A $(2+\varepsilon)$-approximation algorithm for preemptive weighted flow time on a single machine

TL;DR

This paper presents a new polynomial-time algorithm that approximates the minimum weighted flow time on a single machine within a factor of 2+epsilon, improving previous bounds and avoiding LP-rounding.

Contribution

It introduces a novel reduction to a geometric problem and a dynamic programming approach, achieving a 2+epsilon approximation without LP-rounding.

Findings

Achieves a 2+epsilon approximation ratio.

Reduces the problem to a geometric problem with minimal loss.

Avoids LP-rounding by leveraging structural properties.

Abstract

Weighted flow time is a fundamental and very well-studied objective function in scheduling. In this paper, we study the setting of a single machine with preemptions. The input consists of a set of jobs, characterized by their processing times, release times, and weights and we want to compute a (possibly preemptive) schedule for them. The objective is to minimize the sum of the weighted flow times of the jobs, where the flow time of a job is the time between its release date and its completion time. It had been a long-standing open problem to find a polynomial time $O(1)$-approximation algorithm for this setting. In a recent break-through result, Batra, Garg, and Kumar (FOCS 2018) found such an algorithm if the input data are polynomially bounded integers, and Feige, Kulkarni, and Li (SODA 2019) presented a black-box reduction to this setting. The resulting approximation ratio is a (not explicitly stated) constant which is at least $10.000$. In this paper we improve this ratio to $2+\varepsilon$. The algorithm by Batra, Garg, and Kumar (FOCS 2018) reduces the problem to Demand MultiCut on trees and solves the resulting instances via LP-rounding and a dynamic program. Instead, we first reduce the problem to a (different) geometric problem while losing only a factor $1+\epsilon$, and then solve its resulting instances up to a factor of $2+\epsilon$ by a dynamic program. In particular, our reduction ensures certain structural properties, thanks to which we do not need LP-rounding methods. We believe that our result makes substantial progress towards finding a PTAS for weighted flow time on a single machine.