[Paper Review] Machine Learning Operations (MLOps): Overview, Definition, and Architecture

Dominik Kreuzberger, Niklas Kühl+1 more|arXiv (Cornell University)|May 4, 2022

Industrial Vision Systems and Defect Detection50 citations

TL;DR

This paper provides an aggregated overview of MLOps, offering a definition, architecture, principles, components, roles, workflows, and open challenges based on mixed-method research.

ABSTRACT

The final goal of all industrial machine learning (ML) projects is to develop ML products and rapidly bring them into production. However, it is highly challenging to automate and operationalize ML products and thus many ML endeavors fail to deliver on their expectations. The paradigm of Machine Learning Operations (MLOps) addresses this issue. MLOps includes several aspects, such as best practices, sets of concepts, and development culture. However, MLOps is still a vague term and its consequences for researchers and professionals are ambiguous. To address this gap, we conduct mixed-method research, including a literature review, a tool review, and expert interviews. As a result of these investigations, we provide an aggregated overview of the necessary principles, components, and roles, as well as the associated architecture and workflows. Furthermore, we furnish a definition of MLOps and highlight open challenges in the field. Finally, this work provides guidance for ML researchers and practitioners who want to automate and operate their ML products with a designated set of technologies.

Motivation & Objective

• Define MLOps and clarify its scope for researchers and practitioners.
• Aggregate principles, components, and roles from literature, tools, and expert input.
• Describe architecture and workflows that enable automated ML product operation.
• Identify open challenges and provide guidance for adopting MLOps technologies.

Proposed method

• Conduct a literature review to synthesize existing concepts and practices.
• Perform a tool review to assess available MLOps technologies and ecosystems.
• Carry out expert interviews to capture practitioner perspectives.
• Provide an aggregated overview of principles, components, and architectures.
• Offer a definition of MLOps based on collected evidence.

Experimental results

Research questions

• RQ1 What constitutes MLOps and how should it be defined for research and practice?
• RQ2 What are the essential principles, components, and roles in MLOps?
• RQ3 What architectures and workflows enable reliable operation of ML products?
• RQ4 What open challenges remain in applying MLOps in real-world settings?

Key findings

• MLOps comprises principles, concepts, and development culture surrounding ML product deployment and operation.
• An aggregated set of necessary components, roles, architectures, and workflows can guide implementation.
• Open challenges in MLOps include standardization, tooling integration, and operationalizing ML in production.
• The paper provides guidance for researchers and practitioners on selecting and applying technologies to automate and operate ML products.