[docs/blog] Write technical blog: Control Plane vs Data Plane deep dive

[mesutoezdil]

Context

This issue is part of the HAMi v2.9.0 Documentation Revamp Plan (WS8 - New Technical Content, Blog 1).

This blog post explains how HAMi actually works at runtime: what the control plane does, what the data plane does, and how a GPU sharing request flows through the system from pod creation to execution.

Before you start

Please reach out to the maintainers before picking this up. The architecture narrative in this post must align with #418 (README architecture section) and the diagram work in #413.

• @rootsongjc (author, owns architecture narrative)
• @mesutoezdil (revamp reviewer, scope alignment)

Dependency

Depends on #418 (README architecture positioning). Do not start writing before the architecture narrative in the README is settled.

Scope

Write a technical blog post covering:

• What the HAMi control plane is responsible for (scheduling decisions, resource accounting)
• What the HAMi data plane is responsible for (enforcement, isolation, runtime interception)
• A step-by-step walkthrough of a GPU sharing request: from pod spec to actual GPU time slice
• How this differs from raw GPU passthrough and from MIG
• A data flow diagram (coordinate with [docs/diagram] Inventory all existing diagrams across the repository #413 for tooling)

Target audience: Kubernetes engineers evaluating HAMi or trying to understand how it works internally.

Acceptance criteria

• Blog post added under blog/ following existing post format and frontmatter
• Data flow diagram included (SVG preferred, coordinate with WS4 tooling)
• No first-person language
• No marketing language
• No emoji
• Code blocks have language specifiers
• Admonitions use Docusaurus syntax
• DCO sign-off required on every commit (git commit -s)