Rampart

Rampart is an open system built on top of Kubernetes for MLOps that allows easy provisioning of custom infrastructure from templates. It is designed with the principles of composability, automation, and scalability, for developers to enable simple infrastructure management, modular pipeline development, seamless deployment, and experiment management.

Developers can use the drag-and-drop GUI to spin up infrastructure from a template and get started on a new project in minutes, without any help from a system admin. As an extensible system for interoperability, Rampart can be used to deploy any software (i.e., OSS, third-party software, or any custom code) in addition to a library of native Rampart components as parts of custom infrastructure.

Its Infrastructure-as-a-Graph approach represents commonly used AIML application patterns as components, such as Batch Data Uploader, Batch Executor, Ray cluster and Tensorboard visualizer. Components are connected through “flows” which can be NFS volumes, pulsar streaming or repositories. Infrastructure graphs can be composed quickly and replicated, customized, deployed, or updated live.

Rampart is one of a series of tools that make up the MLOps OSS repertoire that Petuum releases. We are excited to introduce more useful tools to the developer community in the coming days.

Installation

This is covered by a separate repo which is coming soon.

Usage

Rampart does many things. The quickest way to familiarize with Rampart is by creating an MLOps deployment, which you can usually start with deploying a graph. You can import a graph represented in the YAML below:

apiVersion: rampart.petuum.com/v1beta2
kind: RampartGraph
metadata:
  name: demo
spec:
  graph:
    components:
      batch-uploader:
        type: batch-uploader
        chart:
          name: batch-uploader
          repo: rampart-repo
          version: latest
        config:
          imagePullSecrets:
            - name: stagingsecret
        inputs: {}
        outputs:
          output_1:
            type: volume
            flow: batch-input
    flows:
      batch-input:
        type: volume
        config:
          size: 1Gi

This is a simple Rampart Graph demonstrating how a component called Batch Uploader is connected to a Volume Flow named batch-input.

Note: A Rampart Graph is deployed on a Kubernetes cluster as a custom resource (CR) named RampartGraph. The following command requires the kubectl command line tool to run.

Run this code to deploy this graph:

kubectl apply -f <path/to/graph.yaml> -n default

Go to Rampart UI at http://<node-ip>:30800/

Use

curl -F "files[]=@<file-path>" -F "batch-name=1" -F "edge-name=output_1" http://<node-ip>:30800/default/demo/batch-uploader/multipart/upload-batch-atomic/

to upload a file. Where <file-path> is the path to any file you want to upload, and <node-ip> is the IP address of a cluster node. “succeeded” will be printed on the terminal when file uploading is completed.

Contributing

Contributions are welcome! If you would like to contribute to Rampart OSS, please follow these steps:

1. Fork the repository.
2. Make your changes.
3. Submit a pull request.

Open Source Progress

• Rampart controller (this repo)
• [] Rampart helm chart
• [] Rampart GUI
• [] Rampart components

Developer Documentation

See the documentation for the Rampart architecture, Rampart Controller and Rampart Backend Service within this repository for details.