DFS-Perf FAQ

This page lists the frequent asked general questions about the DFS-Perf project. If you have some problems during usage, please put forward your questions in the user mailing list.

• How does DFS-Perf relate to HDFS/Alluxio …?
  DFS-Perf is a benchmark framework on top of Distributed File Systems(DFS) such as HDFS, Alluxio, etc. It allows users to run workloads to test the performance of DFS.

• How can I run DFS-Perf on a cluster?
  To run on a cluster, you need a DFS as test target. Then, you can refer Running DFS-Perf on a Cluster to get DFS-Perf up and running on your own cluster.

• Can I run DFS-Perf on different file systems?
  Of courese yes. DFS-Perf is designed to support different widely used DFS. The very simple way to switch to another DFS is to modify the path schema in the configuration file, e.g. hdfs://master:9000 to alluxio://master:19998.

• What’s the difference between DFS-Perf and YCSB, HiBench, …?
  Now there exists benchmark frameworks for big data computing, such as YCSB, HiBench, etc. However, they are not targeted on DFS. YCSB is a benchmark framework for evaluating the performance of the new generation “key-value” and “cloud” serving stores. And HiBench is a Hadoop Benchmark Suite which contains a series of typical Hadoop MapReduce workloads. Indeed, there is no testing framework that can execute different workloads on different DFS. Our goal is to design such a general benchmarking framework for different DFS with the guidance of those existing benchmarks. That is DFS-Perf.

• Where can I get more help?
  Please post on the DFS-Perf user mailing list. We’ll be glad to help there!