NVIDIA-NeMo / NVIDIA-NeMo/DataDesigner

Document async scheduling architecture, capacity, and operations

Open
#660 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

plan task
Dominant language
Python
Stars
2.2k
Forks
211
Avg merge
2d 6h
Merged PRs (30d)
40

Description

Priority Level

Medium

Epic: #645
Depends on: #652, #654, #657, #648, #649
Related: #641, #644, #635, #647, #650, #651
Target branch: epic/645-async-scheduling while the epic is active.

Source of Truth

Specification details live in PR #658 under plans/645/, especially README.md, architecture.md, contracts.md, capacity-model.md, observability.md, benchmark-plan.md, migration-and-cleanup.md, and issue-map.md. This issue tracks the final docs slice and gates only.

Documentation Scope

Document the complete async scheduling architecture after the metadata, task-admission, request-admission, capacity, observability, and benchmark issues land.

This issue owns:

  • maintainer/developer docs for the two-stage task/request scheduling model;
  • operator docs for capacity layers, runtime snapshots, and telemetry interpretation;
  • links from plugin-facing SchedulingMetadata docs (#652) without exposing scheduler internals as plugin API;
  • migration notes for removed names and their replacements;
  • docs references to benchmark artifacts and how to read them.

This issue does not own the plugin-facing SchedulingMetadata guide itself; #652 owns that narrower documentation.

Quality Gates

  • Docs match the UML-level architecture and terminology in plans/645.
  • Task admission, request admission, capacity, observability, and benchmarks are explained as separate but correlated layers.
  • Docs do not present removed Throttle*, legacy hint resolver, or old LLM-wait names as current durable APIs.
  • Public/plugin docs do not expose scheduler queue/admission internals as extension points.
  • Cross-links cover the relevant epic issues and plan sections.

Validation

Run the repository docs build/preview workflow for the touched surface, make check if required by the docs workflow, link checks if available, and stale-term searches over docs/current package docs for the removed names listed in plans/645/migration-and-cleanup.md.

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start with PR #658 and the specified files under plans/645/, especially README.md, architecture.md, contracts.md, capacity-model.md, observability.md, benchmark-plan.md, migration-and-cleanup.md, and issue-map.md. After the dependent epic issues land, update the maintainer, operator, and migration documentation, then run the repository docs build/preview workflow, make check if required, link checks, and stale-term searches; done means the quality gates are satisfied.

Written by the indexing model from the issue text.

Assessment

Domain
documentation
Issue type
Documentation
Difficulty
4/5
Estimated time
3-5 days
Activity status
Quiet
Clarity
Mostly clear
Newbie friendliness
48/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.