Work in Progress: Power-aware Scheduling Strategy for Multiple DAGs in the Heterogeneous Cloud

摘要

High energy consumption has become a major problem of cloud platform. Most of the current task scheduling methods neglect the heterogeneity of cloud platform, which may consume more power consumption of heterogeneous cloud platform. In this paper, we propose a power-aware scheduling strategy (PASS) for multiple DAGs workflow in the heterogeneous cloud with the goal of minimizing the energy consumption. First, we predict the PM energy consumption considering VM status after scheduling tasks, and then we formulate the power-aware DAGs task scheduling as a NP-hard problem, which tries to minimize the energy consumption of heterogeneous cloud platform. Second, we propose a multiple DAGs workflow scheduling algorithm to solve the formulated NP-hard problem. We consider the combination of the coarse-grained sorting for DAGs and the fine-grained sorting for sub-tasks to obtain the optimal sorting of DAG workflows. We then assign tasks to the appropriate computing nodes considering the heterogeneity of the cloud platform to minimize its energy consumption. Third, experiments are conducted to evaluate PASS, and the experimental results verify its efficiency.