Shortest paths problems are subject to extensive studies in classic distributed models such as the CONGEST or congested clique, which describe the way in which nodes may communicate in order to solve such a problem. %where nodes initially know only the distance to their neighbors in some graph and must compute distances to any other node with as few communication rounds as possible. This article focuses on hybrid networks, which give nodes access to multiple, different modes of communication, in particular the HYBRID model which combines unrestricted local communication along edges of the input graph alongside heavily restricted global communication between arbitrary pairs of nodes. Previous work [Augustine et al, SODA'20, Kuhn et al. PODC'20] showed that each node learning its distance to $k$ dedicated source nodes (aka the $k$-SSP problem) takes at least $\tilOm(\!\sqrt{k})$ rounds in the HYBRID model, even for polynomial approximations. This lower bound was matched with algorithmic solutions for $k \geq n^{2/3}$. However, as $k$ gets smaller, the gap between the known upper and lower bounds diverges and even becomes exponential for the single source shortest paths problem (SSSP). In this work we plug this gap for the whole range of $k$ (up to terms that are polylogarithmic in $n$), by giving algorithmic solutions for $k$-SSP in $\tilO\big(\!\sqrt k\big)$ rounds for any $k$.
翻译:最短路径问题受到经典分布式模型(如CONGEST或拥塞丛)的广泛研究,它们描述了节点如何进行通信以解决这样的问题。本文重点研究混合网络,它们为节点提供了多个不同的通信模式,特别是HYBRID模型,它将输入图的无限制本地通信与任意对节点之间的严格限制的全局通信相结合。以前的工作[Augustine等人,SODA'20,Kuhn等人,PODC'20]表明,在HYBRID模型中,每个节点学习到其到k个专用源节点(称为k-SSP问题)的距离至少需要$\tilOm(\!\sqrt{k})$回合,即使进行多项式逼近也是如此。该下界已经匹配了$k \geq n^{2/3}$的算法解决方案。但是,随着k的减小,已知上下限之间的差距越来越大,甚至在单源最短路径问题(SSSP)中变成指数级。在这项工作中,我们通过给出任何k的$\tilO\big(\!\sqrt k\big)$回合的算法解决方案来填补这个差距,覆盖了整个k范围(直到$n$的对数的多项式级)。