History of Kubernetes

Kubernetes did not emerge in a vacuum. It carries over a decade of hard-won operational knowledge from running the world's largest fleets of containers at Google. Understanding this lineage is not merely academic — the architectural decisions made in Borg and Omega explain why Kubernetes works the way it does today, and why certain design choices that seem unusual have deep roots in production experience at scale.

1 · Google Borg (2003–2013)

1.1 Origin and Purpose

Around 2003, Google's infrastructure team built the first version of what would become Borg. The immediate trigger was the rapid growth of Google's internal services — the Web search crawler, Gmail, Maps, YouTube — each of which needed to run on thousands of machines simultaneously. Managing this by hand was impossible.

Borg's design goal was to maximise utilisation of Google's data centre hardware while providing a reliable, scalable runtime for both long-running services (user-facing production jobs) and batch workloads (MapReduce, analytics jobs) on the same shared fleet of machines.

1.2 Borg Architecture

1.3 Borg Key Concepts That Influenced Kubernetes

1.4 Lessons Borg Taught Google (and Kubernetes)

2 · Google Omega (2011–2013)

2.1 What Omega Was

While Borg continued evolving, a separate Google infrastructure team designed Omega as a research prototype to address Borg's architectural limitations. Omega was published academically in the 2013 EuroSys paper "Omega: flexible, scalable schedulers for large compute clusters".

2.2 Omega's Key Innovations

2.3 Architecture Comparison: Borg vs Omega vs Kubernetes

3 · Birth of Kubernetes (2013–2014)

3.1 The Origin Story

In mid-2013, three engineers who had worked on Borg and Omega at Google — Joe Beda, Brendan Burns, and Craig McLuckie — began a skunkworks project internally to build an open-source container orchestrator. Their insight was that Docker's container format was becoming an industry standard, but nobody had built the orchestration layer for it yet.

They were joined by Brian Grant (who led much of the API design), Tim Hockin (networking), and many others. The project was internally code-named "Project Seven" (a reference to Star Trek's Seven of Nine — a Borg character, fittingly).

3.2 Founding Design Principles

The founding team made several explicit architectural decisions that deliberately diverged from Borg:

3.3 First Public Commit and Launch

4 · Kubernetes v1.0 and CNCF (2015)

5 · Major Version Milestones (1.0 → 1.32)

6 · The Docker Shim Removal — A Case Study in API Evolution

One of the most misunderstood events in Kubernetes history is the removal of the Docker shim in v1.24. This section explains exactly what happened and why it was the right decision.

6.1 What Was the Docker Shim?

When Kubernetes introduced the Container Runtime Interface (CRI) in v1.5, Docker did not natively implement CRI. To continue supporting Docker, the Kubernetes team added a shim — a translation layer built into the kubelet that converted CRI calls into Docker API calls. This shim was maintained inside the kubelet source tree.

6.2 Removal Timeline

7 · CNCF Ecosystem Growth

Kubernetes' donation to the CNCF in 2015 seeded an entire ecosystem of complementary projects. The CNCF now has 160+ projects spanning every layer of the cloud-native stack.

7.1 Key CNCF Projects by Layer

7.2 The Container Orchestrator Wars (2015–2017)

Kubernetes did not become the de-facto standard without competition. Three orchestrators competed from 2015 to 2017:

8 · Managed Kubernetes Services

The most significant accelerator for enterprise Kubernetes adoption was managed services — cloud providers abstracting away control plane management entirely.

8.1 On-Premises Distributions

9 · Community Governance and SIGs

Kubernetes is governed by the CNCF Technical Oversight Committee (TOC) and the Kubernetes Steering Committee. The project is divided into Special Interest Groups (SIGs) and Working Groups (WGs), each owning a specific domain.

9.1 Key SIGs and Their Scope

9.2 The KEP Process — How Features Get Added

All significant changes to Kubernetes go through a Kubernetes Enhancement Proposal (KEP). This is a design document that follows a lifecycle:

Provisional → Implementable → Implemented → Deferred / Withdrawn

KEP lifecycle stages:
  Alpha   (opt-in, hidden behind feature gate) — one release minimum
  Beta    (on by default, may have API changes) — two releases minimum
  Stable  (GA, cannot be removed without deprecation period)

Feature gates control alpha/beta features:
  --feature-gates=NewFeature=true     # Enable in kubelet/apiserver

# List all feature gates and their status in your cluster
kubectl get --raw /healthz/ping
# Or check apiserver flags:
ps aux | grep kube-apiserver | grep feature-gates

# Check feature gate defaults for your version
# https://kubernetes.io/docs/reference/command-line-tools-reference/feature-gates/

10 · API Deprecation and Removal Policy

One of the most operationally important aspects of Kubernetes history is its API deprecation policy. Understanding this prevents surprise breakage during upgrades.

# Detect deprecated API usage in your cluster before upgrading
# Option 1: pluto (recommended tool)
pluto detect-all-in-cluster --target-versions k8s=v1.26

# Option 2: Kubernetes built-in audit log with deprecated API filter
kubectl get --raw /metrics | grep apiserver_requested_deprecated_apis

# Option 3: kubent (kube no-trouble)
kubent

11 · Production Implications of Kubernetes History

Understanding this history directly impacts how you run Kubernetes in production:

Next Files to Study

References

• Large-scale cluster management at Google with Borg — Verma et al., EuroSys 2015. The foundational paper.
• Omega: flexible, scalable schedulers for large compute clusters — Schwarzkopf et al., EuroSys 2013.
• Borg, Omega, and Kubernetes — Burns, Grant, Oppenheimer, Brewer, Wilkes. ACM Queue, 2016. The best single overview of the lineage.
• Kubernetes GitHub — github.com/kubernetes/kubernetes — CHANGELOG.md for per-version history
• Kubernetes Enhancement Proposals — github.com/kubernetes/enhancements
• CNCF Landscape — landscape.cncf.io
• Kubernetes release notes — kubernetes.io/docs/setup/release/notes
• The Kubernetes Book — Nigel Poulton (updated annually)