-
Initiative
-
Resolution: Unresolved
-
Undefined
-
None
-
None
-
None
-
False
-
-
False
-
0% To Do, 50% In Progress, 50% Done
Feature title: AIPCC Infrastructure Pruning & Simplification
Feature Overview:
This feature systematically prunes obsolete build infrastructure and simplifies remaining configurations across the entire AIPCC ecosystem, including builder, containers, base images, and CI/CD pipelines.
The feature solves the problem of resource wastage, complexity, and maintenance overhead caused by supporting obsolete accelerator versions (e.g., CUDA 12.8, ROCm 6.4.3) and maintaining custom build processes for standard packages.
Value Statement/Executive Summary: This initiative results in a faster, more reliable, and easier-to-maintain AIPCC infrastructure. This leads to quicker build times, reduced operational costs across all base images, containers, and pipelines.
Product(s) associated:
RHAIIS: Yes
RHEL AI: Yes
RHOAI: Yes
Goals:
- Systematically deprecate and remove build artifacts for unsupported or superseded accelerator versions to reduce operational overhead. Subsequently, filter the remaining active AIPCC collections/pipelines to focus only on packages that utilize specific builder customizations (overrides or package plugins).
- Expected User Outcomes:
-
- Reduced build queue and pipeline execution times due to fewer active jobs.
-
- Lower infrastructure costs from reduced compute/storage utilization across the build environment.
-
- Simplified maintenance and debugging for all AIPCC components.
Who benefits from this Feature, and how?
**
| Beneficiary | How They Benefit |
| Internal Engineering/DevOps Teams | Significantly Reduced Maintenance: Less time spent debugging failures in obsolete build environments and clearer focus on actively supported configurations. |
| AI/ML Developers (Internal & External) | Faster Time to Release/Deployment: Quicker and more predictable build process for new images and updates. Clearer Support Matrix: All AIPCC artifacts reflect only officially supported versions. |
What is the difference between today’s current state and a world with this Feature?
| Current State (Today) | Future State (With Pruning & Simplification) |
| AIPCC infrastructure contains jobs/pipelines/containers for obsolete/unused accelerator versions (e.g., CUDA 12.8, ROCm 6.4.3). | Obsolete build artifacts are permanently removed across all systems, freeing up resources and simplifying base images. |
| Remaining active collections/pipelines build all packages, including those that don't require specialized logic. | Remaining active collections/pipelines are streamlined to only build packages with overrides or package plugins. |
| Increased build times, higher storage/compute costs, and increased complexity for maintenance. | Faster builds, lower costs, and a simplified, focused AIPCC infrastructure. |
Requirements:
- The AIPCC system must identify and list all builder jobs, collections, pipelines, and associated base image tags currently configured for the specified unused accelerator versions (e.g., CUDA 12.8 and ROCm 6.4.3).
- A robust and reversible mechanism must be implemented to safely remove/archive these identified build artifacts from the builder system, pipeline configuration, and container/image registry.
- The remaining active AIPCC collections/pipelines must be analyzed to identify packages that do not have associated builder overrides or package plugins.
- The build configuration must be modified such that these identified packages (without overrides/plugins) are removed from the active, custom collections/pipelines and rely only on standard, base package builds.
- All changes must be validated to ensure that actively supported accelerator versions (e.g., CUDA 12.9, ROCm 7.x) and their associated packages continue to build successfully and as intended.
Done - Acceptance Criteria:
- Obsolete Cleanup Verified: All build jobs, collections, pipelines, and associated base images/container tags explicitly targeting CUDA 12.8 and ROCm 6.4.3 have been successfully removed and are no longer accessible or runnable on the main branches.
- Collection Streamlining Complete: A review of the remaining active AIPCC collections/pipelines confirms that they only contain packages that have a defined builder override or utilize a package plugin.
- Non-Custom Package Removal Verified: A sample of packages that do not have an override or plugin are confirmed to have been successfully removed from the custom collections/pipelines.
Use Cases - i.e. User Experience & Workflow:
Include use case diagrams, main success scenarios, alternative flow scenarios.
Out of Scope:
- Removal of all packages from active collections: Only packages without overrides or package plugins are removed; all others remain.
- Modification of customer-facing components that are not directly generated by the AIPCC build system.
- Support for any future/unannounced accelerator versions (e.g., future ROCm 7.x releases).
Documentation Considerations :
Provide information that needs to be considered and planned so that documentation will meet customer needs. If the feature extends existing functionality, provide a link to its current documentation.
- is depended on by
-
AIPCC-7288 Versioned images names are not consistent
-
- New
-