N. G. Duffield, J. Horowitz, F. Lo Presti, and D. Towsley. Multicast topology inference from measured end-to-end loss. IEEE Trans. on Information Theory, 48:26--45, Jan. 2002.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....information through a network to a set of receivers denoted by R. Assume that the routes from the sender to the receivers are fixed. The problem we address is the identification of the network topology based on end to end measurements that measure the degree of correlation between receivers [1, 2, 3, 4, 5]. With this limited information, it is only possible to identify the so called logical topology defined by the branching points between paths to different receivers. This corresponds to a tree structured topology with the sender at the root and the receivers at the leaves as depicted in Figure ....

....18;19 will be strictly greater than i;19 for all i 2 R=f18; 19g, revealing that receivers 18 and 19 have a common parent in the logical tree. The property can be exploited in this manner to devise a simple and effective bottom up merging algorithms that identify the full, logical topology [1, 2, 3, 4]. Metrics possessing the Monotonicity Property can be estimated from a number of different end to end measurements including counts of losses, counts of zero delay events (utilization) and delay correlations [1, 2, 3, 4] These estimated metrics, denoted fx i;j g, can be interpreted as ....

[Article contains additional citation context not shown here]

N.G. Duffield, J. Horowitz, F. Lo Presti, and D. Towsley, "Multicast topology inference from measured end-to-end loss," to appear in IEEE Trans. Information Theory, 2002.

....aimed at tracking nonstationary network behavior which involve analogs of classical Kalman ltering methods [34, 26] Another variation on the basic problem (1) is obtained by assuming that the routing matrix A is not known precisely. This leads to the so called topology identi cation problem [30, 40, 41, 42, 43, 44, 45], and is somewhat akin to blind deconvolution or system identi cation problems. 3 Link Level Network Inference Link level network tomography is the estimation of link level network parameters (loss rates, delay distributions) from path level measurements. Link level parameters can be estimated ....

....of their paths) the larger the covariance between the two. Metrics possessing this type of monotonicity property can be estimated from a number of di erent end to end measurements including counts of losses, counts of zero delay events (utilization) delay correlations, and delay di erences [30, 40, 41, 42, 43, 45, 44]. Using such metrics, topology identi cation can be cast as a Maximum Likelihood estimation problem as follows. The estimated metrics x fx i;j g, where the indices i; j refer to di erent pairs of receivers, can be interpreted as observations of the true metric values f i;j g contaminated ....

N. G. Dueld, J. Horowitz, F. Lo Presti, and D. Towsley. Multicast topology inference from measured end-to-end loss. IEEE Trans. Info. Theory, 48(1):26-45, January 2002.

....aimed at tracking nonstationary network behavior which involve analogs of classical Kalman ltering methods [34, 26] Another variation on the basic problem (1) is obtained by assuming that the routing matrix A is not known precisely. This leads to the so called topology identi cation problem [30, 40, 41, 42, 43, 44, 45], and is somewhat akin to blind deconvolution or system identi cation problems. 3 Link Level Network Inference Link level network tomography is the estimation of link level network parameters (loss rates, delay distributions) from path level measurements. Link level parameters can be estimated ....

....and proprietary concerns increase. For situations in which common tools such as traceroute are not applicable, a number of methods have been proposed for the identi cation of network (routing) topology based on end18 to end measurements that measure the degree of correlation between receivers [30, 40, 41, 43, 44, 45]. Most of these approaches have concentrated on identifying the tree structured topology connecting a single sender to multiple receivers. It is assumed that the routes from the sender to the receiver are xed. With only end to end measurements, it is only possible to identify the logical topology ....

[Article contains additional citation context not shown here]

N.G. Dueld, J. Horowitz, F. Lo Presti, and D. Towsley. Multicast topology inference from measured end-to-end loss. IEEE Trans. Information Theory, 2002.

....packet stream can use such information to locate fellow participants that are well situated to replace packets that have been lost and that cannot otherwise be reconstituted. Topology and loss rate can be estimated from end to end data using minc (multicast estimation of network characteristics) [15, 13, 11, 28, 27, 30, 14, 1]. However, the loss traces required for minc inference can consume significant bandwidth in comparison to the data tra#c upon which they are based if each receiver must send its entire loss trace. In this dissertation we study ways to either thin the traces (Chapter 2) or to obtain useful results ....

....al. s [11] November 1999 Transactions on Information Theory article. Loss based inference of multicast network topology is described in Caceres et al. s [11] December 1999 CDC paper and in Du#eld et al. s [28] September 2000 IP Tra#c Measurement, Modelling, and Management paper. Du#eld et al. s [27] forthcoming Transactions on Information Theory article provides a detailed evaluation of three di#erent topology inference techniques. 128 There have been extensions to the loss based work. Inference in the face of missing data is described in Du#eld et al. s [30] forthcoming J SAC article. ....

Du#eld, N.G., Horowitz, J., Lo Presti, F., and Towsley, D. Multicast topology inference from measured end-to-end loss. IEEE Transactions in Information Theory . To appear.

.... estimators have been developed to estimate the characteristics of the logical links of the multicast tree from end to end measurements between the source and the receivers, in particular packet loss rates [3] delay distributions [17] delay moments [12] and even the underlying logical topology [10]. Using clusters of unicast packets to emulate multicast packets, it is possible to draw the same type of inference from unicast measurements; see [7] 13] Realization of MINC in the Internet must address two issues: the availability of participating end hosts, and the transmission of ....

....portion could lead to the type of correlations allowed for in the model. In the work reported here the focus is on estimating the link loss rates for a known topology. However, the techniques can be extended to infer multicast topology, at least using data from probes with complete reports, as in [10]. We remark that, under the assumptions of independent probe loss both between different probes, and across different links the estimators in [3] are consistent, i.e. they converge to the true values as the number of probes grows. In [3] it is shown that the presence of temporal correlations ....

N.G. Duffield, J. Horowitz, F. Lo Presti, D. Towsley, "Multicast Topology Inference from Measured End-to-End Loss", IEEE Trans. on Information Theory, vol. 48, pp. 26--45, 2002.

.... estimators have been developed to estimate the characteristics of the logical links of the multicast tree from end to end measurements between the source and the receivers, in particular packet loss rates [3] delay distributions [17] delay moments [12] and even the underlying logical topology [10]. Using clusters of unicast packets to emulate multicast packets, it is possible to draw the same type of inference from unicast measurements; see [7] 13] Realization of MINC in the Internet must address two issues: the availability of participating end hosts, and the transmission of ....

....portion could lead to the type of correlations allowed for in the model. In the work reported here the focus is on estimating the link loss rates for a known topology. However, the techniques can be extended to infer multicast topology, at least using data from probes with complete reports, as in [10]. Having outlined the assumptions and workings behind the inference engine, we now proceed to the main topic of the paper, namely, the architecture by which loss reports are conveyed from the receivers to the inference engine. III. ARCHITECTURE In the MINC architecture there are three distinct ....

N.G. Duffield, J. Horowitz, F. Lo Presti, D. Towsley, "Multicast Topology Inference from Measured End-to-End Loss", IEEE Trans. on Information Theory, to appear.

....that as the number of packets grows multicast receivers sharing a longer portion of the multicast distribution tree also have higher shared loss rates; this information could in turn be used to reconstruct the topology by recursively grouping the pair of nodes with the highest shared loss. In [8] the correctness this algorithm was proven and the approach was extended to general topologies by introducing several other loss based algorithms. More recently, algorithms have been proposed for identifying multicast topologies based on delay measurements instead. By observing that the approach ....

....this algorithm was proven and the approach was extended to general topologies by introducing several other loss based algorithms. More recently, algorithms have been proposed for identifying multicast topologies based on delay measurements instead. By observing that the approach in [16] and [8] can be generalized to any performance measures that (i) monotonically increases as the packet traverse the tree, and (ii) can be estimated on the sole basis of end to end measurements at the receivers, in [7] several algorithms are specified based on delay performance measures as link ....

[Article contains additional citation context not shown here]

N.G. Duffield, J.Horowitz, F. Lo Presti and D. Towsley, "Multicast Topology Inference from Measured End-to-End Loss", submitted for publication.

No context found.

N. G. Duffield, J. Horowitz, F. Lo Presti, and D. Towsley. Multicast topology inference from measured end-to-end loss. IEEE Trans. on Information Theory, 48:26--45, Jan. 2002.

No context found.

N. G. Duffield, J. Horowitz, F. L. Presti and D. Towsley, "Multicast topology inference from measured end-to-end loss," in IEEE Trans. Inform. Theory, to appear.

No context found.

N. Duffield, J. Horowitz, F. Lo Presti, and D. Towsley, "Multicast topology inference from measured end-to-end loss," to appear in IEEE Transaction in Information Theory.

No context found.

N. Duffield, J. Horowitz, F. Lo Presti, and D. Towsley, "Multicast topology inference from measured end-to-end measurements," in ITC Seminar on IP Traffic, Measurement, and Modeling, Monterey, CA, September 2000.

No context found.

N. Duffield, J. Horowitz, F. L. Presti, and D. Towsley. Multicast Topology Inference From Measured End-to-End Loss. IEEE Trans. on Information Theory, 48:26--45, 2002.

No context found.

N. Duffield, J. Horowitz, F. L. Presti, and D. Towsley. Multicast Topology Inference From Measured End-to-End Measurements. In ITC Seminar on IP Traffic, Measurement and Modelling, September 2000.

No context found.

N. Duffield, J. Horowitz, F. L. Presti, and D. Towsley, "Multicast topology inference from measured end-to-end loss," IEEE Trans. Info. Theory, vol. 48, no. 1, pp. 26--45, January 2002.

No context found.

N.G. Duffield, J. Horowitz, F. Lo Presti, and D. Towsley, "Multicast topology inference from measured end-to-end loss," IEEE Trans. Info. Theory, vol.48, pp.26-45, Jan. 2002.

No context found.

N.G. Duffield, J. Horowitz, F. Lo Presti, and D. Towsley, "Multicast topology inference from mea- sured end-to-end loss," IEEE Trans. Info. Theory, vol. 48, no. 1, pp. 26--45, January 2002.

No context found.

N. Duffield, J. Horowitz, F. LoPresti, and D. Towsley. Multicast topology inference from measured end-to-end loss. IEEE Transactions on Information Theory, 48(1):26--45, 2002.

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