diff --git a/assets/images/gluten_framework.png b/assets/images/gluten_framework.png new file mode 100644 index 0000000..6d14ff2 Binary files /dev/null and b/assets/images/gluten_framework.png differ diff --git a/assets/images/spark_operator_performance.png b/assets/images/spark_operator_performance.png new file mode 100644 index 0000000..7970abb Binary files /dev/null and b/assets/images/spark_operator_performance.png differ diff --git a/assets/images/velox_tpcds-like_sf3000_time_comparison.png b/assets/images/velox_tpcds-like_sf3000_time_comparison.png new file mode 100644 index 0000000..b92208a Binary files /dev/null and b/assets/images/velox_tpcds-like_sf3000_time_comparison.png differ diff --git a/assets/images/velox_tpcds-like_sf3000_top20_speedup.png b/assets/images/velox_tpcds-like_sf3000_top20_speedup.png new file mode 100644 index 0000000..553b417 Binary files /dev/null and b/assets/images/velox_tpcds-like_sf3000_top20_speedup.png differ diff --git a/assets/images/velox_tpch-like_sf3000_speedup.png b/assets/images/velox_tpch-like_sf3000_speedup.png new file mode 100644 index 0000000..7e444b0 Binary files /dev/null and b/assets/images/velox_tpch-like_sf3000_speedup.png differ diff --git a/assets/images/velox_tpch-like_sf3000_time_comparison.png b/assets/images/velox_tpch-like_sf3000_time_comparison.png new file mode 100644 index 0000000..29c37c9 Binary files /dev/null and b/assets/images/velox_tpch-like_sf3000_time_comparison.png differ diff --git a/contact-us.md b/contact-us.md index 63f9bcb..1ff2b4f 100644 --- a/contact-us.md +++ b/contact-us.md @@ -2,6 +2,7 @@ layout: page title: Contact Us nav_order: 9 +permalink: /contact/ --- # Contact Us diff --git a/contributing.md b/contributing.md index 24feb44..67c62ac 100644 --- a/contributing.md +++ b/contributing.md @@ -2,6 +2,7 @@ layout: page title: Contributing to Gluten nav_order: 7 +permalink: /contributing/ --- # How to become a committer diff --git a/index.md b/index.md index fddc91f..b8480e1 100644 --- a/index.md +++ b/index.md @@ -15,7 +15,7 @@ Apache Gluten(incubating) is a middle layer responsible for offloading JVM-based Apache Spark is a stable, mature project that has been developed for many years. It is one of the best frameworks to scale out for processing petabyte-scale datasets. However, the Spark community has had to address performance challenges that require various optimizations over time. As a key optimization in Spark 2.0, Whole Stage Code Generation is introduced to replace Volcano Model, which achieves 2x speedup. Henceforth, most optimizations are at query plan level. Single operator's performance almost stops growing.

- +

On the other side, SQL engines have been researched for many years. There are a few libraries like Clickhouse, Arrow and Velox, etc. By using features like native implementation, columnar data format and vectorized data processing, these libraries can outperform Spark's JVM based SQL engine. However, these libraries only support single node execution. @@ -48,7 +48,7 @@ You can click below links for more related information. The overview chart is like below. Substrait provides a well-defined cross-language specification for data compute operations (see more details [here](https://substrait.io/)). Spark physical plan is transformed to Substrait plan. Then Substrait plan is passed to native through JNI call. On native side, the native operator chain will be built out and offloaded to native engine. Gluten will return Columnar Batch to Spark and Spark Columnar API (since Spark-3.0) will be used at execution time. Gluten uses Apache Arrow data format as its basic data format, so the returned data to Spark JVM is ArrowColumnarBatch.

- +

Currently, Gluten only supports Clickhouse backend & Velox backend. Velox is a C++ database acceleration library which provides reusable, extensible and high-performance data processing components. More details can be found from https://github.com/facebookincubator/velox/. Gluten can also be extended to support more backends. @@ -59,3 +59,121 @@ There are several key components in Gluten: * **Fallback Mechanism**: supports falling back to Vanilla spark for unsupported operators. Gluten ColumnarToRow (C2R) and RowToColumnar (R2C) will convert Gluten columnar data and Spark's internal row data if needed. Both C2R and R2C are implemented in native code as well * **Metrics**: collected from Gluten native engine to help identify bugs, performance bottlenecks, etc. The metrics are displayed in Spark UI. * **Shim Layer**: supports multiple Spark versions. We plan to only support Spark's latest 2 or 3 releases. Currently, Spark-3.2, Spark-3.3 & Spark-3.4 (experimental) are supported. + +# 3 How to Use + +There are two methods for utilizing Gluten. The first is to use a pre-built JAR for a quick test of Gluten's acceleration capabilities. The second is to compile the JAR yourself, ensuring it runs on your target platform and delivers the best possible performance. + +# 3.1 Use a pre-Built Jar + +One way is to use released jar. Here is a simple example. Currently, only centos7/8 and ubuntu20.04/22.04 are well supported. +You can find a pre-built gluten release jar in [Apache Gluten Download](https://downloads.apache.org/incubator/gluten/). +Please be aware that the pre-built JAR is compiled using static linking via vcpkg, based on the Intel® Xeon® Gold 6252 processor. +It should be compatible with most operating systems, including CentOS and Ubuntu, although performance is not guaranteed. +For the optimal performance experience, we recommend using the 3.2 Custom Build to compile Gluten from source tailored to your specific platform. + +``` +spark-shell \ + --master yarn --deploy-mode client \ + --conf spark.plugins=org.apache.gluten.GlutenPlugin \ + --conf spark.memory.offHeap.enabled=true \ + --conf spark.memory.offHeap.size=20g \ + --conf spark.shuffle.manager=org.apache.spark.shuffle.sort.ColumnarShuffleManager \ + --jars gluten-velox-bundle-spark3.2_2.12-centos_7_x86_64-1.2.0.jar +``` + +# 3.2 Custom Build + +To utilize Gluten with Spark, you can compile Gluten from the source and then configure it to enable the Gluten plugin. Below is a straightforward example. For more comprehensive instructions, please refer to the detailed guidance provided in the corresponding backend section. + +``` +export gluten_jar = /PATH/TO/GLUTEN/backends-velox/target/ +spark-shell \ + --master yarn --deploy-mode client \ + --conf spark.plugins=org.apache.gluten.GlutenPlugin \ + --conf spark.memory.offHeap.enabled=true \ + --conf spark.memory.offHeap.size=20g \ + --conf spark.driver.extraClassPath=${gluten_jar} \ + --conf spark.executor.extraClassPath=${gluten_jar} \ + --conf spark.shuffle.manager=org.apache.spark.shuffle.sort.ColumnarShuffleManager + ... +``` + +## 3.2.1 Build and install Gluten with Velox backend + +If you want to use Gluten **Velox** backend, see [Build with Velox](https://gluten.apache.org/docs/getting-started/velox-backend/) to build and install the necessary libraries. +You can also find more information in [Gluten with Velox backend](https://gluten.apache.org/docs/velox-backend/) and [velox](https://github.com/facebookincubator/velox). + +## 3.2.2 Build and install Gluten with ClickHouse backend + +If you want to use Gluten **ClickHouse** backend, see [Build with ClickHouse Backend](http://gluten.apache.org/docs/getting-started/clickhouse-backend/). +ClickHouse backend is developed by [Kyligence](https://kyligence.io/), please visit [Kyligence's ClickHouse](https://github.com/Kyligence/ClickHouse) for more infomation. + +## 3.2.3 Build options + +See [Build Parameters](https://gluten.apache.org/docs/getting-started/build-guide/). + + +# 4 Join the Community + +## Contact Us + +Gluten was initiated by Intel and Kyligence in 2022. Several companies are also actively participating in the development, such as BIGO, Meituan, Alibaba Cloud, NetEase, Baidu, Microsoft, etc. If you are interested in Gluten project, please contact and subscribe below mailing lists for further discussion. + +Please see [contact us](https://gluten.apache.org/contact/) for more information. + +## Source Code + +Please see [gluten source code](https://github.com/apache/incubator-gluten) for more information. + +## How to Contribute to Gluten + +Please see [contributing guide](https://gluten.apache.org/contributing/) about how to make contributions. + +## Wechat Group + +There is a Wechat group (in Chinese) which may be more friendly for PRC developers/users. Due to the limitation of wechat group, please mail to dev@gluten.apache.org. + +## Slack channel + +For Velox backend, we recommend to use [velox community](https://github.com/facebookincubator/velox?tab=readme-ov-file#community). + + +# 5 Performance + +We use TPCH-like and TPCDS-like as decison support benchmarks to evaluate Gluten's performance. +TPCH-like is a query set for modified from [TPC-H benchmark](http://tpc.org/tpch/default5.asp) and TPCDS-like is a query set for modified from [TPC-DS benchmark](https://tpc.org/tpcds/default5.asp). We use Parquet file format for Velox testing & MergeTree file format for Clickhouse testing, compared to Parquet file format as baseline. See [Decision Support Benchmark](https://github.com/apache/incubator-gluten/tree/main/tools/workload). + +## Gluten + Velox backend Performance + +The below test environment: single node with 3TB data on Intel® Xeon® Platinum 8592+; Spark-3.3.1 for both baseline and Gluten. + +The TPCH-like result (tested in Mar. 2024) shows an overall speedup of 3.34x and up to 23.45x speedup in a single query with Gluten + Velox backend used. + +![Performance](/assets/images/velox_tpch-like_sf3000_time_comparison.png) + +![Performance](/assets/images/velox_tpch-like_sf3000_speedup.png) + +The TPCDS-like result (tested in Mar. 2024) shows an overall speedup of 3.02x and up to 13.75x speedup in a single query with Gluten + Velox backend used. + +![Performance](/assets/images/velox_tpcds-like_sf3000_time_comparison.png) + +![Performance](/assets/images/velox_tpcds-like_sf3000_top20_speedup.png) + +Notices & Disclaimers + +Performance varies by use, configuration and other factors. Learn more on the Performance Index site. Performance results are based on testing as of dates shown in configurations and may not reflect all publicly available updates. See backup for configuration details. No product or component can be absolutely secure. Your costs and results may vary. Intel technologies may require enabled hardware, software or service activation. © Intel Corporation. Intel, the Intel logo, and other Intel marks are trademarks of Intel Corporation or its subsidiaries. Other names and brands may be claimed as the property of others. + +## Gluten + ClickHouse backend Performance + +The below testing environment: a 8-nodes AWS cluster with 1TB data; Spark-3.1.1 for both baseline and Gluten. The Decision Support Benchmark1 result shows an average speedup of 2.12x and up to 3.48x speedup with Gluten Clickhouse backend. + +![Performance](/assets/images/clickhouse_decision_support_bench1_22queries_performance.png) + + +# Thanks to our contributors + + + + + diff --git a/release.md b/release.md index af8f01b..42c18ba 100644 --- a/release.md +++ b/release.md @@ -1,7 +1,8 @@ --- layout: page -title: Gluten Release +title: Gluten Releases nav_order: 5 +permalink: /releases/ --- [Gluten](https://github.com/apache/incubator-gluten) is a plugin for Apache Spark to double SparkSQL's performance.