BDEv: Big Data Evaluator


BDEv Features

BDEv supports the following features, which are executed in an automatic way.

  • BDEv performs:
    • Automatic configuration of the frameworks
    • Deployment of the frameworks using different cluster sizes
    • Execution of predefined and user-defined workloads (batch and interactive modes)
  • BDEv automatically records:
    • Output and runtime of the workloads
    • Performance and scalability results
    • Configuration and log directories of the frameworks
    • Resource utilization stats using dstat/dool and BDWatchdog
    • CPU power consumption stats using RAPL and turbostat
    • Microachitecture-level events using Oprofile
    • Energy consumption stats using HPE iLO technology
    • Automatically-generated graphs
  • BDEv allows:
    • To execute the workloads using different frameworks and cluster sizes
    • To unify the configuration of the different frameworks
    • To configure timeouts for the workloads
    • To setup multiple disks to store intermediate data for the frameworks
    • To configure different storage backends (HDFS, NFS)
    • To leverage high-performance resources like IP over InfiniBand (IPoIB)

Supported frameworks in BDEv 4.0.0

Framework Versions Deploy mode Network interfaces
Hadoop-YARN 2.10.2 / 3.0.3 / 3.1.4 / 3.2.4 / 3.3.6 / 3.4.3 / 3.5.0 YARN Ethernet / IPoIB
RDMA-Hadoop-3 0.9.1 YARN Ethernet / IPoIB / RDMA (IB & RoCE)
Spark 2.4.8 / 3.0.3 / 3.1.3 / 3.2.4 / 3.3.4 / 3.5.9 / 4.0.4 / 4.1.3 / 4.2.0 Standalone / YARN Ethernet / IPoIB
Flink 1.15.4 / 1.16.3 / 1.17.2 / 1.18.1 / 1.19.3 / 1.20.5 Standalone / YARN Ethernet / IPoIB

Supported benchmarks in BDEv 4.0.0

Benchmark Input generator Hadoop Spark Flink
Testdfsio -
Wordcount RGen (Hadoop RandomTextWriter)
Sort RGen (Hadoop RandomTextWriter)
Grep RGen (Hadoop RandomTextWriter)
Terasort RGen (Hadoop TeraGen)
TPCx-HS RGen (TPC HSGen)
Pagerank RGen
Connected Components RGen
Bayes RGen
Kmeans RGen (Mahout GenKMeansDataset)
Aggregation RGen
Join RGen
Scan RGen