Results 1 
8 of
8
Robust distributed estimation using the embedded subgraphs algorithm
 IEEE Trans. Signal Process
, 2006
"... Abstract—We propose a new iterative, distributed approach for linear minimum meansquareerror (LMMSE) estimation in graphical models with cycles. The embedded subgraphs algorithm (ESA) decomposes a loopy graphical model into a number of linked embedded subgraphs and applies the classical parallel b ..."
Abstract

Cited by 11 (0 self)
 Add to MetaCart
Abstract—We propose a new iterative, distributed approach for linear minimum meansquareerror (LMMSE) estimation in graphical models with cycles. The embedded subgraphs algorithm (ESA) decomposes a loopy graphical model into a number of linked embedded subgraphs and applies the classical parallel block Jacobi iteration comprising local LMMSE estimation in each subgraph (involving inversion of a small matrix) followed by an information exchange between neighboring nodes and subgraphs. Our primary application is sensor networks, where the model encodes the correlation structure of the sensor measurements, which are assumed to be Gaussian. The resulting LMMSE estimation problem involves a large matrix inverse, which must be solved innetwork with distributed computation and minimal intersensor communication. By invoking the theory of asynchronous iterations, we prove that ESA is robust to temporary communication faults such as failing links and sleeping nodes, and enjoys guaranteed convergence under relatively mild conditions. Simulation studies demonstrate that ESA compares favorably with other recently proposed algorithms for distributed estimation. Simulations also indicate that energy consumption for iterative estimation increases substantially as more links fail or nodes sleep. Thus, somewhat surprisingly, sensor network energy conservation strategies such as lowpowered transmission and aggressive sleep schedules could actually prove counterproductive. Our results can be replicated using MATLAB code from www.dsp.rice.edu/software. Index Terms—Asynchronous iterations, distributed estimation, graphical models, matrix splitting, sensor networks, Wiener filter. I.
Computing A DiameterConstrained Minimum Spanning Tree
, 2001
"... In numerous practical applications, it is necessary to find the smallest possible tree with a bounded diameter. A diameterconstrained minimum spanning tree (DCMST) of a given undirected, edgeweighted graph, G, is the smallestweight spanning tree of all spanning trees of G which contain no path wi ..."
Abstract

Cited by 8 (0 self)
 Add to MetaCart
In numerous practical applications, it is necessary to find the smallest possible tree with a bounded diameter. A diameterconstrained minimum spanning tree (DCMST) of a given undirected, edgeweighted graph, G, is the smallestweight spanning tree of all spanning trees of G which contain no path with more than k edges, where k is a given positive integer. The problem of finding a DCMST is NPcomplete for all values of k; 4 k (n  2), except when all edgeweights are identical. A DCMST is essential for the efficiency of various distributed mutual exclusion algorithms, where it can minimize the number of messages communicated among processors per critical section. It is also useful in linear lightwave networks, where it can minimize interference in the network by limiting the traffic in the network lines. Another practical application requiring a DCMST arises in data compression, where some algorithms compress a file utilizing a tree datastructure, and decompress a path in the tree to access a record. A DCMST helps such algorithms to be fast without sacrificing a lot of storage space. We present a survey of the literature on the DCMST problem, study the expected diameter of a random labeled tree, and present five new polynomialtime algorithms for an approximate DCMST. One of our new algorithms constructs an approximate DCMST in a modified greedy fashion, employing a heuristic for selecting an edge to be added to iii the tree in each stage of the construction. Three other new algorithms start with an unconstrained minimum spanning tree, and iteratively refine it into an approximate DCMST. We also present an algorithm designed for the special case when the diameter is required to be no more than 4. Such a diameter4 tree is also used for evaluating the quality of o...
Evaluating steiner tree heuristics and diameter variations for application layer multicast,”AcceptedforpublicationinComputerNetworksonComplexComputerandCommunicationNetworks
, 2008
"... Latency reduction in distributed interactive applications has been studied intensively. Such applications may have stringent latency requirements and dynamic user groups. We focus on applicationlayer multicast with a centralized approach to the group management. The groups are organized in overlay ..."
Abstract

Cited by 2 (2 self)
 Add to MetaCart
Latency reduction in distributed interactive applications has been studied intensively. Such applications may have stringent latency requirements and dynamic user groups. We focus on applicationlayer multicast with a centralized approach to the group management. The groups are organized in overlay networks that are created using graph algorithms that create a tree structure for the group. A tree has no cycles and uses a small routing table, as opposed to a connected overlay mesh. We investigate a group of spanning tree problems that are referred to as Steiner tree problems, and we have a particular focus on reducing the diameter of a tree, which is the maximum pairwise latency in a tree. In addition, we focus on reducing the time it takes to execute membership changes. In that context, we use coreselection heuristics to find wellplaced client nodes, and edgepruning algorithms to reduce the number of edges in an otherwise fully meshed overlay. Our edgepruning algorithms strongly connect wellplaced client nodes to the remaining group members, to create new and pruned group graphs. Consequently, when a tree algorithm is applied to a pruned group graph, it is manipulated into creating trees with a smaller diameter. We devised new Steinertree heuristics that reduced the diameter, and also proposed new edgepruning algorithms to make the tree construction faster. These heuristics and algorithms were implemented and analyzed experimentally along with several spanningtree and coreselection heuristics found in the literature. We found that a fullmesh of shortest paths makes it difficult for Steinertree heuristics to find better trees than spanning tree algorithms. The network seen from the application layer is in fact a full mesh of shortest paths. In addition, we found that faster Steinertree heuristics that do not explicitly optimize the diameter are able to compete with slower heuristics that do optimize it.
Constructing LowLatency Overlay Networks: Tree vs. Mesh Algorithms
"... Abstract—Distributed interactive applications may have stringent latency requirements and dynamic user groups. These applications may benefit from a group communication system, and to improve the system support for such applications, we investigate graph algorithms that construct lowlatency overlay ..."
Abstract

Cited by 1 (0 self)
 Add to MetaCart
Abstract—Distributed interactive applications may have stringent latency requirements and dynamic user groups. These applications may benefit from a group communication system, and to improve the system support for such applications, we investigate graph algorithms that construct lowlatency overlay networks for applicationlayer multicast. In particular, we focus on reducing the diameter and the pairwise latencies in the overlay. The overlay construction time is also considered, as it is often timedependent in our dynamic target applications. Here, we have implemented and experimentally analyzed spanningtree heuristics and mesh construction heuristics, and compared their performance and applicability to distributed interactive applications. We found that trees are faster to construct and save considerable amounts of resources in the network. Meshes, on the other hand, yield lower pairwise latencies and increases the fault tolerance, but at the expense of increased resource consumption. I.
Efficient Execution Plans for Distributed Skyline Query Processing
"... In this paper, we study the generation of efficient execution plans for skyline query processing in largescale distributed environments. In such a setting, each server stores autonomously a fraction of the data, thus all servers need to process the skyline query. An execution plan defines the order ..."
Abstract

Cited by 1 (0 self)
 Add to MetaCart
In this paper, we study the generation of efficient execution plans for skyline query processing in largescale distributed environments. In such a setting, each server stores autonomously a fraction of the data, thus all servers need to process the skyline query. An execution plan defines the order in which the individual skyline queries are processed on different servers, and influences the performance of query processing. Querying servers consecutively reduces the amount of transferred data and the number of queried servers, since skyline points obtained by one server prune points in the subsequent servers, but also increases the latency of the system. To address this tradeoff,weintroducea novel framework, called SkyPlan, for processing distributed skyline queries that generates execution plans aiming at optimizing the performance of query processing. Thus, we quantify the gain of querying consecutively different servers. Then, execution plans are generated that maximize the overall gain, while also taking into account additional objectives, such as bounding the maximum number of hops required for the query or balancing the load on different servers fairly. Finally, we present an algorithm for distributed processing based on the generated plan that continuously refines the execution plan during innetwork processing. Our framework consistently outperforms the stateoftheart algorithm.
OntheInfluenceofLatencyEstimationonDynamicGroup
"... Distributed interactive applications tend to have stringent latency requirements and some may have high bandwidth demands. Many of them have also very dynamic user groups for which alltoall communication is needed. In online multiplayer games, for example, such groups are determined through region ..."
Abstract
 Add to MetaCart
Distributed interactive applications tend to have stringent latency requirements and some may have high bandwidth demands. Many of them have also very dynamic user groups for which alltoall communication is needed. In online multiplayer games, for example, such groups are determined through regionofinterest management in the application. We have investigated a variety of group management approaches for overlay networks in earlier work and shown that several useful tree heuristics exist. However, these heuristics require full knowledge of all overlay link latencies. Since this is not scalable, we investigate the effects that latency estimation techniques have on the quality of overlay tree constructions. We do this by evaluating one example of our group management approaches in Planetlab and examining how latency estimation techniques influence their quality. Specifically, we investigate how two wellknown latency estimation techniques, Vivaldi andNetvigator, affect the quality of tree building. Many types of distributed interactive applications are popular today. Examples include audio/video conferencing, online games, virtual museums and shopping malls, etc. In these applications, the media types may range from text to continuous media, each which must be delivered with stringent latency bounds while satisfying some bandwidth requirements. Especially, the interactivity imposes demanding restrictions on network latency. Thus, to provide a satisfactory user service,
Contents lists available at ScienceDirect Computer Networks
"... journal homepage: www.elsevier.com/locate/comnet Evaluating Steinertree heuristics and diameter variations ..."
Abstract
 Add to MetaCart
journal homepage: www.elsevier.com/locate/comnet Evaluating Steinertree heuristics and diameter variations