Hydrospheredata / Hydro Serving

Hydrosphere Serving

Platform for deploying your Machine Learning to production
Check out Hydrosphere.io docs »

Report bug · Contact Us

Hydrosphere Serving is a cluster for deploying and versioning your machine learning models in production.

• Framework Agnostic. Serve machine learning models developed in any language or framework. Hydrosphere Serving will wrap them in a Docker image and deploy on your production cluster, exposing HTTP, gRPC and Kafka interfaces.
• Traffic shadowing. Shadow your traffic between different model versions to examine how different model versions behave on the same traffic.
• Model Version Control. Version control your models and pipelines as they are deployed.

Getting Started

You can refer to our documentation to see tutorials, check out example projects, and learn about all features of Hydrosphere.

Installation

There are two main ways of installing Hydropshere:

• Docker;
• Kubernetes.

Docker

Before installing Hydrosphere Serving, please install its prerequisites:

• Docker 18.0+;
• Docker Compose 1.23+.

To install Hydrosphere Serving, follow the instructions below:

1. Download the latest release version from the releases page;

   export HYDROSPHERE_RELEASE=2.4.3
   wget -O hydro-serving-${HYDROSPHERE_RELEASE}.tar.gz https://github.com/Hydrospheredata/hydro-serving/archive/${HYDROSPHERE_RELEASE}.tar.gz
   

2. Unpack the tar ball;

   tar -xvf hydro-serving-${HYDROSPHERE_RELEASE}.tar.gz
   

3. Set up an environment.

   cd hydro-serving-${HYDROSPHERE_RELEASE}
   docker-compose up
   

To check installation, open http://localhost/. By default Hydrosphere UI is available at port 80.

Note, other installation options are described in the documentation.

Kubernetes

Before installing Hydrosphere Serving, please install its prerequisites:

• Helm 2 starting from 2.9+ with the tiller installed on the cluster;
• Kubernetes 1.8 and above with beta APIs enabled.

To install Hydrosphere Serving, follow the instructions below:

helm repo add hydrosphere https://hydrospheredata.github.io/hydro-serving/helm/
helm install --name serving --namespace hydrosphere hydrosphere/serving

To reach the cluster, port-forward ui service locally.

kubectl port-forward -n hydrosphere svc/serving-ui 8080:9090

To check installation, open http://localhost:8080/.

Note, other installation options are described in the documentation.

Community

Keep up to date and get Hydrosphere.io support via , or contact us directly at [email protected]

Contributing

We'd be glad to receive any help from the community!

Check out our issues for anything labeled with help-wanted, they will be the perfect starting point! If you don't see any, just let us know, we would be happy to hear from you.