arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 1610.00253cs.LO
收藏

基于不动点的计算域的异步分布式执行

Asynchronous Distributed Execution Of Fixpoint-Based Computational Fields

Alberto Lluch Lafuente, Michele Loreti, Ugo Montanari

摘要:

协调对于动态分布式系统是必不可少的,其组件表现出交互性和自治行为。空间分布的、局部交互的、传播的计算场对于允许组件以很少或没有开销的方式加入和离开特别有吸引力。计算领域是聚合编程的关键组成部分,聚合编程是一种很有前途的软件工程方法,尤其与物联网相关。在我们的方法中,空间拓扑结构表示为一个固定的图形形状的字段,即一个网络的属性上的节点和弧,弧代表节点之间的交互能力。我们提出了一个SMuC演算,其中μ演算样的模态公式表示如何存储在邻居节点的值应该结合起来更新本节点。固定点操作可以全局地理解为递归定义,或者局部地理解为异步收敛传播过程。我们提出了一个分布式实现我们的演算。翻译首先完成映射SMuC程序到范式,纯迭代程序,然后到分布式程序。一些关键的结果表明,收敛的不动点计算公平的duplex和重新初始化的节点。第一个结果允许节点以不同的速度前进,而第二个结果提供了对某些类型故障的鲁棒性。我们说明了我们的方法与案例研究的基础上的灾难恢复的情况下,在一个原型模拟器,我们用来评估恢复策略的性能。

英文摘要:

Coordination is essential for dynamic distributed systems whose components exhibit interactive and autonomous behaviors. Spatially distributed, locally interacting, propagating computational fields are particularly appealing for allowing components to join and leave with little or no overhead. Computational fields are a key ingredient of aggregate programming, a promising software engineering methodology particularly relevant for the Internet of Things. In our approach, space topology is represented by a fixed graph-shaped field, namely a network with attributes on both nodes and arcs, where arcs represent interaction capabilities between nodes. We propose a SMuC calculus where mu-calculus- like modal formulas represent how the values stored in neighbor nodes should be combined to update the present node. Fixpoint operations can be understood globally as recursive definitions, or locally as asynchronous converging propagation processes. We present a distributed implementation of our calculus. The translation is first done mapping SMuC programs into normal form, purely iterative programs and then into distributed programs. Some key results are presented that show convergence of fixpoint computations under fair asynchrony and under reinitialization of nodes. The first result allows nodes to proceed at different speeds, while the second one provides robustness against certain kinds of failure. We illustrate our approach with a case study based on a disaster recovery scenario, implemented in a prototype simulator that we use to evaluate the performance of a recovery strategy.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑