distributed computing (DC) projects tackle large computational problems by exploiting the donated processing power of thousands of volunteered computers, connected through the Internet. To efficiently employ the compu...
详细信息
distributed computing (DC) projects tackle large computational problems by exploiting the donated processing power of thousands of volunteered computers, connected through the Internet. To efficiently employ the computational resources of one of world's largest DC efforts, GPUGRID, the project scientists require tools that handle hundreds of thousands of tasks which run asynchronously and generate gigabytes of data every day. We describe RBoinc, an interface that allows computational scientists to embed the DC methodology into the daily work-flow of high-throughput experiments. By extending the Berkeley Open Infrastructure for Network computing (BOINC), the leading open-source middleware for current DC projects, with mechanisms to submit and manage large-scale distributed computations from individual workstations, RBoinc turns distributed grids into cost-effective virtual resources that can be employed by researchers in work-flows similar to conventional supercomputers. The GPUGRID project is currently using RBoinc for all of its in silico experiments based on molecular dynamics methods, including the determination of binding free energies and free energy profiles in all-atom models of biomolecules. (C) 2010 Elsevier B.V. All rights reserved.
Consider a set of n > 2 identical mobile computational entities in the plane, called robots, operating in Look-Compute-Move cycles, without any means of direct communication. The GATHERING PROBLEM is the primitive ...
详细信息
Consider a set of n > 2 identical mobile computational entities in the plane, called robots, operating in Look-Compute-Move cycles, without any means of direct communication. The GATHERING PROBLEM is the primitive task of all entities gathering in finite time at a point not fixed in advance, without any external control. The problem has been extensively studied in the literature under a variety of strong assumptions (e.g., synchronicity of the cycles, instantaneous movements, complete memory of the past, common coordinate system, etc.). In this paper we consider the setting without those assumptions, that is, when the entities are oblivious (i.e., they do not remember results and observations from previous cycles), disoriented (i.e., have no common coordinate system), and fully asynchronous (i.e., no assumptions exist on timing of cycles and activities within a cycle). The existing algorithmic contributions for such robots are limited to solutions for n <= 4 or for restricted sets of initial configurations of the robots;the question of whether such weak robots could deterministically gather has remained open. In this paper, we prove that indeed the GATHERING PROBLEM is solvable, for any n > 2 and any initial configuration, even under such restrictive conditions.
This paper presents a distributed computing architecture for solving a distribution optimal power flow (DOPF) model based on a smart grid communication middleware (SGCM) system. The system is modeled as an unbalanced ...
详细信息
This paper presents a distributed computing architecture for solving a distribution optimal power flow (DOPF) model based on a smart grid communication middleware (SGCM) system. The system is modeled as an unbalanced three-phase distribution system, which includes different kind of loads and various components of distribution systems. In this paper, fixed loads are modeled as constant impedance, current and power loads, and neural network models of controllable smart loads are integrated into the DOPF model. A genetic algorithm is used to determine the optimal solutions for controllable devices, in particular load tap changers, switched capacitors, and smart loads in the context of an energy management system for practical feeders, accounting for the fact that smart loads consumption should not be significantly affected by network constraints. Since the number of control variables in a realistic distribution power system is large, solving the DOPF for real-time applications is computationally expensive. Hence, to reduce computational times, a decentralized system with parallel computing nodes based on an SGCM system is proposed. Using a "MapReduce" model, the SGCM system runs the DOPF model, communicates between master and worker computing nodes, and sends/receives data among different parts of parallel computing system. Compared to a centralized approach, the proposed architecture is shown to yield better optimal solutions in terms of reducing energy losses and/or energy drawn from the substation within adequate practical run-times for a realistic test feeder.
Inspired by social networks and complex systems, we propose a core-periphery network architecture that supports fast computation for many distributed algorithms, is robust and uses a linear number of links. Rather tha...
详细信息
Inspired by social networks and complex systems, we propose a core-periphery network architecture that supports fast computation for many distributed algorithms, is robust and uses a linear number of links. Rather than providing a concrete network model, we take an axiom-based design approach. We provide three intuitive and independent algorithmic axioms and prove that any network that satisfies all axioms enjoys an efficient algorithm for a range of tasks (such as MST, sparse matrix multiplication, and more). We also show the necessity of our axiom set: for networks that satisfy any subset of the axioms, the same efficiency cannot be guaranteed for any deterministic algorithm. (C) 2016 Elsevier Inc. All rights reserved.
distributed computing is a process through which a set of computers connected by a network is used collectively to solve a single problem. In this paper, we propose a distributed computing methodology for training neu...
详细信息
distributed computing is a process through which a set of computers connected by a network is used collectively to solve a single problem. In this paper, we propose a distributed computing methodology for training neural networks for the detection of lesions in colonoscopy Our approach is based on partitioning the training set across multiple processors using a parallel virtual machine. In this way, interconnected computers of varied architectures can be used for the distributed evaluation of the error function and gradient values, and, thus, training neural networks utilizing various learning methods. The proposed methodology has large granularity and low synchronization, and has been implemented and tested. our results indicate that the parallel virtual machine implementation of the training algorithms developed leads to considerable speedup, especially when large network architectures and training sets are used. (c) 2005 Elsevier Ireland Ltd. All rights reserved.
We study the problem of the amount of information (advice) about a graph that must be given to its nodes in order to achieve fast distributed computations. The required size of the advice enables to measure the inform...
详细信息
We study the problem of the amount of information (advice) about a graph that must be given to its nodes in order to achieve fast distributed computations. The required size of the advice enables to measure the information sensitivity of a network problem. A problem is information sensitive if little advice is enough to solve the problem rapidly (i.e., much faster than in the absence of any advice), whereas it is information insensitive if it requires giving a lot of information to the nodes in order to ensure fast computation of the solution. In this paper, we study the information sensitivity of distributed graph coloring.
distributed computing is one of the paradigms in the world of information technology. Middleware is the essential tool for implementing distributed computing for overtaking the heterogeneity of platform and language. ...
详细信息
distributed computing is one of the paradigms in the world of information technology. Middleware is the essential tool for implementing distributed computing for overtaking the heterogeneity of platform and language. DRDO's intranet, DRONA, has the potential of hosting distributed applications across the network. This paper deals with the essentials of distributed computing, architecture of DRONA network, and the scope of distributed computing in Defence applications. It also suggests a few possible applications of distributed computing.
Applications of probit-based stochastic user equilibrium (SUE) principle on large-scale networks have been largely limited because of the overwhelming computational burden in solving its stochastic network loading pro...
详细信息
Applications of probit-based stochastic user equilibrium (SUE) principle on large-scale networks have been largely limited because of the overwhelming computational burden in solving its stochastic network loading problem. A two-stage Monte Carlo simulation method is recognized to have satisfactory accuracy level when solving this stochastic network loading. This paper thus works on the acceleration of the Monte Carlo simulation method via using distributed computing system. Three distributed computing approaches are then adopted on the workload partition of the Monte Carlo simulation method. Wherein, the first approach allocates each processor in the distributed computing system to solve each trial of the simulation in parallel and in turns, and the second approach assigns all the processors to solve the shortest-path problems in one trial of the Monte Carlo simulation concurrently. The third approach is a combination of the first two, wherein both different trials of the Monte Carlo simulation as well as the shortest path problems in one trial are solved simultaneously. Performances of the three approaches are comprehensively tested by the Sioux-Falls network and then a randomly generated network example. It shows that computational time for the probit-based SUE problem can be largely reduced by any of these three approaches, and the first approach is found out to be superior to the other two. The first approach is then selected to calculate the probit-based SUE problem on a large-scale network example. Copyright (c) 2011 John Wiley & Sons, Ltd.
The paper studies three fundamental problems in graph analytics, computing connected components (CCs), biconnected components (BCCs), and 2-edge-connected components (ECCs) of a graph. With the recent advent of big da...
详细信息
The paper studies three fundamental problems in graph analytics, computing connected components (CCs), biconnected components (BCCs), and 2-edge-connected components (ECCs) of a graph. With the recent advent of big data, developing efficient distributed algorithms for computing CCs, BCCs and ECCs of a big graph has received increasing interests. As with the existing research efforts, we focus on the Pregel programming model, while the techniques may be extended to other programming models including MapReduce and Spark. The state-of-the-art techniques for computing CCs and BCCs in Pregel incur O(m x #supersteps) total costs for both data communication and computation, where m is the number of edges in a graph and #supersteps is the number of supersteps. Since the network communication speed is usually much slower than the computation speed, communication costs are the dominant costs of the total running time in the existing techniques. In this paper, we propose a new paradigm based on graph decomposition to compute CCs and BCCs with O(m) total communication cost. The total computation costs of our techniques are also smaller than that of the existing techniques in practice, though theoretically almost the same. Moreover, we also study distributed computing ECCs. We are the first to study this problem and an approach with O(m) total communication cost is proposed. Comprehensive empirical studies demonstrate that our approaches can outperform the existing techniques by one order of magnitude regarding the total running time.
Most existing distributed systems are structured as statically compiled processes communicating with each other via messages. The system's ''intelligence'' is embodied in the processes, while the m...
详细信息
Most existing distributed systems are structured as statically compiled processes communicating with each other via messages. The system's ''intelligence'' is embodied in the processes, while the messages contain simple, passive pieces information (the communicating objects paradigm). In the autonomous objects paradigm, a message has its own identity and behavior. It decides at runtime where it wants to propagate and what tasks to perform there;the nodes become simply generic interpreters that enable messages to navigate and compute. In this scenario, an application's ''intelligence'' is embodied in and carried by messages as they propagate through the network, much as human agent or a robot would move in space, visiting different locales as it performs tasks. The autonomous objects paradigm is more flexible than the communicating objects paradigm because it allows developers to change the program's behavior after it has started to run. We based our system, MESSENGERS, on autonomous objects, and intended it for the composition and coordination of concurrent activities in a distributed environment. It combines powerful navigational capabilities found in other autonomous objects-based systems with efficient dynamic linking mechanisms supported by some new programming languages, like Java. MESSENGERS allows the dynamic construction of arbitrarily complex control sequences, which are carried through the network. The sequences can invoke node-resident computational programs and coordinate their operation by carrying information among them.
暂无评论