commit | 5fa0b14d20fab007d9e2d954eb4a34155a6f199f | [log] [tgz] |
---|---|---|
author | Kenneth Knowles <klk@google.com> | Thu Sep 07 11:38:22 2017 -0700 |
committer | Kenneth Knowles <klk@google.com> | Thu Sep 07 11:38:22 2017 -0700 |
tree | 1cd93287596ca07146da0ca50f89e409a62735e7 | |
parent | 2fa4fdecd8ef06534e369e527b15ae8193823c8b [diff] | |
parent | 32aeb7ac3d49ade0dc3ad79e711e7b624091d485 [diff] |
This closes #3705: [BEAM-165] Initial implementation of the MapReduce runner mr-runner: Removes WordCountTest, fixes checkstyle, findbugs, and addressed comments. mr-runner-hack: disable unrelated modules to shorten build time during development. mr-runner: support SourceMetrics, this fixes MetricsTest.testBoundedSourceMetrics(). mr-runner: introduces duplicateFactor in FlattenOperation, this fixes testFlattenInputMultipleCopies(). mr-runner: translate empty flatten into EmptySource, this fixes few empty FalttenTests. mr-runner: ensure Operation only start/finish once for diamond shaped DAG, this fixes ParDoLifecycleTest. mr-runner: Graph.getSteps() to return with topological order, this fixes few CombineTests. mr-runner: fail early in the runner when MapReduce job fails. mr-runner: use InMemoryStateInternals in ParDoOperation, this fixed ParDoTest that uses state. mr-runner: use the correct step name in ParDoTranslator, this fixes MetricsTest.testAttemptedCounterMetrics(). mr-runner: remove the hard-coded GlobalWindow coder, and fixes WindowingTest. mr-runner: handle no files case in FileSideInputReader for empty views. mr-runner: fix NPE in PipelineTest.testIdentityTransform(). mr-runner: filter out unsupported features in ValidatesRunner tests. mr-runner: setMetricsSupported to run ValidatesRunner tests with TestPipeline. mr-runner: fix the bug that steps are attached multiple times in diamond shaped DAG. [BEAM-2783] support metrics in MapReduceRunner. mr-runner: setup file paths for read and write sides of materialization. mr-runner: support side inputs by reading in all views contents. mr-runner: support multiple SourceOperations by composing and partitioning. mr-runner: support PCollections materialization with multiple MR jobs. mr-runner: hack to get around that ViewAsXXX.expand() return wrong output PValue. mr-runner: support graph visualization with dotfiles. mr-runner: refactors and creates Graph data structures to handle general Beam pipelines. mr-runner: add JarClassInstanceFactory to run ValidatesRunner tests. mr-runner: support reduce side ParDos and WordCount. core-java: InMemoryTimerInternals expose getTimers() for timer firings in mr-runner. mr-runner: add BeamReducer and support GroupByKey. mr-runner: add ParDoOperation and support ParDos chaining. mr-runner: add JobPrototype and translate it to a MR job. mr-runner: support BoundedSource with BeamInputFormat. MapReduceRunner: add unit tests for GraphConverter and GraphPlanner. MapReduceRunner: add Graph and its visitors. Initial commit for MapReduceRunner.
Apache Beam is a unified model for defining both batch and streaming data-parallel processing pipelines, as well as a set of language-specific SDKs for constructing pipelines and Runners for executing them on distributed processing backends, including Apache Apex, Apache Flink, Apache Spark, and Google Cloud Dataflow.
Beam provides a general approach to expressing embarrassingly parallel data processing pipelines and supports three categories of users, each of which have relatively disparate backgrounds and needs.
The model behind Beam evolved from a number of internal Google data processing projects, including MapReduce, FlumeJava, and Millwheel. This model was originally known as the “Dataflow Model”.
To learn more about the Beam Model (though still under the original name of Dataflow), see the World Beyond Batch: Streaming 101 and Streaming 102 posts on O’Reilly’s Radar site, and the VLDB 2015 paper.
The key concepts in the Beam programming model are:
PCollection
: represents a collection of data, which could be bounded or unbounded in size.PTransform
: represents a computation that transforms input PCollections into output PCollections.Pipeline
: manages a directed acyclic graph of PTransforms and PCollections that is ready for execution.PipelineRunner
: specifies where and how the pipeline should execute.Beam supports multiple language specific SDKs for writing pipelines against the Beam Model.
Currently, this repository contains SDKs for both Java and Python.
Have ideas for new SDKs or DSLs? See the JIRA.
Beam supports executing programs on multiple distributed processing backends through PipelineRunners. Currently, the following PipelineRunners are available:
DirectRunner
runs the pipeline on your local machine.ApexRunner
runs the pipeline on an Apache Hadoop YARN cluster (or in embedded mode).DataflowRunner
submits the pipeline to the Google Cloud Dataflow.FlinkRunner
runs the pipeline on an Apache Flink cluster. The code has been donated from dataArtisans/flink-dataflow and is now part of Beam.SparkRunner
runs the pipeline on an Apache Spark cluster. The code has been donated from cloudera/spark-dataflow and is now part of Beam.Have ideas for new Runners? See the JIRA.
Please refer to the Quickstart[Java, Python] available on our website.
If you'd like to build and install the whole project from the source distribution, you may need some additional tools installed in your system. In a Debian-based distribution:
sudo apt-get install \ openjdk-8-jdk \ maven \ python-setuptools \ python-pip
Then please use the standard mvn clean install
command.
See the Spark Runner README.
To get involved in Apache Beam:
We also have a contributor's guide.