Yicong-Huang opened a new issue, #4681:
URL: https://github.com/apache/texera/issues/4681

   ### Task Summary
   
   Pick up and re-land the work from #3674, which was the first of two PRs 
implementing Basic Ramen (operator-level parallelism via reused runtime 
statistics) on top of the umbrella issue #3605.
   
   #3674 has been closed without merging — it was authored 2025-08-19, has 14 
changed files (+622 / −203), is in a CONFLICTING state against current `main`, 
and has been idle. The work is still desirable; this issue captures the intent 
so it can be revived in a fresh branch.
   
   ### What the closed PR covered
   
   **UDF UI (frontend):**
   - Add a number-of-workers input field on the UDF operator panel.
   - Let users mark a UDF as parallelizable / non-parallelizable.
   - Surface parallelizability for non-UDF operators in read-only form (per the 
design in #3605).
   
   **Backend:**
   - Refactor `ResourceAllocator` to support configurable parallelism.
   - Add `GreedyResourceAllocator` — DFS-based baseline allocator from the 
Ramen paper that picks per-operator parallelism using past runtime stats.
   - Wire `GreedyResourceAllocator` into the UDF operator execution path.
   
   **Configuration:**
   - Workflow-level flag to enable/disable the greedy allocator (Basic Ramen 
mode).
   
   ### Out of scope
   
   PR 2 from the original plan — extending parallelism support to non-UDF 
parallelizable operators — should remain a separate follow-up issue once this 
lands.
   
   ### Pointers
   
   - Umbrella: #3605
   - Closed PR with diff and review history: #3674
   - Original branch (for reference / cherry-picking): `yunyad-basic-ramen-pr-1`
   
   ### Suggested approach
   
   1. Branch off current `main`.
   2. Cherry-pick / port the changes from `yunyad-basic-ramen-pr-1`, resolving 
conflicts (most likely in resource allocator and UDF executor — both have moved 
since August).
   3. Address any unresolved review comments from #3674 before re-opening.


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to