arXiv 2501.092932025-01cs.DS

异构并行网络中具有可证明保证的非分割Coflow调度

Non-Splitting Coflow Scheduling with Provable Guarantees in Heterogeneous Parallel Networks

National Cheng Kung University（国立成功大学）

Chi-Yeh Chen

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

针对异构并行网络中非分割coflow调度问题，提出统一多项式时间近似算法，最小化完工时间，并在纯EPS、纯非全停OCS和纯全停OCS环境下给出近似比。

摘要：

作为一种突出的网络抽象，coflow有效地捕获了数据中心中的通信模式。由于大规模数据中心中的coflow调度是$\mathcal{NP}$-难的，现有文献主要关注具有$m=2$个网络核心的有限环境，并依赖于流分割，这引入了大量的操作开销。关键的是，对于更实际的非分割coflow调度问题，即使对于$m=2$的情况，也没有提出具有可证明性能保证的近似算法，更不用说一般的混合架构了。为了弥合这一关键差距，本文研究了混合异构并行网络中的非分割问题，该网络具有多个网络核心（$m \ge 2$），由电子分组交换机（EPS）、非全停光电路交换机（OCS）和全停OCS组成。我们提出了一种统一的多项式时间近似算法，该算法在混合环境中最小化完工时间，且不产生任何分割开销。令$τ$表示网络中所有端口的最大流度，$N$为输入/输出端口数，$m$为网络核心数。在纯EPS环境中，该算法实现了$\min\left\{τ, 2Nm+1\right\}$的近似保证。对于纯非全停和纯全停OCS环境，保证比率分别为$2\min\left\{τ, 2Nm+1\right\}$和$2\min\left\{2τ-1, 2Nm+τ\right\}$。值得注意的是，当专门针对$m=2$设置时，我们的算法摆脱了网络规模的依赖，对于纯EPS和纯非全停OCS分别产生常数界$2$和$4$，对于纯全停OCS产生$2τ+2$。通过利用这些组成界，我们证明了混合架构中的整体性能保证由网络中性能最差的交换机架构上界决定。

英文摘要：

As a prominent network abstraction, coflow models efficiently capture communication patterns in data centers. Since coflow scheduling in large-scale data centers is $\mathcal{NP}$-hard, the existing literature has predominantly focused on limited environments with $m=2$ network cores, relying on flow splitting, which introduces substantial operational overhead. Crucially, no approximation algorithm with provable performance guarantees has been proposed for the more practical, non-splitting coflow scheduling problem, even for the $m=2$ case, let alone for general hybrid architectures. To bridge this critical gap, this paper investigates the non-splitting problem within a hybrid, heterogeneous parallel network featuring multiple network cores ($m \ge 2$) composed of Electronic Packet Switches (EPS), not-all-stop Optical Circuit Switches (OCS), and all-stop OCS. We propose three unified polynomial-time approximation algorithms that minimize the makespan and the total weighted coflow completion time across this hybrid environment without incurring any splitting overhead. Let $τ$ denote the maximum flow degree across all ports in the network, and let $m$ be the number of network cores. To minimize the makespan, our algorithm achieves an approximation ratio of $2\min\left\{2τ-1, m+τ-1\right\}$ in the hybrid architecture. To minimize the total weighted coflow completion time, our algorithm achieves an approximation ratio of $16\min\left\{2τ-1, 2m+τ-1\right\}$ in the hybrid architecture. Moreover, we characterize the approximation ratios of our algorithm under different architectural combinations.

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。

AI总结：

摘要：

英文摘要：

相关论文