S. Hotz. Routing Information Organization to Support Scalable Interdomain Routing with Heterogeneous Path Requirements. PhD thesis, Univ. of Southern California, 1994.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....overlay latency, hop count distance, and overlay bandwidth at the same time. Network measurements in terms of finding nearby hosts have been examined in the literature. Recent studies have focused on scalable estimation schemes that do not requiring all pair latency tests, including Hotz [18], IDMaps [15] GNP [24] and Binning [29] All these techniques require certain infrastructural support or established landmark hosts. While Saxons network measurement component can utilize any of the existing techniques, we also introduce a light weight random sampling approach to locate nearby ....

S. Hotz. Routing Information Organization to Support Scalable Routing with Heterogeneous Path Requirements. PhD thesis, Dept. of Computer Science, University of Southern California, 1994.

....prob See http: www.stardust.com cdn Our work is not related to the Tiers topology generator by M. Doar. lem. However, it requires Internet wide deployment of the measurement entities. Techniques that require a very limited infrastructure support include a triangulation method, due to Hotz [10], and its weighted variant [9] a distributed binning technique [15] and Beaconing [12] These techniques use a small set of measurement reference points in the network called landmarks [15] or beacons [12] Distances between each application peer and these beacons are measured, and are ....

....peers on the Internet. They classify existing techniques into reactive gathering and proactive gathering categories. Expanding ring searches are classified under the reactive category. Our work can be classified under probing based schemes along with the triangulation based approaches due to Hotz [10] and the weighted variant [9] The beaconing technique [12] also uses similar distance estimation probes to find the nearest peer. However, all these other schemes require incur significantly higher traffic overhead in comparison to the Tiers approach. A completely different approach to finding ....

S. Hotz. Routing information organization to support scalable interdomain routing with heterogeneous path requirements. In PhD thesis, University of Southern California, 1996.

....3161 100 TABLE II BELL LABS WEB CLIENT SET BREAKDOWN we estimate the distance between two clients as the sum of the distances from each client to the closest Traceroute Gateway, and the distance between the two Traceroute Gateways. This method is usually called triangulation in the literature [12], 13] In [14] the authors evaluated its efficacy on estimating distance between two points on the Internet. IV. EXPERIMENT RESULTS Recall from Section III A, in all of our simulations, we use a network of 3,037 nodes, of which Itl 50 are selected as candidate hosts. The choice of which ....

S. Hotz, "Routing information organization to support scalable interdomain routing with hererogenous path requirements," Tech. Rep. Ph.D. Thesis, Univ. of Southern California, CS Dept., 1994.

....CNN content) Name servers might then use the above scheme to select a server for the requesting client. We compare the performance of our binning based server selection to 3 schemes: Random: A client selects one of all available servers at random. Selection using the Hotz metric: Hotz [30], 31] like our binning scheme, uses RTT measurements from a node to a set of well known landmarks to estimate inter node distances. The Hotz metric is computed as follows: Let # # represent the distance from a node # to landmark # . Then for any two nodes # and # and Landmark # , the distance ....

....For server selection, a client selects the server to which its estimated distance is minimum. Note that one advantage in general (not necessarily for server selection) of using bins as a metric rather than the Hotz or Cartesian distance is that the latter metrics require more information. In [30], Hotz defines the metric we describe but does not apply his work to the server selection problem. This application of the Hotz metric is our interpretation of how it might be used for server selection. 0 7803 7476 2 02 17.00 (c) 2002 IEEE. Topology Hotz Crtsn Bin Rand TS 10K 3.50 2.35 2.40 ....

S. Hotz, "Routing information organization to support scalable routing with heterogeneous path requirements," Tech. Rep. PhD thesis (draft), University of Southern California, 1994.

....of the addresses. 1.2 Overview of our approach We introduce a new technique called Beaconing. A set of application level processes in the network act as beacons and we find peers using mutual distance measurements to these beacons. Our solution most closely resembles the triangulation [12] and weighted triangulationapproaches [11] Like these approaches, our scheme does not require any changes to the infrastructure and works over only IP unicast; however, as described later in this paper, our overall approach is quite different comprared to the triangulation approaches. In our ....

....node w. When n decides to join the application, w chooses some node i 2 A uniformly at random as the nearest peer for n. This scheme has extremely low run time overhead, but also incurs a high error rate. 2.3. 2 Triangulation and Weighted Triangulation The triangulation method, due to Hotz [12], also computes distances using a set of measurement points called beacons. Given a host i in A and k beacons B 0 ; B k Gamma1 , the triangulation method described by Hotz [12] defines a n tuple D i = hdist(i; B 0 ) dist(i; B 1 ) dist(n; B k Gamma1 )i . A similar tuple Dn = ....

[Article contains additional citation context not shown here]

S. Hotz. Routing Information Organization To Support Scalable Interdomain Routing with Heterogeneous Path Requirements. PhD thesis, University of Southern California, 1996.

....(i.e. ### # of data) Thus, this approach is able to trade local computations for significantly reduced communication overhead, achieving higher scalability. We study two types of coordinates for distance prediction. The first is a kind of relative coordinates, originally proposed by Hotz [6] to construct the triangulated heuristic. Hotz s goal was to apply this heuristic in the # heuristic search algorithm to reduce the computation overhead of shortest path searches in interdomain graphs. The potential of this heuristic for network distance prediction has not been previously ....

S.M. Hotz, "Routing information organization to support scalable interdomain routing with heterogeneous path requirements," 1994, Ph.D. Thesis (draft), University of Southern California.

....(i.e. O(K 2 ) of data) Thus, this approach is able to trade local computations for significantly reduced communication overhead, achieving higher scalability. We study two types of coordinates for distance prediction. The first is a kind of relative coordinates, originally proposed by Hotz [6] to construct the triangulated heuristic. Hotz s goal was to apply this heuristic in the A heuristic search algorithm to reduce the computation overhead of shortest path searches in interdomain graphs. The potential of this heuristic for network distance prediction has not been previously ....

S.M. Hotz, "Routing information organization to support scalable interdomain routing with heterogeneous path requirements," 1994, Ph.D. Thesis (draft), University of Southern California.

.... (In these experiments, beaconing does not nd the nearest server all the time because we did not enforce the two correctness assumptions) Lastly, our preliminary studies also show that beaconing more accurate and incurs orders of magnitude less overhead than techniques previously reported [23, 13]. Optimizations and re nements The protocol, as described above, requires all k beacon nodes to be available at all times and for each beacon node to maintain state about every server. Further, the set of beacon nodes is xed a priori this can lead to a bottleneck if the number of servers ....

S. Hotz. Routing information organization to support scalable interdomain routing with heterogeneouspath requirements, 1996.

....3161 100 TABLE II BELL LABS WEB CLIENT SET BREAKDOWN we estimate the distance between two clients as the sum of the distances from each client to the closest Traceroute Gateway, and the distance between the two Traceroute Gateways. This method is usually called triangulation in the literature [12], 13] In [14] the authors evaluated its efficacy on estimating distance between two points on the Internet. IV. EXPERIMENT RESULTS Recall from Section III A, in all of our simulations, we use a network of 3,037 nodes, of which jHj = 50 are selected as candidate hosts. The choice of which host ....

S. Hotz, "Routing information organization to support scalable interdomain routing with heterogenous path requirements," Tech. Rep. Ph.D. Thesis, Univ. of Southern California, CS Dept., 1994.

....that is bounded above by , and below by . If either of the two distances is small relative to the other, the bound is tight and the estimate accurate. Deriving a distance estimate from this bound has been referred to as triangulation [34, 35]. A key point to keep in mind is that any time we estimate a distance from to based on distances to an intermediary , we are to some degree relying on what we will term efficient routing: that Internet routing does indeed strive to find low latency paths, and that the routes used by ....

S. Hotz, "Routing Information Organization to Support Scalable Interdomain Routing with Heterogenous Path Requirements," Tech. Rep. Ph.D. Thesis, Univ. of Southern California, CS Dept., 1994.

....3161 100 TABLE II BELL LABS WEB CLIENT SET BREAKDOWN we estimate the distance between two clients as the sum of the distances from each client to the closest Traceroute Gateway, and the distance between the two Traceroute Gateways. This method is usually called triangulation in the literature [12], 13] In [14] the authors evaluated its efficacy on estimating distance between two points on the Internet. IV. EXPERIMENT RESULTS Recall from Section III A, in all of our simulations, we use a network of 3,037 nodes, of which z z are selected as candidate hosts. The choice of which ....

S. Hotz, "Routing information organization to support scalable interdomain routing with heterogenous path requirements," Tech. Rep. Ph.D. Thesis, Univ. of Southern California, CS Dept., 1994.

....triangle inequality we have that (a; c) is bounded above by (a; b) b; c) and below by j(a; b) b; c)j. If either of the two distances is small relative to the other, the bound is tight and the estimate accurate. Deriving a distance estimate from this bound has been referred to as triangulation [34, 35]. A key point to keep in mind is that any time we estimate a distance from a to c based on distances to an intermediary b, we are to some degree relying on what we will term efficient routing: that Internet routing does indeed strive to find low latency paths, and that the routes used by two ....

S. Hotz, "Routing Information Organization to Support Scalable Interdomain Routing with Heterogenous Path Requirements," Tech. Rep. Ph.D. Thesis, Univ. of Southern California, CS Dept., 1994.

....distribution of prefix availability and steadiness is likely to be similar at other backbone providers. 5 Related Work To our knowledge, no prior work has attempted to characterize the Internet inter domain topology and its growth. Several researchers have proposed inter domain topology models [3, 11, 16]. These models posit a three or four level provider hierarchy. Some of these models also assume that domains in a class primarily connect to domains in classes above or below their own. Our analysis confirms the existence of a four level hierarchy. However, our classification exposes a ....

S. M. Hotz. Routing Information Organization To Support Scalable Interdomain Routing With HeterogeneousPath Requirements. PhD thesis, University of Southern California, 1996.

....we have that (a; c) is bounded above by (a; b) b; c) and below by j(a; b) Gamma (b; c)j. If either of the two distances is small relative to the other, then the bound is tight and the estimate accurate. Deriving a distance estimate from this bound has been referred to as triangulation [16, 17]. A key point to keep in mind is that any time we estimate a distance from a to c based on distances to an intermediary b, then we are in fact making an assumption of what we will term efficient routing: that Internet routing does indeed strive to find low latency paths, and that the routes used ....

S. Hotz, "Routing information organization to support scalable interdomain routing with heterogenous path requirements," Tech. Rep. Ph.D. Thesis, Univ. of Southern California, CS Dept., 1994.

....distribution of prefix availability and steadiness is likely to be similar at other backbone providers. 6 Related Work To our knowledge, no prior work has attempted to characterize the Internet inter domain topology and its growth. Several researchers have proposed inter domain topology models [3, 12, 17]. These models posit a three or four level provider hierarchy. Some of these models also assume that domains in a class primarily connect to domains in classes above or below their own. Our analysis confirms the existence of a four level hierarchy. However, our classification exposes a ....

S. M. Hotz. Routing Information Organization To Support Scalable Interdomain Routing With Heterogeneous Path Requirements. PhD thesis, University of Southern California, 1996.

....for their popular WWW server, and by Harvest Brokers [4] Another simple strategy is to configure routing topologies manually. While labor intensive, this approach is widely used, for example in the MBONE [6] and USENET [14] The most closely related work to the current paper is Hotz PhD thesis [12], in which he examines network triangulation as a possible approach to routing and server location. We use some of Hotz definitions and results in the current paper. Another very relevant effort is Danzig et al. s floodd replication system, which uses network probes to compute logical update ....

.... 1: Internet Hop Count Statistics We also need to define a way to compute distances when only partial routing knowledge is available (e.g. allowing a distance to be inferred between a client and server given measurements from a measurement beacon to each) For this purpose we use Hotz definitions [12]. Hotz defines an N tuple s 1 ; s 2 ; s 3 ; s N as the distance to a node from each of N measurement beacons. He then defines the distance between a server at coordinates S = s 1 ; s 2 ; s 3 ; s N and a client at C = c 1 ; c 2 ; c 3 ; c N with the AV G function: AV G(S; ....

[Article contains additional citation context not shown here]

Steven Michael Hotz. Routing information organization to support scalable interdomain routing with heterogeneous path requirements. Technical report, Computer Science Department, University of Southern California, Los Angeles, California, September 1994 (Draft). PhD Thesis.

....A number of proposals have been made examining replica selection in the Internet, both in general and for web applications. This section brie y reviews this work. Guyton and Schwartz examine the cost of locating servers through several schemes [8] including forms of broadcast and triangularization [10]. They compare hop count and round trip packet latency, choosing weighted hop count because of poor clock resolution and RTT variability. They assume an active client and, in some cases, additional servers (beacons) deployed throughout the network. Their work focuses mainly on network bandwidth ....

Steven Michael Hotz. Routing Information Organization to Support Scalable Interdomain Routing with Heterogeneous Path Requirements. PhD thesis, USC, 1996.

....global distribution of topology. Alaettinoglu explored route construction using aggregation and hierarchical heuristics for querying remote aggregates for more detailed information [Ala94] Hotz has studied route construction heuristics based on route fragments and triangulation of graph position [Hot94] In contrast to the above work, QoS routing protocols globally distribute topology and group information. A large body of QoS multicast routing literature uses Steiner trees to optimize tree cost for small, static groups [BKJ83, Wax88, Cho91, KPP92] Salama et al. compare these and other ....

Steven Hotz. "Routing Information Organization to Support Scalable Interdomain Routing with Heterogeneous Path Requirements". PhD thesis, University of Southern California, 1994.

No context found.

S. Hotz. Routing Information Organization to Support Scalable Interdomain Routing with Heterogeneous Path Requirements. PhD thesis, Univ. of Southern California, 1994.

No context found.

S. Hotz, Routing information organization to support scalable interdomain routing with heterogeneous path requirements, Ph.d. thesis, Univ. of Southern California, 1994.

No context found.

S. Hotz. Routing information organization to support scalable interdomain routing with heterogeneous path requirements, 1996.

No context found.

S. Hotz, "Routing information organization to support scalable routing with heterogeneous path requirements," Tech. Rep. PhD thesis (draft), University of Southern California, 1994.

No context found.

Hotz, Steven Michael, Routing Information Organization to Support Scalable Inter-Domain Routing with Heterogeneous Path Requirements, Technical Report, Computer Science Department, University of Southern California, Los Angeles, 1994.

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