Scheduling Coflows With Dependency Graph

Applications in data-parallel computing typically consist of multiple stages. In each stage, a set intermediate parallel data flows ( Coflow ) is produced and transferred between servers to enable starting next stage. While there has been much research on scheduling isolated coflows, the dependency coflows multi-stage jobs largely ignored. this paper, we consider represented by general xmlns:xlink="http://www.w3.org/1999/xlink">DAG s (Directed Acyclic Graphs) shared center network, so as minimize total weighted completion time jobs. This problem significantly more challenging than traditional coflow scheduling, even single job its shown be NP-hard. propose polynomial-time algorithm with approximation ratio $O(\mu \log (m)/\log (\log (m)))$ , where notation="LaTeX">$\mu $ maximum number notation="LaTeX">$m$ servers. For special case that jobs’ underlying graphs are xmlns:xlink="http://www.w3.org/1999/xlink">rooted trees modify improve ratio. To verify performance our algorithms, present simulation results using real traffic traces show up 53% improvement over prior approach. We conclude paper providing result concerning an optimality gap for DAGs.