OpenJDK* Environment Setup
This page describes the details of software environment for the platform used to run Java based workloads. Information about hardware platform setup can be found here. Software configuration: Red Hat Enterprise Linux Server release 6.8 (Santiago) on 64 bits with kernel version 2.6.32.
The source files for JDK9 are downloaded from here. We use Mercurial with the trees extension to clone the repository
hg tclone http://hg.openjdk.java.net/jdk9/jdk9
We build OpenJDK* using gcc-4.4.7, with default parameters. JDK9 requires a bootstrap JDK8. We use jdk-8u72-linux-x64.
./configure --with-boot-jdk=<PATH TO BOOTSTRAP JDK8>
We use the linux-x86_64-normal-server-release configuration when building the binaries.
make images CONF='linux-x86_64-normal-server-release'
- The following benchmarks are run for every JDK9 revision:
- * Java Business Benchmark 1¹
- * Java Business Benchmark 2²
- * Java Virtual Machine Benchmark³
The benchmarks are run automatically once for every revision. The system checks for new revisions on a nightly basis. We use jdk-8u72-linux-x64 as a baseline for the results.
The workflow of each benchmark is very specific. The generic process is the following:
- 1. Compile the benchmark
- 2. Perform X runs
- 3. Preprocess the results to facilitate next steps
Some benchmarks may be precompiled.
Other benchmarks (e.g. Java Business Benchmark 1¹, Java Business Benchmark 2²) require running a server and multiple clients started, making the second step more complicated.
¹ Server side Java measuring throughput and response time with graduated load levels These measurements are using a Server Side Java* middleware internal workload which is very similar to SPECjbb2015* in stressing the different components as well as following methodology for metrics calculation by measuring full system capacity "max ops/sec" and a response time SLA bound throughput "SLA passing ops/sec"
² Server side Java measuring throughput with scaling: These measurements are using a Server Side Java* internal workload which is very similar to SPECjbb2005* in stressing the different components as well as increasing the load for scaling by adding threads in increment of one and determining full system capacity throughput metric "ops/sec".
³Server and client side Java with mixed suites: These measurements are using a Server and client Side Java* internal workload which is very similar to SPECjvm2008* in stressing the different components as well as following methodology of mixing several suites of small workloads and calculating an aggregated throughput metric for full system capacity as "ops/sec".
*Other names and brands may be claimed as property of others. SPEC* and the benchmark names SPECjbb*, SPECjvm* are registered trademarks of the Standard Performance Evaluation Corporation. OpenJDK (the "Name") is a trademark of Oracle America, Inc. ("Oracle")(the "Trademark Owner").
Intel technologies' features and benefits depend on system configuration and may require enabled hardware, software or service activation. Performance varies depending on system configuration.