/Cross-edge-Computation-Offloading

A python package of Cross-edge Computation Offloading (CCO) algorithm and its distributed version, Decor for Mobile Edge Computing (MEC).

Primary LanguagePythonMIT LicenseMIT

Cross-edge Computation Offloading

license

A python package of Cross-edge Computation Offloading (CCO) algorithm and its distributed version, Decor for Mobile Edge Computing (MEC).

This package is an implementation of the proposed offloading algorithm for mobility-aware computation-intensive partitionable applications. Specifically, for a non-convex edge site-selection sub-problem, we propose a Sampling-and-Classification-based (SAC) algorithm to obtain the near optimal solution. Based on Lyapunov optimization CCO algorithm is proposed to jointly determine edge site-selection and energy harvesting without a priori knowledge. The transmission, execution and coordination cost, as well as the penalty for task failure, are chosen as performace metrics.

Citation:

[The corresponding paper is under review now. Leave this empty temporarily.]

Used dataset:

EUA dataset @ https://github.com/swinedge/eua-dataset/.

Referenced Code:

RACOS @ https://github.com/eyounx/RACOS.

Quick Start

Code Structure

  • python package racos: This package contains the algorithm named Racos, which is a specific algorithm based on Sampling-And-Classification (SAC) Framework. Details can be found at http://lamda.nju.edu.cn/yuy/research_sal.ashx.

  • directory dataset: This directory contains the real-life dataset of base stations in Melbourne CBD area.

  • python package cross_edge_offloading: This package contains the algorithms proposed in our work. The package named cco is the implementation of main algorithm. The package named benchmarks contains three benchmark policies for comparsion.

Results

Simulation Results #1 Simulation Results #2