February 06, 2026: Weekly Status Update in Gluten

[GlutenPerfBot]

This weekly update is generated by LLMs. You're welcome to join our Github for in-depth discussions.

Overall Activity Summary

The Gluten project has been highly active over the past week with 49 pull requests and 21 issues, showing strong momentum in development. Key themes include Spark 4.x compatibility improvements, build system enhancements, memory management optimizations, and ANSI mode support expansion. The community is actively working on stabilizing the upcoming 1.6.0 release while addressing critical memory and performance issues.

Key Ongoing Projects

Spark 4.x Compatibility Initiative

A major effort led by @baibaichen to ensure full Spark 4.0/4.1 compatibility, with 400+ test suites being enabled. Recent fixes include:

• [GLUTEN-11550][CORE][UT] Fix GlutenSingleJoinSuite for Spark 4.0+ #11577 by @baibaichen enabling GlutenSingleJoinSuite for Spark 4.0+
• [GLUTEN-11550][UT] Enable GlutenXmlExpressionsSuite for spark4x and exclude 'from_xml- invalid data' #11580 by @baibaichen fixing XML expressions test suite
• [GLUTEN-11550][UT] Fixed GlutenSparkPlanSuite & GlutenSetCommandSuite #11579 by @Surbhi-Vijay fixing multiple test suites for Spark 4.0+

Build System Modernization

@liuneng1994 introduced a complete Gradle build system (#11576) that coexists with Maven, offering 2.5x faster cold builds and 118x faster incremental builds. This represents a significant developer experience improvement.

Memory Management Improvements

Several critical memory-related fixes:

• [GLUTEN-11509][VL] Make TreeMemoryConsumer thread-safe #11553 by @malinjawi making TreeMemoryConsumer thread-safe
• [CORE][VL] Fix protobuf memory leak in JNI_OnUnload #11532 by @clee704 fixing protobuf memory leak in JNI_OnUnload
• [VL] Race condition in VeloxMemoryManager::getOrCreateArrowMemoryPool with weak_ptr #11485 race condition fix in VeloxMemoryManager by @baibaichen

ANSI Mode Support Expansion

@PHILO-HE continues leading the comprehensive ANSI mode support initiative (#10134), with recent additions including string-to-boolean casting and ongoing work on type casting functions.

Priority Items

Critical Memory Issues

• [VL][1.5] Not enough spark off-heap execution memory on rss shuffle writer #11542: RSS shuffle writer memory threshold problems causing OOM errors
• [VL][1.5] Not enough spark off-heap execution memory on window #11540: Window operations hitting memory limits
• [VL][1.5] Not enough spark off-heap execution memory on RowToVeloxColumnar #11541: RowToVeloxColumnar memory allocation issues

Build and CI Infrastructure

• [VL] Java stackoverflow on packaging Spark-4.0 #11582: Java stackoverflow issues during Spark-4.0 packaging
• [VL] Caching java dependencies in testing docker #11501: Docker image optimization for faster CI builds
• [VL] CI: Support running Spark UT in centos9 docker image #11511: CentOS 9 support for Spark unit tests

Performance Optimizations

• [GLUTEN-7548][VL] Optimize BHJ in velox backend #8931: Broadcast hash join optimization by @JkSelf showing 1.29x performance improvement
• [GLUTEN-10215][VL] Delta write: Native statistics tracker to eliminate C2R overhead #11419: Delta write native statistics tracker eliminating C2R overhead by @zhztheplayer

Notable Discussions

Performance Benchmarking

#11554: Community discussion on Velox Bloom Filter inefficiency compared to Databricks Photon at 1TB scale, highlighting the need for better large-scale filtering capabilities.

Release Planning

#11568: Upcoming release manager scheduling for 1.6.0 (February 2026) through 1.10.0, with @zhztheplayer managing the upcoming 1.6.0 release.

Platform Support

#11535: macOS Apple Silicon support discussion, indicating growing interest in local development on modern hardware.

Emerging Trends

1. Spark 4.x Migration Acceleration: The project is rapidly moving toward full Spark 4.x compatibility with extensive test coverage being added.

2. Memory Management Focus: Significant engineering effort is being directed toward solving memory-related issues, particularly around shuffle operations and off-heap memory management.

3. Build System Evolution: The introduction of Gradle alongside Maven shows the project's commitment to developer experience improvements.

4. ANSI Compliance Priority: Growing emphasis on ANSI SQL compliance, especially with Spark 4.0 making ANSI mode the default.

5. Performance Optimization: Multiple PRs focused on reducing overhead and improving performance, particularly for Delta Lake operations and broadcast joins.

Good First Issues

#10134: ANSI Mode Support

Skills needed: Scala, SQL expressions, Spark internals
Why it's good: Well-documented issue with clear task breakdown. Perfect for understanding Spark's expression system and type casting. Each subtask is self-contained.

#11501: Docker Dependencies Caching

Skills needed: Docker, CI/CD, Maven
Why it's good: Infrastructure improvement with clear requirements. Good introduction to Gluten's CI system and build optimization.

#11511: CentOS 9 CI Support

Skills needed: GitHub Actions, Docker, Linux
Why it's good: Straightforward infrastructure task that helps understand the project's CI/CD pipeline and testing infrastructure.

#11383: Velox Bloom Filter Configuration

Skills needed: Java, Configuration management
Why it's good: Simple configuration addition task that introduces Velox backend integration patterns.

#11509: TreeMemoryConsumer Thread Safety

Skills needed: Java, Concurrent programming
Why it's good: Well-defined problem with existing error examples. Excellent for learning about Gluten's memory management architecture.