Improved Application Placement in Fog Environments Through Parallel Collaboration

Abstract

The real-time Quality of Service (QoS) constraints of Internet-of-Things (IoT) applications have motivated the development of distributed systems environments such as fog computing. A challenging issue in fog computing is to identify placement plans by which the IoT applications are placed on fog devices for execution, a problem known as the application placement problem. With a potentially huge number of fog devices and IoT applications, it makes sense to solve this problem in a decentralized manner, where an independent placement problem is solved in parallel for different clusters of fog devices. To cope with the highly dynamic nature of the IoT networks, clusters lacking a sufficient number of fog devices to accommodate all the submitted IoT applications can propagate undeployed applications to other clusters for execution, potentially leading to uncertain fulfillment of QoS constraints. The challenge becomes how to enhance the probability of placing the propagated IoT applications within their predefined QoS constraints, i.e., response time delays. This paper proposes an application placement approach, called PColl, that aims to minimize the total number of delay violations for applications, specifically for the propagated applications, through parallel collaboration. The proposed approach utilizes a heuristic-based algorithm that enhances the probability of placing the propagated applications within their delays through parallel collaboration. This is achieved by duplicating the placement requests of these applications, which enables parallel searching for fog devices. Additionally, this paper proposes a collaboration algorithm where various architecture entities collaborate to manage the problem of parallel searching. The experimental results demonstrate the effectiveness of the proposed approach in minimizing the percentage of applications that experience delay violations, as compared to other placement approaches.

Keywords

• Application/service placement
• fog clusters
• fog computing
• propagated applications
• delay-sensitive applications
• iFogsim