Reference Implementation of paper "Distributed Analytics in Fog Computing Platforms Using TensorFlow and Kubernetes"
The below photo is credited to the author of the paper
Tensorflow 1.1.1
Kubeadm 1.10.2
kubelet 1.10.2
docker 1.13.1
Application/yolo/
single/ //write yolo for one machine
distributed/ //break yolo into three part, which could be handled by different devices
docker-image/ //contains final distibuted yolo version and required Dockerfile
Deployment/
helmChart/yolo/ //contains params to be passed during execution and worker.yaml, master.yaml 1_node.sh 2_node.sh 3_node.sh
Rpi_ENV/
describe my steps to setup a cluster with one pc as master and several Rpis as workers
Experiment/
contains several files that I have experimented before
Because I use a router shared all wifi device in my Lab, therefore the Internet Overhead is big! In the results shown above I just doubled the CPU & Memory Resources that each operator could use on the same Rpi to simulate the 2 devices speed.