Skip to content

Caribou is a framework for geo-distributed deployment of serverless workflows to save carbon emissions.

License

Notifications You must be signed in to change notification settings

ubc-cirrus-lab/caribou

Repository files navigation

Caribou

Build Status GitHub license

Caribou is a framework for running and deploying complex serverless workflows multi-constraint and multi-objective aware on hyper-scalers over multiple regions.

⚡️ Quickstart

See the Quick Start guide.

 Introduction

Caribou is a framework for running and deploying complex serverless workflows multi-constraint and multi-objective aware on hyper-scalers over multiple regions. The framework is designed to be self-adapting and self-optimizing with regards to the carbon footprint, cost, and performance of the workflows. A workflow that is developed and deployed using the Caribou Python package can be run and solved by the framework, which will then automatically, depending on constraints and objectives, adapt the workflow deployment to optimize the workflow's performance, cost, and carbon footprint. The priority of optimization is determined by the objectives set by the workflow developer in the corresponding workflow deployment manifest. Optimization is done, when warranted by the type of application and frequency of its invocation, by the deployment solver that solves for new multi-region deployment configurations. The overhead of the system plays a crucial role in the optimization process and sets the frequency as well as the granularity of the optimization process.

 Example

An example workflow can be found in examples/small_sync_example including the source code of the workflow in the app.py file as well as the deployment manifest in the .caribou directory. More information on the example workflow can be found in the respective README.

See the Installation section on how to get set up and the Quick Start guide on how to run the example workflow.

 Installation

See the Installation guide.

Testing

See the Testing guide.

Benchmarks

See the Benchmarks guide.

Troubleshooting

See the Troubleshooting guide.

Paper

Our paper on Caribou has been recently accepted to the 30th ACM Symposium on Operating Systems Principles (SOSP 2024). It will be presented in November 2024. If you use Caribou in your research, please cite our paper:

V. Gsteiger, P. H. Long, Y. Sun, P. Javanrood, and M. Shahrad, "Caribou: Fine-Grained Geospatial Shifting of Serverless Applications for Sustainability", in Proceedings of the 30th ACM Symposium on Operating Systems Principles (SOSP 2024), 2024.

Architecture

See the architecture outline for more information on the framework.

About

Caribou is being developed at the Cloud Infrastructure Research for Reliability, Usability, and Sustainability Lab at the University of British Columbia. If you have any questions or feedback, please open a GitHub issue.

 Contributing

See the Contributing guide.

License

Apache License 2.0