We examine the power and limitations of the weakest vertex relabelling system which allows to change a label of a vertex in function of its own label and of the label of one of its neighhours. We characterize the grap...
详细信息
We examine the power and limitations of the weakest vertex relabelling system which allows to change a label of a vertex in function of its own label and of the label of one of its neighhours. We characterize the graphs for which two important distributed algorithmic problems are solvable in this model : naming and election.
In this paper, we analyze scheduling gathering multitype data in a star network. Certain numbers of datasets of different types have to be collected on a single computer, either by downloading them from remote nodes, ...
详细信息
In this paper, we analyze scheduling gathering multitype data in a star network. Certain numbers of datasets of different types have to be collected on a single computer, either by downloading them from remote nodes, or by producing them locally. The time required to download a dataset depends on its type and initial location, and the time needed to produce a dataset depends only on its type. The scheduling problem is to decide which datasets should be downloaded from which nodes, and how many datasets of each type should be generated locally, in order to minimize the total time necessary for gathering all data. We show that this problem is NP-hard, indicate special cases solvable in polynomial time, and propose a fully polynomial time approximation scheme for a special case which is still NP-hard. For the general case, we present an integer linear programming formulation, a dynamic programming algorithm and a polynomial 2-approximation algorithm. The performance of these algorithms is tested in computational experiments. (c) 2021 Elsevier B.V. All rights reserved.
In dynamic distributed systems, the topology of networks changes over time which makes difficult the design and much harder the proof of distributed algorithms. These unavoidable changes of the topology make the elect...
详细信息
In dynamic distributed systems, the topology of networks changes over time which makes difficult the design and much harder the proof of distributed algorithms. These unavoidable changes of the topology make the election and the maintenance of the elected leader a complex task. The maintaining problem is not considered in a static context. To encode distributed algorithms, we adopt the local computation model. Distributed algorithms are formally presented by transitions systems. Based on both, the mobile agent paradigm and the local computation model, we present in this paper a distributed algorithm that elects a leader in a tree. A set of topological events that may affect the structure of the tree: we focus on the appearance and the disappearance of places as well as the communication channels. Our goal is to maintain always a tree with a single leader or a forest of trees where each one has his own leader. 1877-0509 (C) 2017 The Authors. Published by Elsevier B.V.
In this paper, we propose a formal framework enhancing the termination detection property of distributed algorithms and reusing their specifications as well as their proofs. By relying on refinement and composition, w...
详细信息
ISBN:
(纸本)9783319336008;9783319335995
In this paper, we propose a formal framework enhancing the termination detection property of distributed algorithms and reusing their specifications as well as their proofs. By relying on refinement and composition, we show that an algorithm specified with local termination detection, can be reused in order to compute the same algorithm with global termination detection. The main idea relies upon the development of distributed algorithms following a top/down approach and the integration of additional computation steps developed in a pre-defined module. This module is specified in a generic and scalable way in order to be composed with particular developments. Once the composition link is proven, the global termination emerges automatically.
Due to the lack of knowledge of the global state and the non determinism in the execution of the processes, distributed algorithms are considered to be very complex to design and to prove. However, it becomes crucial ...
详细信息
ISBN:
(纸本)9781479942497
Due to the lack of knowledge of the global state and the non determinism in the execution of the processes, distributed algorithms are considered to be very complex to design and to prove. However, it becomes crucial to guarantee that these algorithms run as designed. Modularization mechanism in formal development provides a simple way to manage this complexity. In this paper, we rely on the modularization mechanism of the Event-B method and on local computations model to propose a reuse based approach for modelling classes of distributed algorithms. The proposed approach consists in developing a formal pattern defined as a set of proved logical entities called modules. These modules are developed separately and, when needed, can be incorporated and instantiated in a given system development. Such a mechanism can save efforts on modelling and proving the computation steps in distributed algorithms.
Most of distributed algorithms encoded by means of local computations [3] need to solve k-local election problems to ensure a faithful relabeling of disjoint subgraphs. Due to a result stated in [1], it is not possibl...
详细信息
ISBN:
(纸本)3540259201
Most of distributed algorithms encoded by means of local computations [3] need to solve k-local election problems to ensure a faithful relabeling of disjoint subgraphs. Due to a result stated in [1], it is not possible to solve the k-local election problem for k >= 3 in anonymous networks. Based on distributed computations of rooted trees of minimal paths, we present in this paper a simple randomized algorithm which, with very high probability, solves the k-local election problem (k >= 2) in an anonymous graph.
Global predicates in distributed systems are predicates considering the state of more than one process. They are a useful concept for debugging parallel programs, e.g., for specifying assertions, breakpoints or termin...
详细信息
ISBN:
(纸本)0769521606
Global predicates in distributed systems are predicates considering the state of more than one process. They are a useful concept for debugging parallel programs, e.g., for specifying assertions, breakpoints or termination. In this paper we propose an algorithm for detecting the validity of local and global state predicates in distributed algorithms encoded by means of local computations. This algorithm is based on ideas introduced by Cooper and Marzullo and on the approach of Garg and Waldecker. It uses the Lamport's happened-before ordering relationship to construct a partial sorted sequence of all the events that have changed the local state of any process during a computation. While the sequence is traversed, all the global states through which the system has passed are computed and the validity of the given properties is checked. Under the same assumptions, our algorithm is significantly more efficient than the algorithms by Cooper and Marzullo and by Garg and Waldecker.
Distributed systems have been studied thoroughly. Many applications are based on huge systems and one of the main property an algorithm running on such systems should verified is fault-tolerance. This paper presents a...
详细信息
ISBN:
(纸本)9781509024612
Distributed systems have been studied thoroughly. Many applications are based on huge systems and one of the main property an algorithm running on such systems should verified is fault-tolerance. This paper presents a new randomized algorithm to solve the handshake problem in a distributed system. This algorithm is designed for an asynchronous distributed network of anonymous processes under the message passing communication protocol. In addition to its fault-tolerance, this algorithm is more effective than the existing ones thanks to its asynchronous aspect. To highlight its performance, we provide an experimental evaluation by comparing it to the most representative algorithm solving the handshake problem. The experimentation is done using Visidia, a tool for simulating and visualizing distributed algorithms.
In dynamic distributed systems, the topology of networks changes over time which makes difficult the design and much harder the proof of distributed algorithms. These unavoidable changes of the topology make the elect...
详细信息
In dynamic distributed systems, the topology of networks changes over time which makes difficult the design and much harder the proof of distributed algorithms. These unavoidable changes of the topology make the election and the maintenance of the elected leader a complex task. The maintaining problem is not considered in a static context. To encode distributed algorithms, we adopt the local computation model. Distributed algorithms are formally presented by transitions systems. Based on both, the mobile agent paradigm and the local computation model, we present in this paper a distributed algorithm that elects a leader in a tree. A set of topological events that may affect the structure of the tree: we focus on the appearance and the disappearance of places as well as the communication channels. Our goal is to maintain always a tree with a single leader or a forest of trees where each one has his own leader.
暂无评论