arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2502.12697cs.DC
收藏

弱通信下的极简领导者选举

Minimalist Leader Election Under Weak Communication

Robin Vacus, Isabella Ziccardi

AI总结:

提出一种在弱通信模型下仅需6种状态、无需唯一标识符和网络先验知识的均匀Leader Election协议,系统几乎必然收敛至单领导者状态,耗时约$O(D^2 \log n)$轮,适用于最简单的分布式及生物系统。

摘要:

我们提出了一种在弱通信模型(如beeping model或stone-age model)下解决Leader Election的协议。与以往大多数工作不同,我们的算法仅需6种状态,不要求唯一标识符,且不假设网络规模或拓扑的先验知识,即它是均匀的。我们证明,在我们的协议下,系统几乎必然收敛到仅有一个节点处于领导者状态的配置。在大概率下,这会在少于$O(D^2 \log n)$轮内发生,其中$D$为网络直径。我们还证明,当已知$D$的常数因子近似时,该时间可降至$O(D \log n)$。该方法的主要缺点是,与要求更强的算法相比,运行时间存在$\TildeΩ(D)$的开销,且节点无法感知何时达到单领导者配置。然而,极简的假设与自然吸引力使该方案特别适合在最简单的分布式系统(尤其是生物系统)中实现。

英文摘要:

We propose a protocol to solve Leader Election within weak communication models such as the beeping model or the stone-age model. Unlike most previous work, our algorithm operates on only six states, does not require unique identifiers, and assumes no prior knowledge of the network's size or topology, i.e., it is uniform. We show that under our protocol, the system almost surely converges to a configuration in which a single node is in a leader state. With high probability, this occurs in fewer than $O(D^2 \log n)$ rounds, where $D$ is the network diameter. We also show that this can be decreased to $O(D \log n)$ when a constant factor approximation of $D$ is known. The main drawbacks of our approach are a $\TildeΩ(D)$ overhead in the running time compared to algorithms with stronger requirements, and the fact that nodes are unaware of when a single-leader configuration is reached. Nevertheless, the minimal assumptions and natural appeal of our solution make it particularly well-suited for implementation in the simplest distributed systems, especially biological ones.

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。

↑