In this paper we are interested in exploring the Edge-Fog-Cloud architecture as an alternative approach to the Cloud-based IoT data analytics. Given the limitations of Fog in terms of limited computational resources that can also be shared among multiple analytics with continuous operators over data streams, we introduce a holistic cost model that accounts both the network and computational resources available in the Edge-Fog-Cloud architecture. Then, we propose scheduling algorithms RCS and SOO-CPLEX for placing continuous operators for data stream analytics at the network edge. The former dynamically places continuous operators between the Cloud and the Fog according to the evolution of data streams rates and uses as less as possible Fog computational resources to satisfy the constraints regarding the usage of both computational and network resources. The latter statically places continuous operators between the Cloud and the Fog to minimize the overall computational and network resource usage cost. Based on thorough experiments, we evaluate the effectiveness of SOO-CPLEX and RCS using simulation.