Abstract
Fog computing is a promising technology that leverages the resources to provide services for requests of IoT (Internet of Things) devices at the cloud edge. The high dynamic and heterogeneous nature of devices at the cloud edge causes failures to be a popular event and therefore fault tolerance became indispensable. Most early scheduling and fault-tolerant methods did not highly consider time-sensitive requests. This increases the possibility of latencies for serving these requests which causes unfavorable impacts. This paper proposes a fault-tolerant scheduling method (FTSM) for allocating services' requests to the most sufficient devices in fog-cloud IoT-based environments. The main purpose of the proposed method is to reduce the latency and overheads of services and to increase the reliability and capacity of the cloud. The method depends on categorizing devices that can issue requests into three classes according to the type of service required. These classes are time-sensitive, time-tolerant and core. Each time-sensitive request is directly mapped to one or more edge devices using a pre-prepared executive list of devices. Each time-tolerant request may be assigned to one or more devices at the cloud edge or the cloud core. Core requests are assigned to resources at the cloud core. In order to achieve fault tolerance, the proposed method selects the most suitable fault-tolerant technique from replication, checkpointing and resubmission techniques for each request while most existing methods consider only one technique. The effectiveness of the proposed method is assessed using average service time, throughput, operation costs, success rate and capacity percentage as performance indicators.