Impact of Distributed Processing on Power Consumption for IoT Based Surveillance Applications

With the rapid proliferation of connected devices in the Internet of Things (IoT), the centralized cloud solution faces several challenges, out of which, power consumption is one of the top priorities among the research community. In this paper, we evaluate the impact of demand splitting over heterogeneous processing resources in an IoT platform, supported by Fog and Cloud infrastructure. We develop a Mixed Integer Linear Programming (MILP) model to study the gains of splitting resource intensive demands among IoT nodes, Fog devices and Cloud servers. A surveillance application is considered, which consists of multiple smart cameras capable of capturing and analyzing real-time video streams. The PON access network aggregates IoT layer demands for processing in the Fog or the Cloud, which is accessed through the IP/WDM network. For typical video analysis workloads, the results show that, splitting medium demand sizes among IoT and Fog resources yields a total power consumption saving of up to 32%, even if these layers can host only 10% of the total workload and this can reach up to 93% for lower number of demands, compared to the centralized cloud solution. However, the gains in power savings from splitting decreases as the number of splits increases.