cluster-benchmark

How to use

1. Download it.

git clone https://github.com/hdinsight/cluster-benchmark && cd cluster-benchmark

2. Compile it.

mvn package

3. Run it.

hadoop jar target/cluster-benchmark.jar [BenchmakrName]

Currently following benchmarks are available:

• Sleep: Streeses YARN components (RM,NM,Historyserver) etc by creating a job with given number of tasks which sleeps for given amount of time.

• Compute: Measures the compute capacity available on the cluster with a job which repeatedly generaetes random numbers.

• Read: Measures the read capactiy of the file system.

• DFSWrite: Measures the write capacity of the distributed file system (WASB,HDFS,ADLS, etc.)

• Write: Measures the write capacity of local disk writes.

• Shuffle: Measures the shuffle performance of the cluster.

• ReadCompute: Generates a job that does given amount of read and compute.

• ReadComputeWrite: Generates a job that does given amount of read, compute and write.

• Mix: Generates a job which generates a mix of compute,read and write.

• Mix2: Genereates a job which generates a mix of compute,read and write.