GitLab: The Reference Architecture for the Complete DevOps Platform

TL;DR

Modern software development requires more than just code hosting. While hyperscalers like AWS attempt to lock developers into their platforms with a fragmented chain of individual services (CodeCommit, CodeBuild, CodePipeline), GitLab follows the ‘Single Application’ approach. It combines Source Code Management (SCM), CI/CD, Security Scanning, and Package Registry into a single, coherent interface. This reduces complexity, accelerates feedback loops, and ensures that your intellectual property (the code) and processes remain portable.

1. The Architecture Principle: The One DevOps Platform

In many companies, the toolchain resembles a patchwork: GitHub for code, Jenkins for CI, Jira for tickets, Artifactory for packages. This requires constant context switching and complex integrations.

GitLab eliminates these friction losses through an integrated architecture.

• Seamless CI/CD: The pipeline definition (.gitlab-ci.yml) is directly with the code. There is no separation between repository and build server. A push immediately triggers the pipeline, and the result is directly visible in the merge request.
• Review Apps: GitLab can automatically spin up a temporary live environment in the Kubernetes cluster for each branch. Developers and PMs test changes at a real URL before the code is merged.

2. Core Feature: GitLab Runner on Kubernetes

The heart of automation is the GitLab Runner. Unlike monolithic Jenkins servers, which are difficult to scale, the GitLab Runner integrates natively with Kubernetes.

• Elastic Scaling: A new, isolated pod in the cluster is started for each job. If 50 developers push simultaneously on a Friday afternoon, the cluster scales up automatically. Once the jobs are finished, the pods disappear.
• Security: Each job runs in a clean container. There are no “legacy” or artifacts from previous builds that could distort the result.

3. Container Registry & Security Scanning

GitLab is not just a code repository but also an artifact repository. The integrated container registry allows storing Docker images directly in the project—without needing to configure external services like Docker Hub or AWS ECR.

Additionally, GitLab offers (depending on the edition) deep DevSecOps functions: SAST (Static Application Security Testing), Secret Detection, and Dependency Scanning run automatically in the pipeline, blocking insecure code before it goes into production.

4. Operating Models Compared: AWS CodeSuite vs. ayedo Managed GitLab

Here, the decision is whether your development processes are tied to a cloud provider or if you retain control over your SDLC (Software Development Life Cycle).

Scenario A: AWS CodeSuite (The Fragmented Dependency)

Those who rely on AWS CodeCommit, CodeBuild, and CodePipeline enter into a deep lock-in.

• Proprietary Logic: The pipeline definitions (buildspec.yml) and deployment logics are specific to AWS. Switching to another provider or on-premise is impossible without rewriting the entire CI/CD logic.
• Lack of Developer Experience: AWS tools are built for Ops, not Devs. Code reviews (pull requests) in CodeCommit are rudimentary compared to the powerful merge request interface of GitLab.
• Strategic Risk: AWS has announced closing CodeCommit for new customers. This clearly shows: AWS does not prioritize these services. Building on this is building on a declining branch.

Scenario B: GitLab with Managed Kubernetes by ayedo

In the ayedo app catalog, GitLab (or the GitLab Runner) is provided as the central DevOps engine.

• Full Data Sovereignty: Your code, your tickets, and your pipelines reside in your cluster (Self-Managed GitLab) or you use GitLab.com with your own runners in your cluster. In both cases, the data and processes belong to you.
• Portable Pipelines: A .gitlab-ci.yml is standard. You can use the same process to deploy to AWS, Azure, Google Cloud, or on bare metal.
• Unified Experience: Developers have one tool for everything. This increases productivity and significantly reduces the onboarding time for new employees.

Technical Comparison of Operating Models

FAQ: GitLab & DevOps Strategy

Should I host GitLab myself (Self-Managed) or use SaaS?

This depends on your compliance requirements. The SaaS version (GitLab.com) is maintenance-free. However, with Self-Managed GitLab (deployed via ayedo), you have absolute control: The code never leaves your infrastructure, and you can run GitLab behind your VPN/firewall. This is often mandatory for highly regulated industries (banks, health). A hybrid approach is also possible: SaaS for the UI, but self-hosted runner in your own cluster for job execution.

How do I migrate from Jenkins to GitLab CI?

Jenkins is often a maintenance nightmare (“plugin hell”). Migrating to GitLab CI is almost always worthwhile. The logic shifts from the click-based Jenkins configuration to the versioned .gitlab-ci.yml. GitLab offers keyword equivalents for almost all Jenkins functions. The biggest gain is the disposable environment of the runners: Never again “It works on my machine but not on Jenkins,” as each job runs in a fresh Docker container.

Isn’t GitLab too resource-intensive for Kubernetes?

The GitLab core (Rails monolith) indeed requires substantial resources (RAM). However, the runners are extremely efficient. In the ayedo stack, the deployment is optimized so that components (Gitaly, Sidekiq, Postgres) are best distributed across the nodes. For smaller teams, the model is often recommended: Code on GitLab.com, but the runners (compute) in your own ayedo cluster.

Does GitLab support GitOps (like ArgoCD)?

Yes. GitLab has an integrated “Agent for Kubernetes” that offers GitOps functions. However, the common best practice in the industry (and in the ayedo stack) is the combination: GitLab CI for Continuous Integration (testing, building, security scan) and ArgoCD/Flux for Continuous Delivery (deployment). GitLab pushes the finished image, and ArgoCD synchronizes it. That is “Best of Breed.”

Conclusion

Code is the heart of digital value creation. Binding this core and the processes around it (CI/CD) to proprietary AWS tools results in a loss of innovation speed and flexibility. GitLab offers the only true end-to-end platform that developers love and Ops teams find transparent. With the ayedo managed stack, you get a robust, scalable DevOps environment that ensures your most important asset—your code and automation—remains portable and under your control at all times.