The elastic resource provision, no interfering resource sharing and flexible customized configuration provided by the Cloud infrastructure has shed light on efficient execution of many scientific applications modeled as Directed Acyclic Graph (DAG) structured workflows. However, the energy cost on running the increasingly deployed Cloud data centers around the globe together with the amount of CO2 emissions have skyrocketed. In order to maintain sustainable Cloud computing facing with ever-increasing problem complexity and big data size, we propose an energy-efficient scientific workflow scheduling algorithm to minimize energy consumption and CO2 emission while satisfying certain Quality of Service (QoS). Our multiple-step resource provision and allocation algorithm applies Dynamic Voltage and Frequency Scaling (DVFS) technology to reduce energy consumption within acceptable performance bounds, and minimize the Virtual Machine (VM) overhead for further reduced energy consumption and higher resource utilization rate. The candidacy of multiple data centers from the energy and performance efficiency perspectives is also evaluated. The simulation results show that our algorithm is able to achieve an average up to 30% of energy savings and increase the resource utilization rate for about 25% leading to higher profit and less CO2 emissions.