Application Placement in the Cloud Continuum With Resource Overbooking

The pervasiveness of IoT devices in many real-world domains, such as healthcare, agriculture, manufacturing, transportation, offers new interesting opportunities, but at the same time opens new challenges. In fact, to ensure the effectiveness of the innovative smart applications being deployed nowadays in these ecosystems, the mapping between applications and resources of the cloud continuum plays a key role. In this context, we propose a lightweight placement policy aimed at minimizing the network delays experienced by IoT applications by fully exploiting the resources of the cloud continuum. For this purpose, our heuristic tries to allocate application modules as close as possible to their data sources. In addition, by introducing the concept of resource overbooking, a module can be allocated to a resource even if the desired processing capacity of the module exceeds to some extent the available processing capacity of the resource. The results of the extensive simulation experiments have clearly demonstrated the benefits of our policy. For example, as the overbooking factor increases, the network delay tends to decrease even significantly. In contrast, the potential contentions caused by resource overbooking introduce a slight increase in the processing time although it does not generally exceed the desired processing time of the applications, thus ensuring their QoS requirements.