Efficient Algorithms for Scheduling Moldable Tasks

TL;DR

This paper introduces a new model for scheduling moldable tasks that combines linear and monotonic speedup behaviors, and proposes approximation algorithms with improved bounds for makespan minimization and throughput maximization.

Contribution

It presents a novel hybrid speedup model for moldable tasks and develops approximation algorithms with better bounds than existing methods.

Findings

Achieves a pproximation for makespan minimization with improved bounds.

Provides a pproximation for throughput maximization under a common deadline.

Demonstrates the model's generality between classic monotonic and linear-speedup models.

Abstract

We study the problem of scheduling $n$ independent moldable tasks on $m$ processors that arises in large-scale parallel computations. When tasks are monotonic, the best known result is a $(\frac{3}{2}+\epsilon)$-approximation algorithm for makespan minimization with a complexity linear in $n$ and polynomial in $\log{m}$ and $\frac{1}{\epsilon}$ where $\epsilon$ is arbitrarily small. We propose a new perspective of the existing speedup models: the speedup of a task $T_{j}$ is linear when the number $p$ of assigned processors is small (up to a threshold $\delta_{j}$) while it presents monotonicity when $p$ ranges in $[\delta_{j}, k_{j}]$; the bound $k_{j}$ indicates an unacceptable overhead when parallelizing on too many processors. The generality of this model is proved to be between the classic monotonic and linear-speedup models. For any given integer $\delta\geq 5$, let $u=\left\lceil \sqrt[2]{\delta} \right\rceil-1\geq 2$. In this paper, we propose a $\frac{1}{\theta(\delta)} (1+\epsilon)$-approximation algorithm for makespan minimization where $\theta(\delta) = \frac{u+1}{u+2}\left( 1- \frac{k}{m} \right)$ ($m\gg k$). As a by-product, we also propose a $\theta(\delta)$-approximation algorithm for throughput maximization with a common deadline.