A $(2+\epsilon)$-approximation for scheduling parallel jobs in platforms

to full text (284 kB)   ZIP
involved person(s) / institution(s)author :
datePublished :
  • February 2013
size33 S.

We consider the problem of \textsc{Scheduling parallel Jobs in heterogeneous Platforms}: We are given a set $\mathcal{J}=\{1,\ldots,n\}$ of $n$ jobs, where a job $j\in\mathcal{J}$ is described by a pair $(p_j,q_j)$ of a processing time $p_j\in\mathbb{Q}_{>0}$ and the number of processors $q_j\in\mathbb{N}$ that are required to execute $j$. We are also given a set $\mathcal{B}$ of $N$ heterogeneous platforms $P_1,\ldots,P_N$, where each $P_i$ contains $m_{i}$ processors for $i\in\{1,\ldots, N\}.$ The objective is to find a schedule for the jobs in the platforms minimizing the makespan, i.e. the latest finishing time of a job. Unless $\mathcal{P}=\mathcal{NP}$ there is no approximation algorithm with absolute ratio strictly better than $2$ for the problem. We give a $(2+\epsilon)$-approximation for the problem improving the previously best known absolute approximation ratio $3$.
Static URLhttps://www.uni-kiel.de/journals/receive/jportal_jparticle_00000035
 
URN:NBNurn:nbn:de:101:1-201303157479
IDNumber of report :
  • TR_1217