Power-Delay Trade-offfor Optimum Data Storage in a Cloud-Fog-Mist Architecture

Abstract

The high growth in Internet-of-Things (IoT) data has resulted in the typical Cloud architecture creating a bottleneck that degrades the quality of Cloud-based services. Fog and Mist computing are proposed to ooad the Cloud by storing and processing data locally and distributing tasks between network layers. The work in this paper investigates the distribution of data in a Cloud-Fog-Mist architecture and studies the influence on power consumption and network delay. A Mixed Integer Linear Programming (MILP) Model is proposed to minimize the power consumption of a Cloud-Fog-Mist architecture by optimizing the location of data files generated at the Mist layer. A second MILP model aims at minimizing network delay by storing data at its optimum location. In addition, we propose a Joint Power-Delay Optimization model to minimize network power consumption while guaranteeing not exceeding an acceptable maximum network delay. The results demonstrate that storing data at the Mist is the most power-efficient solution with maximum savings of 92.3% in power consumption. However, limited processing capabilities result in high processing delay even though data transmission delay within the Mist is negligible. On the other hand, Cloud computing remains the most efficient solution for delay-sensitive applications, regardless of the high energy consumed in the Cloud architecture. A power-delay trade-off can be achieved where power is minimized without exceeding acceptable delay.