried out in standardization committees (IETF, ANSI, ISO, ATM, etc. Dee89, Coh91, Mou92, Coc92, ISO93, MPC95] and projects CIO [Mil93, Mat93] Berkom [Del93] Cesame ....

Rajagopalan B., "Reliability and Scaling Issues in Multicast Communications", Sigcomm'92, pp188198, 1992.

....[12] and teleconferencing, a significant portion of the overall traffic in future broadband networks will be of the multi destination type. As a result, a significant amount of research has been aimed at designing efficient multicast mechanisms for next generation communications environments [13, 14, 15, 16]. In the context of single hop networks, the issue of multi destination traffic has been addressed in [17] where a controlchannel based multicast protocol is presented, and in [18] where we investigated the throughput characteristics of various multicast schemes. In this paper we suggest, ....

B. Rajagopalan. Reliability and scaling issues in multicast communication. In Proceedings of SIGCOMM '92, pages 188--198. ACM, 1992.

....layers. Second, they are primarily concerned with routing messages (datagrams) to from an MH regardless of its location in the network. Third, their focus has been restricted to point to point messages. On the other hand, there exists a number of multicast routing protocols for static hosts, e.g. [12, 17, 30], but the effects of mobility on multicast routing [4] has started to received attention only of late. Similarly, there also exists a body of work on reliable multicast protocols [18, 21, 28] for static hosts. In theory, such protocols can support mobile endpoints if the underlying network is able ....

....is distributed amongst multiple MSSs at the expense of a higher latency for each message to traverse the multiple levels of the hierarchy. Alternatively, instead of using point to point message delivery between MSSs, a (best effort delivery) multicasting service provided at the network layer [12, 17, 30] could be used to efficiently route copies of multicast, view change( or delete( messages amongst the MSSs comprising a host view, with 7 This array stores the sequence numbers of the last message received by h from each MSS in the current incarnation of H G . additional transport layer ....

Bala Rajagopalan. Reliability and scaling issues in multicast communication. In Proc. ACM SIGCOMM, pages 188--197, 1992.

....protocols apply to local area networks and do not scale well in wide area networks, mainly because the entities involved in the protocol need to exchange several control messages for coordination among themselves. In addition, they do not address fundamental issues of acknowledgment implosion [11] and propagation delays in wide area networks. The goal of this paper is to describe the design and implementation of a reliable multicast transport protocol for wide area networks. Called RMTP, the protocol provides sequenced, lossless delivery of bulk data from a sender to a group of receivers. ....

B. Rajagopalan, "Reliability and Scaling issues in Multicast Communication," Proceedings of ACM SIGCOMM '92, pages 188--198, Aug. 1992.

....network are both special cases of multicast. Multimedia applications are envisioned to benefit most from multicast communication capabilities. Some frequently cited examples include video or tele conferencing, distributed databases, distributed games, mass mailings, and video on demand services [1 3]. In a video on demand application, for instance, a single server provides a one to many transmission for customers who join the same movie at approximately the same time. A video conference, on the other hand, is likely to be a many to many transmission with multiple parties wishing to transmit ....

B. Rajagopalan, "Reliability and scaling issues in multicast communication," in Proc. of ACM SIGCOMM, (Baltimore, MD), pp. 188--198, October 1992.

....maintaining low end to end delay. This is achieved by reducing unnecessary retransmissions by the sender. In addition, we adopt a novel technique of grouping receivers into local regions and generating a single acknowledgment per local region to avoid the acknowledgment implosion problem [R92] inherent in any reliable multicasting scheme. We also use the principle of periodic sending of state information from the receivers to the transmitter to avoid complex error recovery procedures [NRS90] Finally we use a selective repeat retransmission scheme to achieve high throughput. In this ....

B. Rajagopalan, "Reliability and Scaling Issues in Multicast Communication," Proceedings of ACMSIGCOMM '92, Pages 188-198, September 1992.

....These range from introducing reliable multicast at the network layer, to building new transport layer protocols which support reliable multicast, to combining the functionality of both the transport and network layers into a new layer. Reliability at the network layer has been introduced in RM [3], ST II [4] and URGC [5] RM offers mechanisms for partial multicast and supports reverse communication from group members to source by a gather operation. It involves constructing and maintaining trees for each group. ST II guarantees end to end delay and bandwidth, and involves reserving ....

B. Rajagopalan, "Reliability and scaling issues in multicast communication", Proceedings of ACM SIGCOMM, pp. 188-198, 1992.

....group decision making when all the users subscribe simultaneously. In [BFC93] different (TCP IP packet switched) network environments is assumed, without explicit real time guarantees. A more comprehensive survey of existing statistical models and other published work, such as [BFC93, CESZ91, Raj92] will be provided in the full version of the paper. 2 The model To simplify, we are considering here the case of permanent subscription, or, equivalently, case in which all subscriptions are for identical slots of time, say, 1 2 hour slot for ABC news from 7 till 7.30 p.m. and the customers can ....

Bala Rajagopalan. Reliability and scaling issues in multicast communication. In Proc. of the Annual ACM SIGCOMM Symposium on Communication Architectures and Protocols, Baltimore, MD, pages 188--197, August 1992.

....environment in which the corresponding multicast protocol operates. For instance, the protocol given in [7] assumes a broadcast environment and the protocol developed in [3] does not tolerate the partitioning of the network. Similarly, of the few multicast transport schemes that have been proposed [6,11,12,13,14], none consider the group membership management in wide area networks in any detail. Recently, some work has begun to appear concerning this issue in detail [15,16] In [15] the authors consider a protocol that maintains membership information at each member in the presence of topological and ....

B. Rajagopalan "Reliability and Scaling Issues in Multicast Communication," Proc. SIGCOMM '92, pp 188-198, August, 1992.

No context found.

Rajagopalan, B. Reliability and scaling issues in multicast communication. In Proc. ACM SIGCOMM (1992), pp. 188--197.

No context found.

B. Rajagopalan, "Reliability and Scaling Issues in Multicast Communication," Proceedings of SIGCOMM'92, Aug. 1992.

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