Addressing the topic of ‘Core’ through Spider

 

August 23, 2013

The word “core” carries with it a wide range of meaning and implication for those involved within the OpenStack community. What we’ve discovered through ongoing discussions is that the goals of one audience simply aren’t necessarily the goals of the other audiences – in fact, in some cases, they’re opposed!

We first began the journey to refine the definition of core through a work effort called IncUp.  IncUp was a combined effort of TC and board members that resulted with improvements to the Incubation process. That effort also helped draw out the complex nature of defining core.  The complexity is due to the varied impact core has upon OpenStack users and contributors, as well as the ecosystem as a whole.

Addressing The Problem
On assignment from the Board, Rob Hirschfeld and I began meeting to further the progress of defining core. The first step in this process was to examine and define the goals of each audience, which Rob detailed in a series of blogs[1] outlining the thought processes we went through.

Shortly after we began this process, we quickly realized that the goals amongst these various audiences would either be aligned or in tension. Using different colors to map the goals and tensions on a whiteboard, we created a massive multi-colored “spider” graph, which became more of a mind map. Thus, we refer to the current effort as the “spider discussions” or simply spider.

Problem Statement/Goal
At the highest level, this spider effort is about advancing the OpenStack mission of producing the ubiquitous Open Source cloud computing platform. The key efforts within this process include defining the set of components, processes and criteria needed to protect the brand, provide users a good experience, and reinforce a collaborative community.OpenStack marks are the tools that the Foundation has to define and defend those keys.
In order for OpenStack implementers to use the OpenStack mark, the Board was chartered to provide specific and verifiable criteria. However, these do not currently exist in a usable form.  Thus, our “what is core” discussions and efforts seek to determine those criteria.
Statements of Position
Spider focuses on breaking down those criteria of core into referenceable position statements.  Those statements will form a basis to draw upon so that we may keep the decision making – and eventual implementation – on a progressive track.There are currently 10 evolving position statements:1.     Implementations that are core can utilize the OpenStack trademark (OpenStack™)
1.     This is the legal definition of core, as well as why it matters to  the community.

  1.  We want to make sure that the OpenStack™ mark means something.
  2. The OpenStack™ mark is not the same as the OpenStack brand; rather, the Board uses its control of the mark as a proxy to help manage the brand.

2.     Core is a subset of the whole project

  1. The OpenStack project is intended to be a broad and diverse community with new projects entering incubation, with new implementations frequently added. This innovation is vital to OpenStack but separate from the definition of core.
  2. There may be other marks that are managed separately by the foundation and available for the platform ecosystem at the Board’s discretion.
  3. The “OpenStack API Compatible” mark is not part of this discussion and should be not be assumed.
3.     Core definition can be applied equally to all usage models

  1.  There should not be multiple definitions of OpenStack, regardless of the operator (public, private, community, etc.).
  2. While expected that each deployment is identical, the differences must be quantifiable.

4.     Claiming OpenStack requires use of designated code frameworks

  1. Implementations claiming the OpenStack™ mark must use the approved API framework code.\
  2. Entities which only pass all the tests but do not use the API framework will not be considered or be approved as part of OpenStack.This prevents entities from using the API without joining the community, as well as surfaces bit-rot in alternate implementations to the larger community.
  3. By implementing this requirement,  interoperability is greatly improved, as there is more shared code between implementation

5.     Projects must have an open reference implementation

  1. OpenStack will require an open source reference base plug-in implementation for all projects (if not part of OpenStack, license model for reference plug-in must be compatible).
  2. We define a plug-in as alternate backend implementations with a common API framework that use common _code_ to implement the API.
  3. This establishes the expectation that projects (where technically feasible) will implement plug-in or extension architecture.
  4. This is already in place for several projects and addresses ecosystem support, further enabling innovation.
  5. Reference plug-ins are, by definition, the complete capability set. It is not acceptable to have core features that are not functional in the reference plug-in.
  6. This will enable alternate implementations to offer innovative or differentiated features without forcing changes to the reference plug-in implementation.
  7. This will also enable the reference to expand without forcing other alternate implementations to match all features and recertify.

6.     Vendors may substitute alternate implementations

  1. If a vendor plug-in passes all relevant tests then it can be considered a full substitute for the reference plug-in.
  2. If a vendor plug-in does not pass all relevant tests then the vendor is required to include the open source reference in the implementation.
  3. Alternate implementations may pass any tests that make sense and should add tests to validate new functionality.
  4. They are also required to have all the must-pass tests (see #10) to claim the OpenStack mark.

7.     OpenStack implementations are verified by open community tests

  1. OpenStack implementations by vendors must achieve 100% of must-have coverage.
  2. Implemented tests can be flagged as must-have required list.
  3. Certifiers will be required to disclose their testing gaps.
  4. This will put a lot of pressure on the Tempest project.
  5. Maintenance of the testing suite will become a core Foundation responsibility, which may require additional resources.
  6. Implementations and products are allowed to have variation based on publication of compatibility.
  7. Consumers must have a way to determine how the system is different from reference (posted, discovered, etc.).
  8. Testing must respond in an appropriate way in BOTH pass and fail (the wrong return rejects the entire suite).

8.     Tests can be remotely or self-administered

  1. Plug-in certification is driven by Tempest self-certification model.
  2. Self-certifiers are required to publish their results.
  3. Self-certifier are also required to publish enough information that a third party could build the reference implementation to pass the tests.
  4. Self-certifiers must include the operating systems that have been certified.
  5. It is preferred for self-certified implementations to refer back to an OpenStack reference architecture “flavor” instead of defining their own reference (a way to publish and agree on flavors is needed).
  6. The Foundation needs to define a mechanism of dispute resolution (i.e. a trust but verify model).
  7. All ecosystem partners need to make a “works against OpenStack” statement that is supportable.
  8. An API consumer can claim working against the OpenStack API if it works against any implementation passing all the “must have” tests.
  9. API consumers can state they are working against the OpenStack API with some “may have” items as requirements.
  10. API consumers are expected to write tests that validate their required behaviors (submitted as “may have” tests)

9.     A subset of tests are chosen by the Foundation as “must-pass”

  1. An OpenStack body will recommend which tests are elevated from “may-have” to “must-have.”
  2. The selection of “must-pass” tests should be based on quantifiable information when possible.
  3. Must-pass tests should be selected from the existing body of may-pass tests.  This encourages people to write tests for cases they want supported.
  4. We will have a process by which tests are elevated from “may” to “must” lists.
  5. Essentially, the hope is that the User Committee will nominate tests that elevate to the Board.

10.  OpenStack core means passing all “must-pass” tests

  1. The OpenStack Board owns the responsibility to define core (to approve ‘musts’).
  2. We are NOT defining which items are on the list in the Spider effort; rather, just making the position that this is how we will define core.
  3. May-have tests include items in the integrated release, but which are not core.
  4. Must-haves must comply with the core criteria defined from the IncUp committee results.
  5. Projects in Incubation or pre-Incubation are not to be included in the “may” list.

For a quick visual way to understand how these position statements interconnect Rob posted, in his blog series, an ‘OpenStack Core flowchart’. [2]

Help Us Finalize the 10

These position statements are an ongoing work in process. It is our goal to finalize them for the November Board meeting and Hong Kong Summit. Yet in order to meet that goal, we want your help and, as such, would love to hear from you. I encourage you all to strike up a conversation on IRC with Rob, myself or any board member, or simply post your feedback to the Foundation mailing list. We’re looking to you to help us reach the goal of the Foundation’s mission!

Alan Clark
OpenStack Board Chair

[1] http://robhirschfeld.com/2013/07/23/introducing-the-openstack-spider-graph-untangling-our-web-of-interdependencies/
[2] http://robhirschfeld.com/2013/08/07/visualizing-the-openstack-core/

Leave a Reply

Your email address will not be published. Required fields are marked *