Skip to content
mateiz edited this page Jun 12, 2012 · 78 revisions

Spark is a MapReduce-like cluster computing framework designed to support low-latency iterative jobs and interactive use from an interpreter. It is written in Scala, a high-level language for the JVM, and exposes a clean language-integrated syntax that makes it easy to write parallel jobs. Spark runs on top of the Apache Mesos cluster manager.

Downloading

Get Spark by checking out the master branch of the Git repository, using git clone git://github.com/mesos/spark.git.

Building

Spark requires Scala 2.9. In addition, to run Spark on a cluster, you will need to install Mesos, using the steps in Running Spark on Mesos. However, if you just want to run Spark on a single machine (possibly using multiple cores), you do not need Mesos.

To build and run Spark, you will need to have Scala's bin directory in your PATH, or you will need to set the SCALA_HOME environment variable to point to where you've installed Scala. Scala must be accessible through one of these methods on Mesos slave nodes as well as on the master.

Spark uses Simple Build Tool, which is bundled with it. To compile the code, go into the top-level Spark directory and run

sbt/sbt compile

Testing the Build

Spark comes with a number of sample programs in the examples directory. To run one of the samples, use ./run <class> <params> in the top-level Spark directory (the run script sets up the appropriate paths and launches that program). For example, ./run spark.examples.SparkPi will run a sample program that estimates Pi. Each of the examples prints usage help if no params are given.

Note that all of the sample programs take a <host> parameter that is the Mesos master to connect to. This can be a Mesos master URL, or local to run locally with one thread, or local[N] to run locally with N threads. You should start by using local for testing.

Finally, Spark can be used interactively from a modified version of the Scala interpreter that you can start through ./spark-shell. This is a great way to learn Spark.

A Note About Hadoop

Spark uses the Hadoop core library to talk to HDFS and other Hadoop-supported storage systems. Because the HDFS protocol has changed in different versions of Hadoop, you must build Spark against the same version that your cluster runs. You can change the version by setting the HADOOP_VERSION variable at the top of project/SparkBuild.scala, then rebuilding Spark (sbt/sbt clean compile).

Where to Go from Here

Other Resources

Community

To keep up with Spark development or get help, sign up for the spark-users mailing list.

If you're in the San Francisco Bay Area, there's a regular Spark meetup every few weeks. Come by to meet the developers and other users.

If you'd like to contribute code to Spark, read how to contribute.