EcoMobiFog–Design and Dynamic Optimization of a 5G Mobile-Fog-Cloud Multi-Tier Ecosystem for the Real-Time Distributed Execution of Stream Applications
By Enzo Baccarelli, Michele Scarpiniti, Alireza Momenzadeh. 2019
36 citations recorded by the source corpus.
The emerging 5G paradigm will enable multi-radio smartphones to run high-rate stream applications. However, since current smartphones remain resource and battery-limited, the 5G era opens new challenges on how to actually support these applications. In principle, the service orchestration capability of the Fog and Cloud Computing paradigms could be an effective means of dynamically providing resource-augmentation to smartphones. Motivated by these considerations, the peculiar focus of this paper is on the joint and adaptive optimization of the resource and task allocations of mobile stream applications in 5G-supported multi-tier Mobile-Fog-Cloud virtualized ecosystems. The objective is the minimization of the computing-plus-network energy of the overall ecosystem under hard constraints on the minimum streaming rate and the maximum computing-plus-networking resources. To this end: 1) we model the target ecosystem energy by explicitly accounting for the virtualized and multi-core nature of the Fog/Cloud servers; 2) since the resulting problem is non-convex and involves both continuous and discrete variables, we develop an optimality-preserving decomposition into the cascade of a (continuous) resource allocation sub-problem and a (discrete) task-allocation sub-problem; and 3) we numerically solve the first sub-problem through a suitably designed set of gradient-based adaptive iterations, while we approach the solution of the second sub-problem by resorting to an ad-hoc-developed elitary Genetic algorithm. Finally, we design the main blocks of EcoMobiFog, a technological virtualized platform for supporting the developed solver. The extensive numerical tests confirm that the energy-delay performance of the proposed solving framework is typically within a few per-cent the benchmark one of the exhaustive search-based solution.
Subfields: Computer Networks and Communications, Electrical and Electronic Engineering, ML Systems & Hardware
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Computer Networks and Communications
In Electrical and Electronic Engineering
Citations in the corpus
- Why Should We Add Early Exits to Neural Networks? (2020, background): “Alternatively, more sophisticated distributed optimization strategies, possibly taking into account also communication constraints, are an open interesting research area [ 6, 4 ] .”
- Differentiable Branching In Deep Networks for Fast Inference (2020, background): “In addition, this class of ‘branching’ neural networks can be distributed across multiple tiers of distributed computing platforms, making them a perfect fit for scenarios in Fog computing [13] and 5G [3,14].”