This repository contains the software used for evaluations in Digital Ocean cloud environment.
This software is used in the following paper:
Slack Squeeze Coded Computing for Adaptive Straggler Mitigation
Krishna Giri Narra, Zhifeng Lin, Mehrdad Kiamari, Salman Avestimehr and Murali Annavaram
arXiv:1904.07098
This software is a proof-of-concept meant only for performance evaluations of the Slack Squeeze Coded Computing(S2C2) framework.
The workload to be evaluated is to be dockerized first and then run on a kubernetes cluster.
For bootstraping a k8s cluster, please refer to this link for more information -- https://kubernetes.io/docs/setup/production-environment/tools/kubeadm/create-cluster-kubeadm/.
Once a k8s cluster is up and running, please refer to the README in the k8s/ folder for how to launch the workload to the cluster for evaluations.
coded_computation_docker/codedComputation.Dockerfile is the Dockerfile used to dockerize the workloads.
coded_computation_docker/apps contains the workloads used for evaluations.
coded_computation_docker/apps/s2c2 contains the codes for S2C2 workload used for evaluations.
coded_computation_docker/apps/static contains the codes for conventional MDS workload used for evaluations.
The dataset, that is computed on, needs to be encoded into multiple partition files and stored. Please follow the naming in corresponding files inside apps/ folder.
MIT License