# apache/gluten

Gluten is a middle layer responsible for offloading JVM-based SQL engines' execution to native engines.

Repository: https://github.com/apache/gluten
Canonical: https://ross.abutalabs.com/products/gluten
Homepage: https://gluten.apache.org/
Language: Scala
License: Apache-2.0
License Family: permissive
Topics: clickhouse, simd, spark-sql, vectorization, velox, arrow
Last push: 2026-08-26T16:42:02+00:00

## Health v2 (maintenance only)
Score: 91/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 99, release rhythm 75, longevity 100
- inputs: {"age_days": 1731, "days_push": 7, "days_rel": 7, "gap_med": 131.5, "n_releases_24m": 7}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1592, forks 646 (observed 2026-08-28T04:05:08.736269+00:00)

## What it is
Apache Gluten is a middle-layer plugin that offloads JVM-based SQL engine execution (primarily Spark SQL) to high-performance native engines like Velox and ClickHouse. It translates Spark physical plans into Substrait plans executed natively via JNI, delivering up to 3x+ speedups with zero changes to user queries or DataFrame code.

## Use cases
- accelerate spark sql queries without changing application code
- speed up etl pipelines running on spark
- offload jvm sql execution to native vectorized engines
- run tpc-h and tpc-ds benchmarks faster on spark
- switch between velox and clickhouse backends for spark workloads
- add columnar shuffle and columnar processing to spark
- reduce compute cost of petabyte-scale spark jobs

## When to choose
- you run Spark SQL workloads and want major performance gains without rewriting queries
- your bottleneck is per-operator execution performance rather than query planning
- you want to leverage native engines like Velox or ClickHouse while keeping the Spark ecosystem
- you need automatic fallback to vanilla Spark for unsupported operators

## When to avoid
- you don't use a JVM-based SQL engine like Spark
- your queries rely heavily on operators unsupported by the native backends, causing frequent fallback
- you cannot enable off-heap memory or add Spark plugins in your environment
- you need a lightweight solution and can't tolerate the added operational complexity of native bundle JARs

## Facets
- artifact type: plugin
- maturity: active
- function: database, streaming, developer-tools
- domain: big-data, databases, analytics, performance
- platform: jvm, cloud
- tags: spark, spark-sql, velox, clickhouse, substrait, vectorized-execution, columnar, jni, native-execution, query-acceleration, simd, apache-arrow, scala, data-engineering, linux, docker

## Member repositories
- apache/gluten (main) score 91

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:08.736269+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T03:54:40.142081+00:00, confidence not recorded.
  - readme: https://github.com/apache/gluten (fetched 2026-08-28T04:05:08.736269+00:00, sha 8ff508c8ec3d)
  - homepage: https://gluten.apache.org/ (fetched 2026-08-29T11:25:08.245030+00:00, sha 411c21b51d03)
- Data as of 2026-08-30T08:39:29.467469+00:00.