Downloads & Free Reading Options - Results

We present two algorithms in the Quantum CONGEST- CLIQUE model of distributed computation that succeed with high probability; one for producing an approximately optimal Steiner Tree, and one for producing an exact directed minimum spanning tree, each of which uses O ̃(n1/4) rounds of communication and O ̃(n9/4) messages, achieving a lower asymptotic round and message complexity than any known algorithms in the classical CONGEST-CLIQUE model. At a high level, we achieve these results by combining classical algorithms with fast quantum subroutines. Additionally, we characterize the constants and logarithmic factors involved in our algorithms, as well as related classical algorithms, revealing that advances are needed to render both practical.

This paper is concerned with distributed computation of several commonly used centrality measures in complex networks. In particular, we propose deterministic algorithms, which converge in finite time, for the distributed computation of the degree, closeness and betweenness centrality measures in directed graphs. Regarding eigenvector centrality, we consider the PageRank problem as its typical variant, and design distributed randomized algorithms to compute PageRank for both fixed and time-varying graphs. A key feature of the proposed algorithms is that they do not require to know the network size, which can be simultaneously estimated at every node, and that they are clock-free. To address the PageRank problem of time-varying graphs, we introduce the novel concept of persistent graph, which eliminates the effect of spamming nodes. Moreover, we prove that these algorithms converge almost surely and in the sense of $L^p$. Finally, the effectiveness of the proposed algorithms is illustrated via extensive simulations using a classical benchmark.

The project seeks to develop methods to more efficiently simulate aerospace vehicles. The goals are to reduce model development time, increase accuracy (e.g.,by allowing the integration of multidisciplinary models), facilitate collaboration by geographically- distributed groups of engineers, support uncertainty analysis and optimization, reduce hardware costs, and increase execution speeds. These problems are the subject of considerable contemporary research (e.g., Biedron et al. 1999; Heath and Dick, 2000).

The design of survivable algorithms requires a solid foundation for executing them. While hardware techniques for fault-tolerant computing are relatively well understood, fault-tolerant operating systems, as well as fault-tolerant applications (survivable algorithms), are, by contrast, little understood, and much more work in this field is required. We outline some of our work that contributes to the foundation of ultrareliable operating systems and fault-tolerant algorithm design. We introduce our consensus-based framework for fault-tolerant system design. This is followed by a description of a hierarchical partitioning method for efficient consensus. A scheduler for redundancy management is introduced, and application-specific fault tolerance is described. We give an overview of our hybrid algorithm technique, which is an alternative to the formal approach given.

A new approach is developed for solving the moving target detection and tracking problem using highly cluttered images. The unknown target is assumed to be moving over a cluttered background in the presence of foreground noise. Using a Markov random field model for the target and a probabilistic description of the noise, the posterior distribution of the target is a Gibbs distribution. The maximum aposteriori target image is found by a randomized search process. Both batch and recursive formulations are developed, with the recursive approach yielding superior results. Numerical results indicate that this approach can successfully detect and track small targets in environments where the target is essentially made invisible by noise. The algorithms are almost completely parallelizable: for n pixels a total of n/4 processors may be used, with the result that solutions would require on the order of 2 seconds on current machines for the examples presented. Keywords: Motion; Optical flow.

During the period of the grant, 6/15/95-6/30/98, we investigated problems in major areas: (1) Image Analysis Tasks, including image interpretation, object recognition, and tracking; (2) Distributed communication networks and distributed resource allocation. In the former area, we developed and explored three paradigms: hierarchical compositional models; deformable templates for a host of applications including medical tasks; and HMM/deformable templates for tracking and recognition of moving objects. In the later area, we developed a novel scheme for managing buffer overflows; a new framework for network security; and a mathematical framework for synthesizing distributed algorithms.

In this paper we propose distributed flooding-based storage algorithms for large-scale wireless sensor networks. Assume a wireless sensor network with $n$ nodes that have limited power, memory, and bandwidth. Each node is capable of both sensing and storing data. Such sensor nodes might disappear from the network due to failures or battery depletion. Hence it is desired to design efficient schemes to collect data from these $n$ nodes. We propose two distributed storage algorithms (DSA's) that utilize network flooding to solve this problem. In the first algorithm, DSA-I, we assume that every node utilizes network flooding to disseminate its data throughout the network using a mixing time of approximately O(n). We show that this algorithm is efficient in terms of the encoding and decoding operations. In the second algorithm, DSA-II, we assume that the total number of nodes is not known to every sensor; hence dissemination of the data does not depend on $n$. The encoding operations in this case take $O(C\mu^2)$, where $\mu$ is the mean degree of the network graph and $C$ is a system parameter. We evaluate the performance of the proposed algorithms through analysis and simulation, and show that their performance matches the derived theoretical results.

In this paper we propose distributed flooding-based storage algorithms for large-scale wireless sensor networks. Assume a wireless sensor network with $n$ nodes that have limited power, memory, and bandwidth. Each node is capable of both sensing and storing data. Such sensor nodes might disappear from the network due to failures or battery depletion. Hence it is desired to design efficient schemes to collect data from these $n$ nodes. We propose two distributed storage algorithms (DSA's) that utilize network flooding to solve this problem. In the first algorithm, DSA-I, we assume that every node utilizes network flooding to disseminate its data throughout the network using a mixing time of approximately O(n). We show that this algorithm is efficient in terms of the encoding and decoding operations. In the second algorithm, DSA-II, we assume that the total number of nodes is not known to every sensor; hence dissemination of the data does not depend on $n$. The encoding operations in this case take $O(C\mu^2)$, where $\mu$ is the mean degree of the network graph and $C$ is a system parameter. We evaluate the performance of the proposed algorithms through analysis and simulation, and show that their performance matches the derived theoretical results.

In this paper we propose distributed flooding-based storage algorithms for large-scale wireless sensor networks. Assume a wireless sensor network with $n$ nodes that have limited power, memory, and bandwidth. Each node is capable of both sensing and storing data. Such sensor nodes might disappear from the network due to failures or battery depletion. Hence it is desired to design efficient schemes to collect data from these $n$ nodes. We propose two distributed storage algorithms (DSA's) that utilize network flooding to solve this problem. In the first algorithm, DSA-I, we assume that every node utilizes network flooding to disseminate its data throughout the network using a mixing time of approximately O(n). We show that this algorithm is efficient in terms of the encoding and decoding operations. In the second algorithm, DSA-II, we assume that the total number of nodes is not known to every sensor; hence dissemination of the data does not depend on $n$. The encoding operations in this case take $O(C\mu^2)$, where $\mu$ is the mean degree of the network graph and $C$ is a system parameter. We evaluate the performance of the proposed algorithms through analysis and simulation, and show that their performance matches the derived theoretical results.

Distributed systems deadlock is similar to single-processor system deadlock, but is worse. It is harder to avoid, prevent or detect and is harder to cure, when it is tracked down because all the relevant information is scattered over many machines. In some systems, such as distributed database systems, it can be extremely serious, so it is important to understand how it differs from ordinary deadlock and what can be done about it. Two important deadlock prevention algorithms in distributed systems are wait-die and wound-wait. Their problem is that they just attend to the time stamp of processes, but not priority of them. In a real operating system, attending to priority of processes is very important. The proposed improved algorithms are attending to both priority and time stamp of processes.

The PageRank algorithm employed at Google assigns a measure of importance to each web page for rankings in search results. In our recent papers, we have proposed a distributed randomized approach for this algorithm, where web pages are treated as agents computing their own PageRank by communicating with linked pages. This paper builds upon this approach to reduce the computation and communication loads for the algorithms. In particular, we develop a method to systematically aggregate the web pages into groups by exploiting the sparsity inherent in the web. For each group, an aggregated PageRank value is computed, which can then be distributed among the group members. We provide a distributed update scheme for the aggregated PageRank along with an analysis on its convergence properties. The method is especially motivated by results on singular perturbation techniques for large-scale Markov chains and multi-agent consensus.

The PageRank algorithm employed at Google assigns a measure of importance to each web page for rankings in search results. In our recent papers, we have proposed a distributed randomized approach for this algorithm, where web pages are treated as agents computing their own PageRank by communicating with linked pages. This paper builds upon this approach to reduce the computation and communication loads for the algorithms. In particular, we develop a method to systematically aggregate the web pages into groups by exploiting the sparsity inherent in the web. For each group, an aggregated PageRank value is computed, which can then be distributed among the group members. We provide a distributed update scheme for the aggregated PageRank along with an analysis on its convergence properties. The method is especially motivated by results on singular perturbation techniques for large-scale Markov chains and multi-agent consensus.

The PageRank algorithm employed at Google assigns a measure of importance to each web page for rankings in search results. In our recent papers, we have proposed a distributed randomized approach for this algorithm, where web pages are treated as agents computing their own PageRank by communicating with linked pages. This paper builds upon this approach to reduce the computation and communication loads for the algorithms. In particular, we develop a method to systematically aggregate the web pages into groups by exploiting the sparsity inherent in the web. For each group, an aggregated PageRank value is computed, which can then be distributed among the group members. We provide a distributed update scheme for the aggregated PageRank along with an analysis on its convergence properties. The method is especially motivated by results on singular perturbation techniques for large-scale Markov chains and multi-agent consensus.

In this paper, we study the impact of heterogeneity on homogeneous protocols. Most of the protocols for increasing the lifetime of wireless sensor networks are designed for homogeneous networks. The load balancing protocol with adjustable sensing range (ALBPS) and deterministic energy efficient protocol with adjustable sensing range (ADEEPS), are two important homogeneous protocols. In this work, we incorporate 3-level heterogeneity to the ALBPS and ADEEPS and the resultant protocols are named as heterogeneous load balancing protocol with adjustable sensing range (HALBPS) and heterogeneous deterministic energy efficient protocol with adjustable sensing range (HADEEPS), respectively. We compare the performance of the ALBPS and ADEEPS protocols with that of the HALBPS and HADEEPS protocols. The simulation results indicate that the heterogeneous protocols further prolong the network lifetime.

Six variations of the token execution strategy of a distributed algorithm are described, applied to an algorithm for computing the minimum-weight spanning tree simulated on three network topologies, and compared with the chaotic execution strategy. In a chaotic execution, a processor may transmit a message M as soon as it generates M. In a token execution, a processor may transmit a message only when it holds a unique token. The token execution limits the number of messages in transit at the time. Execution with one token allows the user to observe the response to each message sequentially; execution with a fixed number of tokens provides a congestion control strategy. The best combination of variations of the token execution strategy uses 5.6% to 15.8% more messages and 0% to 12.3% more execution time than the chaotic execution, depending on network topology.

Sensor nodes in a wireless sensor network (WSN) can establish a link with a UAV by using beam forming techniques to from a random array with position errors. The position errors' effect in the array performance is examined using a MATLAB-based simulation model. In order to spread the processing and communication load among the nodes, two new distributed algorithms for beam forming in WSN, based on the least squares (LS) approximation of the desired array response, are proposed. The first is a distributed implementation of the QR decomposition, and the second is an iterative method for solving the LS problem. Results indicate that the processing load is effectively shared among the nodes. Especially, in the second approach, the processing load can be lower than that of the centralized approach, depending on the algorithm's convergence. For both algorithms, the tradeoff for the ability to spread the processing load is the increased communication cost, which could cause an overall increase in the total power consumption in the network. However, the average power per participating sensor node is still lower than that required by the cluster head in the centralized approach. Consequently, the network's susceptibility to failures due to excessive power consumption is greatly reduced.

In this paper we present generic distributed algorithms for assembling and repairing shapes using modular self-reconfiguring robots. The algorithms work in the sliding cube model. Each module independently evaluates a set of local rules using different evaluation models. Two methods are used to determine the correctness of the algorithmsa graph analysis technique which can prove the rule set is correct for specific instances of the algorithm, and a statistical technique which can produce arbitrary bounds on the likelihood that the rule set functions correctly. An extension of the assembly algorithm can be used to produce arbitrary non-cantilevered convex shapes without holes. The algorithms have been implemented and evaluated in simulation.

Multisensor data fusion combines data from multiple sensor systems to achieve improved performance and provide more inferences than could be achieved using a single sensor system. One of the most important aspects of data fusion is data association. This dissertation develops new algorithms for data association, including measurement to track association, track to track association and track fusion, in distributed multisensor multitarget environment with overlapping sensor coverage. The performance of the proposed algorithms is compared to that of existing techniques. Computational complexity analysis is also presented. Numerical results based on Monte Carlo simulations and real data collected from the United States Coast Guard Vessel Traffic Services system are presented. The results show that the proposed algorithms reduce the computational complexity and achieve considerable performance improvement over those previously reported in the literature.

The authors use the paradigm of diffusing computation, introduced by Dijkstra and Scholten, to solve a class of graph problems. They present a detailed solution to the problem of computing shortest paths from a single vertex to all other vertices, in the presence of negative cycles. (Author)

Data collection in a distributed embedded system requires dealing with failures: data corruptions by malicious devices and arbitrary message delay/loss in the network. Replication of data collection devices deals with such failures by voting among the replica devices to move a correct data to the end-user. Here, a data voted upon can be large-sized and/or take a long time to be compiled. The goal of this paper is to engineer the voting protocols for good performance while meeting the reliability requirements of data delivery in a high assurance setting. Two metric quantify the effectiveness of voting protocols: Data Transfer Efficiency (DTE) and Time-to-Complete (TTC) data delivery. DTE captures the network bandwidth wasted and/or the energy drain in wireless-connected devices; whereas, TTC captures the degradation in user-level Quality of Service (QoS) due to delayed/missed data deliveries. Given the distributed nature of voting, the protocol-level optimizations to improve DTE and TTC reduce the movement of user-level data over the network, the number of control messages generated, and the latency in effecting a data delivery. The paper describes these optimizations, and reports experimental results from a prototype voting system.

Data collection in a distributed embedded system requires dealing with failures: data corruptions by malicious devices and arbitrary message delay/loss in the network. Replication of data collection devices deals with such failures by voting among the replica devices to move a correct data to the end-user. Here, a data voted upon can be large-sized and/or take a long time to be compiled. The goal of this paper is to engineer the voting protocols for good performance while meeting the reliability requirements of data delivery in a high assurance setting. Two metric quantify the effectiveness of voting protocols: Data Transfer Efficiency (DTE) and Time-to-Complete (TTC) data delivery. DTE captures the network bandwidth wasted and/or the energy drain in wireless-connected devices; whereas, TTC captures the degradation in user-level Quality of Service (QoS) due to delayed/missed data deliveries. Given the distributed nature of voting, the protocol-level optimizations to improve DTE and TTC reduce the movement of user-level data over the network, the number of control messages generated, and the latency in effecting a data delivery. The paper describes these optimizations, and reports experimental results from a prototype voting system.

We discuss a method for controlling a group of mobile robots in a distributed manner. The method is fully distributed in the sense that each robot plans its motion individually based upon a given goal of the group and the observed positions of other robots. We illustrate the method by showing how a large number of robots can form an approximation of a circle, a simple polygon or a line segment in the plane. We also show how the robots can distribute themselves nearly uniformly within a circle or a convex polygon in the plane. Finally, we show how the robots can be divided into two or more groups. It turns out that in many cases most robots execute an identical, simple algorithm. The performance of the method is demonstrated by simulation

In this thesis we study the performance of update algorithms for replicated data in a distributed database. In doing so, we also investigate several other related issues. We start by presenting a simple model of a distributed database which is suitable for studying updates and concurrency control. We also develop a performance model and a set of parameters which represent the most important performance features of a distributed database. The distributed database models are used to study the performance of update algorithms for replicated data. This is done in two steps. First the algorithms are analyzed in the case of completely replicated databases in a no failure, update only environment. Then, the restrictions that we made are eliminated one at a time, and the impact on the system performance of doing this is evaluated. For the first step, we develop a new technique for analyzing the performance of update algorithms. This iterative technique is based on queueing theory. Several well known update algorithms are analyzed using this technique. The performance results are verified through detailed simulations of the algorithms. The results show that centralized control algorithms nearly always perform better than the more popular distributed control algorithms.

The Chandrasekhar equations arising in optimal control problems for linear distributed parameter systems are considered. The equations are derived via approximation theory. This approach is used to obtain existence, uniqueness, and strong differentiability of the solutions and provides the basis for a convergent computation scheme for approximating feedback gain operators. A numerical example is presented to illustrate these ideas.

Several techniques to perform static and dynamic load balancing techniques for vision systems are presented. These techniques are novel in the sense that they capture the computational requirements of a task by examining the data when it is produced. Furthermore, they can be applied to many vision systems because many algorithms in different systems are either the same, or have similar computational characteristics. These techniques are evaluated by applying them on a parallel implementation of the algorithms in a motion estimation system on a hypercube multiprocessor system. The motion estimation system consists of the following steps: (1) extraction of features; (2) stereo match of images in one time instant; (3) time match of images from different time instants; (4) stereo match to compute final unambiguous points; and (5) computation of motion parameters. It is shown that the performance gains when these data decomposition and load balancing techniques are used are significant and the overhead of using these techniques is minimal.

A new method for enhancing the convergence rate of iterative algorithms for the numerical integration of systems of partial differential equations was developed. It is termed the Distributed Minimal Residual (DMR) method and it is based on general Krylov subspace methods. The DMR method differs from the Krylov subspace methods by the fact that the iterative acceleration factors are different from equation to equation in the system. At the same time, the DMR method can be viewed as an incomplete Newton iteration method. The DMR method was applied to Euler equations of gas dynamics and incompressible Navier-Stokes equations. All numerical test cases were obtained using either explicit four stage Runge-Kutta or Euler implicit time integration. The formulation for the DMR method is general in nature and can be applied to explicit and implicit iterative algorithms for arbitrary systems of partial differential equations.

We initiate a thorough study of \emph{distributed property testing} -- producing algorithms for the approximation problems of property testing in the CONGEST model. In particular, for the so-called \emph{dense} testing model we emulate sequential tests for nearly all graph properties having $1$-sided tests, while in the \emph{sparse} and \emph{general} models we obtain faster tests for triangle-freeness and bipartiteness respectively. In most cases, aided by parallelism, the distributed algorithms have a much shorter running time as compared to their counterparts from the sequential querying model of traditional property testing. The simplest property testing algorithms allow a relatively smooth transitioning to the distributed model. For the more complex tasks we develop new machinery that is of independent interest. This includes a method for distributed maintenance of multiple random walks.

Distributed decision problems arise when two or more sensors viewing the same phenomenon must work cooperatively to draw inferences about the observed situation. Typical examples are in target detection and target classification. Such problems are characterized by distributed processing of information and communication between processors over a limited bandwidth data link. This thesis presents some statistical distributed decision algorithms and describes the implementation of one of them on a set of loosely coupled multiprocessor clusters which simulate the distributed environment characterizing multisensor decision problems. The purpose of the implementation was to investigate problems of communication and process synchronization in a pair of processor clusters performing a statistical distributed decision algorithm. This thesis describes how these communication and synchronization problems were addressed and solved.

Efficient and robust algorithms for decentralized estimation in networks are essential to many distributed systems. Whereas distributed estimation of sample mean statistics has been the subject of a good deal of attention, computation of $U$-statistics, relying on more expensive averaging over pairs of observations, is a less investigated area. Yet, such data functionals are essential to describe global properties of a statistical population, with important examples including Area Under the Curve, empirical variance, Gini mean difference and within-cluster point scatter. This paper proposes new synchronous and asynchronous randomized gossip algorithms which simultaneously propagate data across the network and maintain local estimates of the $U$-statistic of interest. We establish convergence rate bounds of $O(1/t)$ and $O(\log t / t)$ for the synchronous and asynchronous cases respectively, where $t$ is the number of iterations, with explicit data and network dependent terms. Beyond favorable comparisons in terms of rate analysis, numerical experiments provide empirical evidence the proposed algorithms surpasses the previously introduced approach.

This paper focuses on the problem of recursive nonlinear least squares parameter estimation in multi-agent networks, in which the individual agents observe sequentially over time an independent and identically distributed (i.i.d.) time-series consisting of a nonlinear function of the true but unknown parameter corrupted by noise. A distributed recursive estimator of the \emph{consensus} + \emph{innovations} type, namely $\mathcal{CIWNLS}$, is proposed, in which the agents update their parameter estimates at each observation sampling epoch in a collaborative way by simultaneously processing the latest locally sensed information~(\emph{innovations}) and the parameter estimates from other agents~(\emph{consensus}) in the local neighborhood conforming to a pre-specified inter-agent communication topology. Under rather weak conditions on the connectivity of the inter-agent communication and a \emph{global observability} criterion, it is shown that at every network agent, the proposed algorithm leads to consistent parameter estimates. Furthermore, under standard smoothness assumptions on the local observation functions, the distributed estimator is shown to yield order-optimal convergence rates, i.e., as far as the order of pathwise convergence is concerned, the local parameter estimates at each agent are as good as the optimal centralized nonlinear least squares estimator which would require access to all the observations across all the agents at all times. In order to benchmark the performance of the proposed distributed $\mathcal{CIWNLS}$ estimator with that of the centralized nonlinear least squares estimator, the asymptotic normality of the estimate sequence is established and the asymptotic covariance of the distributed estimator is evaluated. Finally, simulation results are presented which illustrate and verify the analytical findings.

Interference mitigation has a great potential for improving the performance of interference limited wireless networks. In this paper we introduce novel schemes for interference management in wireless cellular networks using a virtual layer that captures and simplifies the complicated interference situation in the network. We show how a network utility maximization approach in conjunction with suitable short term scheduling and optimization in this virtual layer can be used for autonomous interference minimization by power control. We compare three distributed algorithms and evaluate their applicability for different user mobility assumptions. These algorithms gradually and autonomously steer the network towards a higher utility. Thereby the granularity of control ranges from controlling frequency subband power via controlling the power on a per-beam basis through to only enforcing average power constraints per beam. We use extensive system-level simulations which indicate high gains. In particular, it turns out that larger gains can be achieved by imposing average power constraints and allowing opportunistic scheduling instantaneously, rather than controlling the power in a strict way. Further we introduce a centralized version directly solving the underlying optimization and showing fast convergence, which serves as a performance benchmark for the distributed solutions. Finally, we investigate the deviation from global optimality by comparing to a Branch-And-Bound based solution.

Typical desktop workstations can be a severe bottleneck in the viewing and enhancement of imagery data. Due to the nature of many image processing algorithms, an effective method for alleviating this problem is through parallelism. Parallel hardware can come in many forms, from small clusters of workstations and workstations with many processors to dedicated hardware containing 10's, 100's and 1000's of processing nodes. One of the challenges is developing a portable parallel image processing library in such a potentially diverse environment. These issues led to the development of a Parallel Image Processing Toolkit (PIPT). The toolkit hides the detail of parallelization from the users of the PIPT and provides a uniform programming interface. In developing the toolkit the issues of advanced data handling, load, balancing and parallel visualization were addressed. In addition a specific computationally expensive High Resolution Video Stills algorithm was implemented within the PIPT. The important contributions that resulted from this work transcend the individual tasks that were undertaken. First, the general approach that was taken to implement the PIPT illustrates several important principles for the design and implementation of general purpose parallel libraries. in this regard, the PIPT design can serve as a design pattern for an extensible parallel library. Second, the design pattern of the current implementation of the PIPT contains some notable attempts at programming image processing tasks in a generic fashion.

Distributed Computing Principles, Algorithms, and Systems

Distributed Computing Principles, Algorithms, and Systems

Distributed Computing Principles, Algorithms, and Systems

http://uf.catalog.fcla.edu/uf.jsp?st=UF025576824&ix=pm&I=0&V=D&pm=1

http://uf.catalog.fcla.edu/uf.jsp?st=UF025576824&ix=pm&I=0&V=D&pm=1

In this paper, a class of convex feasibility problems (CFPs) are studied for multi-agent systems through local interactions. The objective is to search a feasible solution to the convex inequalities with some set constraints in a distributed manner. The distributed control algorithms, involving subgradient and projection, are proposed for both continuous- and discrete-time systems, respectively. Conditions associated with connectivity of the directed communication graph are given to ensure convergence of the algorithms. It is shown that under mild conditions, the states of all agents reach consensus asymptotically and the consensus state is located in the solution set of the CFP. Simulation examples are presented to demonstrate the effectiveness of the theoretical results.

Consider the continuous distributed monitoring model in which $n$ distributed nodes, receiving individual data streams, are connected to a designated server. The server is asked to continuously monitor a function defined over the values observed across all streams while minimizing the communication. We study a variant in which the server is equipped with a broadcast channel and is supposed to keep track of an approximation of the set of nodes currently observing the $k$ largest values. Such an approximate set is exact except for some imprecision in an $\varepsilon$-neighborhood of the $k$-th largest value. This approximation of the Top-$k$-Position Monitoring Problem is of interest in cases where marginal changes (e.g. due to noise) in observed values can be ignored so that monitoring an approximation is sufficient and can reduce communication. This paper extends our results from [IPDPS'15], where we have developed a filter-based online algorithm for the (exact) Top-k-Position Monitoring Problem. There we have presented a competitive analysis of our algorithm against an offline adversary that also is restricted to filter-based algorithms. Our new algorithms as well as their analyses use new methods. We analyze their competitiveness against adversaries that use both exact and approximate filter-based algorithms, and observe severe differences between the respective powers of these adversaries.

We propose two novel algorithms for distributed and location-free boundary recognition in wireless sensor networks. Both approaches enable a node to decide autonomously whether it is a boundary node, based solely on connectivity information of a small neighborhood. This makes our algorithms highly applicable for dynamic networks where nodes can move or become inoperative. We compare our algorithms qualitatively and quantitatively with several previous approaches. In extensive simulations, we consider various models and scenarios. Although our algorithms use less information than most other approaches, they produce significantly better results. They are very robust against variations in node degree and do not rely on simplified assumptions of the communication model. Moreover, they are much easier to implement on real sensor nodes than most existing approaches.

Random walk based distributed algorithms make use of a token that circulates in the system according to a random walk scheme to achieve their goal. To study their efficiency and compare it to one of the deterministic solutions, one is led to compute certain quantities, namely the hitting times and the cover time. Until now, only bounds on these quantities were defined. First, this paper presents two generalizations of the notions of hitting and cover times to weighted graphs. Indeed, the properties of random walks on symmetrically weighted graphs provide interesting results on random walk based distributed algorithms, such as local load balancing. Both of these generalization are proposed to precisely represent the behaviour of these algorithms, and to take into account what the weights represent. Then, we propose an algorithm to compute the n^2 hitting times on a weighted graph of n vertices, which we improve to obtain a O(n^3) complexity. This complexity is the lowest up to now. This algorithm computes both of the generalizations that we propose for the hitting times on a weighted graph. Finally, we provide the first algorithm to compute the cover time (in both senses) of a graph. We improve it to achieve a complexity of O(n^3 2^n). The algorithms that we present are all robust to a topological change in a limited number of edges. This property allows us to use them on dynamic graphs.

Random walk based distributed algorithms make use of a token that circulates in the system according to a random walk scheme to achieve their goal. To study their efficiency and compare it to one of the deterministic solutions, one is led to compute certain quantities, namely the hitting times and the cover time. Until now, only bounds on these quantities were defined. First, this paper presents two generalizations of the notions of hitting and cover times to weighted graphs. Indeed, the properties of random walks on symmetrically weighted graphs provide interesting results on random walk based distributed algorithms, such as local load balancing. Both of these generalization are proposed to precisely represent the behaviour of these algorithms, and to take into account what the weights represent. Then, we propose an algorithm to compute the n^2 hitting times on a weighted graph of n vertices, which we improve to obtain a O(n^3) complexity. This complexity is the lowest up to now. This algorithm computes both of the generalizations that we propose for the hitting times on a weighted graph. Finally, we provide the first algorithm to compute the cover time (in both senses) of a graph. We improve it to achieve a complexity of O(n^3 2^n). The algorithms that we present are all robust to a topological change in a limited number of edges. This property allows us to use them on dynamic graphs.

This paper describes a new fast and implicitly parallel approach to neighbour-finding in multi-resolution Smoothed Particle Hydrodynamics (SPH) simulations. This new approach is based on hierarchical cell decompositions and sorted interactions, within a task-based formulation. It is shown to be faster than traditional tree-based codes, and to scale better than domain decomposition-based approaches on hybrid shared/distributed-memory parallel architectures, e.g. clusters of multi-cores, achieving a $40\times$ speedup over the Gadget-2 simulation code.

The forest-of-octrees approach to parallel adaptive mesh refinement and coarsening (AMR) has recently been demonstrated in the context of a number of large-scale PDE-based applications. Although linear octrees, which store only leaf octants, have an underlying tree structure by definition, it is not often exploited in previously published mesh-related algorithms. This is because the branches are not explicitly stored, and because the topological relationships in meshes, such as the adjacency between cells, introduce dependencies that do not respect the octree hierarchy. In this work we combine hierarchical and topological relationships between octree branches to design efficient recursive algorithms. We present three important algorithms with recursive implementations. The first is a parallel search for leaves matching any of a set of multiple search criteria. The second is a ghost layer construction algorithm that handles arbitrarily refined octrees that are not covered by previous algorithms, which require a 2:1 condition between neighboring leaves. The third is a universal mesh topology iterator. This iterator visits every cell in a domain partition, as well as every interface (face, edge and corner) between these cells. The iterator calculates the local topological information for every interface that it visits, taking into account the nonconforming interfaces that increase the complexity of describing the local topology. To demonstrate the utility of the topology iterator, we use it to compute the numbering and encoding of higher-order $C^0$ nodal basis functions. We analyze the complexity of the new recursive algorithms theoretically, and assess their performance, both in terms of single-processor efficiency and in terms of parallel scalability, demonstrating good weak and strong scaling up to 458k cores of the JUQUEEN supercomputer.

We consider optimizations that are required for efficient execution of code segments that consists of loops over distributed data structures. The PARTI (Parallel Automated Runtime Toolkit at ICASE) execution time primitives are designed to carry out these optimizations and can be used to implement a wide range of scientific algorithms on distributed memory machines. These primitives allow the user to control array mappings in a way that gives an appearance of shared memory. Computations can be based on a global index set. Primitives are used to carry out gather and scatter operations on distributed arrays. Communications patterns are derived at runtime, and the appropriate send and receive messages are automatically generated. (kr)

This volume presents the results of research on techniques to develop distributed data processing architectures for specific functions, with particular reference to BMD functions. It describes two methodologies and six designs for two BMD functions. Short discussions of two nonfunctional payoffs are included to aid future workers on these problems. Many suggestions for future research and development activities are given. (Author)

This volume presents the results of research on techniques to develop distributed data processing architectures for specific functions, with particular reference to BMD functions. It describes two methodologies and six designs for two BMD functions. Short discussions of two nonfunctional payoffs are included to aid future workers on these problems. Many suggestions for future research and development activities are given. (Author)

This volume presents the results of research on techniques to develop distributed data processing architectures for specific functions, with particular reference to BMD functions. It describes two methodologies and six designs for two BMD functions. Short discussions of two nonfunctional payoffs are included to aid future workers on these problems. Many suggestions for future research and development activities are given. (Author)

The implementation of a vast majority of machine learning (ML) algorithms boils down to solving a numerical optimization problem. In this context, Stochastic Gradient Descent (SGD) methods have long proven to provide good results, both in terms of convergence and accuracy. Recently, several parallelization approaches have been proposed in order to scale SGD to solve very large ML problems. At their core, most of these approaches are following a map-reduce scheme. This paper presents a novel parallel updating algorithm for SGD, which utilizes the asynchronous single-sided communication paradigm. Compared to existing methods, Asynchronous Parallel Stochastic Gradient Descent (ASGD) provides faster (or at least equal) convergence, close to linear scaling and stable accuracy.

In geographically-distributed systems, communication latencies are non-negligible. The perceived processing time of a request is thus composed of the time needed to route the request to the server and the true processing time. Once a request reaches a target server, the processing time depends on the total load of that server; this dependency is described by a load function. We consider a broad class of load functions; we just require that they are convex and two times differentiable. In particular our model can be applied to heterogeneous systems in which every server has a different load function. This approach allows us not only to generalize results for queuing theory and for batches of requests, but also to use empirically-derived load functions, measured in a system under stress-testing. The optimal assignment of requests to servers is communication-balanced, i.e. for any pair of non perfectly-balanced servers, the reduction of processing time resulting from moving a single request from the overloaded to underloaded server is smaller than the additional communication latency. We present a centralized and a decentralized algorithm for optimal load balancing. We prove bounds on the algorithms' convergence. To the best of our knowledge these bounds were not known even for the special cases studied previously (queuing theory and batches of requests). Both algorithms are any-time algorithms. In the decentralized algorithm, each server balances the load with a randomly chosen peer. Such algorithm is very robust to failures. We prove that the decentralized algorithm performs locally-optimal steps. Our work extends the currently known results by considering a broad class of load functions and by establishing theoretical bounds on the algorithms' convergence. These results are applicable for servers whose characteristics under load cannot be described by a standard mathematical models.