TiSpark

TiSpark is a thin layer built for running Apache Spark on top of TiDB/TiKV to answer complex OLAP queries. It takes advantages of both the Spark platform and the distributed TiKV cluster while seamlessly glues to TiDB, the distributed OLTP database, to provide a Hybrid Transactional/Analytical Processing (HTAP), and serves as a one-stop solution for online transactions and analysis.

Quick start

Read the Quick Start.

Getting TiSpark

The current stable version is 1.0 which is compatible with Spark 2.1.0+.

When using Spark 2.1.0+, please follow the document for Spark 2.1

When using Spark 2.3.0+, please follow the document for Spark 2.3

If you are using maven, add the following to your pom.xml:

<dependency>
  <groupId>com.pingcap.tispark</groupId>
  <artifactId>tispark-core</artifactId>
  <version>1.0</version>
</dependency>

If you're using SBT, add the following line to your build file:

libraryDependencies += "com.pingcap.tispark" % "tispark-core" % "1.0"

For other build tools, you can visit search.maven.org and search with GroupId (This search will also list all available modules of TiSpark including tikv-client).

To build from sources that is compatible with Spark 2.3.0+, please follow the next section.

How to build from sources

TiSpark now supports Spark 2.3.0+. The previous version for Spark 2.1.0+ will only contain bug fixes in future, you may still get Spark 2.1 support until release 1.1.

git clone https://github.com/pingcap/tispark.git

To build all TiSpark modules from sources, please run command under TiSpark root directory:

mvn clean install -Dmaven.test.skip=true

Remember to add -Dmaven.test.skip=true to skip all the tests if you don't need to run them.

How to migrate from Spark 2.1 to Spark 2.3

For users using Spark 2.1 who wish to migrate to latest TiSpark on Spark 2.3, please download or install Spark 2.3+ following instructions on Apache Spark Site and overwrite the old spark version in $SPARK_HOME.

TiSpark Architecture

• TiSpark integrates with Spark Catalyst Engine deeply. It provides precise control of the computing, which allows Spark read data from TiKV efficiently. It also supports index seek, which improves the performance of the point query execution significantly.

• It utilizes several strategies to push down the computing to reduce the size of dataset handling by Spark SQL, which accelerates the query execution. It also uses the TiDB built-in statistical information for the query plan optimization.

• From the data integration point of view, TiSpark + TiDB provides a solution runs both transaction and analysis directly on the same platform without building and maintaining any ETLs. It simplifies the system architecture and reduces the cost of maintenance.

• In addition, you can deploy and utilize tools from the Spark ecosystem for further data processing and manipulation on TiDB. For example, using TiSpark for data analysis and ETL; retrieving data from TiKV as a machine learning data source; generating reports from the scheduling system and so on.

TiSpark depends on the existence of TiKV clusters and PDs. It also needs to setup and use Spark clustering platform.

A thin layer of TiSpark. Most of the logic is inside tikv-client library. https://github.com/pingcap/tispark/tree/master/tikv-client

Quick Start

From Spark-shell:

./bin/spark-shell --jars /wherever-it-is/tispark-${version}-jar-with-dependencies.jar

For TiSpark version >= 2.0:

spark.sql("use tpch_test")

spark.sql("select count(*) from lineitem").show

For TiSpark version < 2.0:

import org.apache.spark.sql.TiContext
val ti = new TiContext (spark)
ti.tidbMapDatabase ("tpch_test")

spark.sql("select count(*) from lineitem").show

Current Version

spark.sql("select ti_version()").show

Configuration

Below configurations can be put together with spark-defaults.conf or passed in the same way as other Spark config properties.

Unsupported MySQL Type List

Statistics information

If you want to know how TiSpark could benefit from TiDB's statistic information, read more here.

How to test

We use docker-compose to provide tidb cluster service which allows you to run test across different platforms. It is recommended to install docker in order to test locally, or you can set up your own TiDB cluster locally as you wish.

If you prefer the docker way, you can use docker-compose up -d to launch tidb cluster service under tispark home directory. If you want to see tidb cluster's log you can launch via docker-compose up. You can use docker-compose down to shutdown entire tidb cluster service. All data is stored in data directory at the root of this project. Feel free to change it.

You can read more about test here.

Follow us

Twitter

@PingCAP

Mailing list

[email protected]

Google Group

License

TiSpark is under the Apache 2.0 license. See the LICENSE file for details.