Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

An Analysis of MLOps Architectures: A Systematic Mapping Study

Faezeh Amou Najafabadi, Justus Bogner, Ilias Gerostathopoulos, Patricia Lago

TL;DR

The paper addresses the fragmented understanding of MLOps architectures by performing a systematic mapping study over 43 peer-reviewed primary studies (published 2020–2024) and applying card sorting to derive a holistic, component-based view. It identifies 35 architecture components grouped into six categories, and maps 76 tools to these components, revealing how tooling supports different parts of MLOps workflows. The authors describe multiple architecture variants (including baseline, inference-service vs inference-engine, and online-training combinations) and present a UML component diagram to illustrate dependencies, enabling practitioners and researchers to choose suitable variants and identify tool gaps. The work aims to lay groundwork toward a reference MLOps architecture and highlights opportunities for future process-oriented views, architectural decisions, and standardized representations to improve communication and design in this evolving field.

Abstract

Context. Despite the increasing adoption of Machine Learning Operations (MLOps), teams still encounter challenges in effectively applying this paradigm to their specific projects. While there is a large variety of available tools usable for MLOps, there is simultaneously a lack of consolidated architecture knowledge that can inform the architecture design. Objective. Our primary objective is to provide a comprehensive overview of (i) how MLOps architectures are defined across the literature and (ii) which tools are mentioned to support the implementation of each architecture component. Method. We apply the Systematic Mapping Study method and select 43 primary studies via automatic, manual, and snowballing-based search and selection procedures. Subsequently, we use card sorting to synthesize the results. Results. We contribute (i) a categorization of 35 MLOps architecture components, (ii) a description of several MLOps architecture variants, and (iii) a systematic map between the identified components and the existing MLOps tools. Conclusion. This study provides an overview of the state of the art in MLOps from an architectural perspective. Researchers and practitioners can use our findings to inform the architecture design of their MLOps systems.

An Analysis of MLOps Architectures: A Systematic Mapping Study

TL;DR

The paper addresses the fragmented understanding of MLOps architectures by performing a systematic mapping study over 43 peer-reviewed primary studies (published 2020–2024) and applying card sorting to derive a holistic, component-based view. It identifies 35 architecture components grouped into six categories, and maps 76 tools to these components, revealing how tooling supports different parts of MLOps workflows. The authors describe multiple architecture variants (including baseline, inference-service vs inference-engine, and online-training combinations) and present a UML component diagram to illustrate dependencies, enabling practitioners and researchers to choose suitable variants and identify tool gaps. The work aims to lay groundwork toward a reference MLOps architecture and highlights opportunities for future process-oriented views, architectural decisions, and standardized representations to improve communication and design in this evolving field.

Abstract

Context. Despite the increasing adoption of Machine Learning Operations (MLOps), teams still encounter challenges in effectively applying this paradigm to their specific projects. While there is a large variety of available tools usable for MLOps, there is simultaneously a lack of consolidated architecture knowledge that can inform the architecture design. Objective. Our primary objective is to provide a comprehensive overview of (i) how MLOps architectures are defined across the literature and (ii) which tools are mentioned to support the implementation of each architecture component. Method. We apply the Systematic Mapping Study method and select 43 primary studies via automatic, manual, and snowballing-based search and selection procedures. Subsequently, we use card sorting to synthesize the results. Results. We contribute (i) a categorization of 35 MLOps architecture components, (ii) a description of several MLOps architecture variants, and (iii) a systematic map between the identified components and the existing MLOps tools. Conclusion. This study provides an overview of the state of the art in MLOps from an architectural perspective. Researchers and practitioners can use our findings to inform the architecture design of their MLOps systems.
Paper Structure (17 sections, 3 figures, 3 tables)

This paper contains 17 sections, 3 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related work
  3. Methodology
  4. Goal and research questions
  5. Research process
  6. Step 1: Automated initial search
  7. Step 2: Paper selection
  8. Step 3: Data extraction
  9. Step 4: Snowballing
  10. Step 5: Data synthesis
  11. Results
  12. MLOps architecture components and their dependencies (RQ1)
  13. Tools used to support or implement the components (RQ2)
  14. Discussion
  15. Threats to validity
  16. ...and 2 more sections

Figures (3)

  • Figure 1: Overview of the research process
  • Figure 2: UML component diagram of MLOps architecture variants
  • Figure 3: The most popular tools and platforms, mapped to the components