[Platforms] Publish definition of "platforms" for cloud computing

[joshgav]

Hi folks - WG Platforms proposes to publish a doc on the meaning, value, attributes and capabilities of platforms to help enterprise leaders, enterprise architects and platform product managers determine the value of platforms and what to expect and require from them.

We'd like to create this doc in the WG Platforms subdirectory and publish it on the TAG and CNCF blogs.

We expect that contributors will raise issues about the doc once it's in GitHub. We'll also run live interviews after initial publication to continuously update our positions. We'll try to get the initial version published by end of 2022 - that gives us ~70 days.

While this first iteration will be mostly based on the observations and analysis of experts and already-published experience reports, we'll follow up by gathering more experience reports from CNCF end users/enterprises to inform the next iteration, targeting summer 2023.

Work is underway in https://docs.google.com/document/d/1UDL3E5BqPyDdzV1wek9o8xR-nLJQXEEMU1qIMCgq5cI/. That doc is open to all to comment; and if you want to edit click "Share" and then "Ask to share". We'll discuss this live in WG Platforms meetings and asynchronously in the WG Platforms Slack as well as this thread.

The report will discuss the following topics; but this list is just to jog our thinking. What other topics and sections should we include?

Thanks in advance for your contributions and collaboration!

Proposed topics

• What is a platform in relation to cloud-native computing?
  • Platform means many things in societies and systems. What are the key characteristics of a "cloud-native computing" platform? What is its purpose, who does it serve, and how can its impact be measured?
  • Who are platforms intended to serve? For example: application developers, ML model developers, COTS operators, information workers. Is there a common "base" platform or will several platforms be needed in big organizations?
  • What names and descriptors are appropriate for various platforms and what do they imply? Terms found in industry chatter include "developer platform", "developer experience", "cloud platform", "app platform", "delivery platform", "MLOps platform" and more.
• What value do platforms and platform engineering promise cloud-native computing practictioners and enterprises?
  • What are the reasons platforms and "platform engineering" have trended recently in the cloud-native community?
  • In what way (if at all) do they differ from DevOps and Reliability Engineering?
• What components are included in typical platforms? Which components are required and which are optional?
  • Is Kubernetes required?
  • Are build and delivery services required?
  • Is a self-service catalog of infrastructure capabilities required?
  • Should we provide an example platform built of CNCF projects?
• How should a platform team be implemented in an enterprise?
  • What is the relationship between a platform team and its consumers - e.g. product teams?
  • What is the relationship between a platform team and its providers - e.g. managed service providers and big clouds?
  • What does it mean to treat an internal platform as a product?

cc @roberthstrand

[joshgav]

We've nearly reached consensus on the top-level sections of the paper in the current revision; these are as follows. Chime in with your feedback here or in the Google doc.

• What a platform is: a definition of platforms used in cloud-native computing
• Why a platform is valuable: values expected to be attained by organizations adopting a platform for cloud-native computing
• How to measure the success of platforms: metrics to track to ensure delivery of the promised values
• Attributes of successful platforms: non-functional attributes which platforms should provide
• Attributes of successful platform teams: how platform teams and engineers should execute their roles
• Capabilities of platforms: functional capabilities platforms should include

Next step is to create a framework and process to merge and publish this set of sections from a directory in git. This can be inspired by previous work like the following:

• operator white paper: https://github.com/cncf/tag-app-delivery/tree/main/operator-whitepaper/v1
• cloud-native security paper: https://github.com/cncf/tag-security/tree/main/security-whitepaper
• cloud-native maturity model: https://github.com/cncf/cartografos/tree/main/reference

Then within that framework we'll add each of the top-level sections listed above, initially copied from the Google doc.

Once this is completed we'll be able to discuss content and updates and release versions in typical GitHub PRs and issues in this repo.

[joshgav]

Initial draft of this paper is now in a branch in this repo, check it out: https://github.com/cncf/tag-app-delivery/blob/platforms-v1alpha1/platforms-whitepaper/v1alpha1/paper.md. Suggest changes by submitting PRs to that branch.

Discussion will be via the issues in this milestone: https://github.com/cncf/tag-app-delivery/milestone/1. We'd love for folks to take responsibility for finishing up a section, volunteer by commenting in that section's issue.

[joshgav]

Status update: in our last project meeting we agreed that we'd complete milestone 1 - "content complete" - by Feb 28, and lots of good folks have stepped up to help shepherd and finalize each section, see assignments on the milestone's issues. Anyone can suggest changes by submitting a PR to the platforms-v1alpha1 branch.

After 2/28 we'll do a final edit pass, add an executive summary, prep a blog post and get the paper published. We'll track these items in milestone 2 and plan to publish the final version in early April, prior to Kubecon EU on April 17.