From T2B Wiki
Jump to navigation Jump to search

Basic System Installation

The basic requirements for this benchmark is to have RHEL ES 4.3 for the x86_64 architecture installed and running on the hardware. The RHEL4 clones CentOS 4.3 and Scientific Linux 4.3 are also valid for the benchmark. We even recommand one of those clones because they provide yum (updater and package installer which is used below) by default. Installation instructions can be found on the sites or are provide with the copy.

All benchmarks will be run in 64-bit mode. In order to complete the benchmarks succesfully, a number of compatibility rpms have to be installed on the system.

Starting from a standard Desktop installation, install the following rpms (.i386 and .x86_64 indicate the rpms for both architecture versions) and their corresponding dependencies. (This is easiest done using yum.):

 yum install libf2c.i386 libf2c.x86_64 openmotif.i386 openmotif.x86_64 xorg-x11-Mesa-libGLU.i386 xorg-x11-Mesa-libGLU.x86_64
 yum install compat-libstdc++-33-3.2.3-47.3

Partition and Filesystem setup

In order to run the benchmark succesfully, there must be approximately 4GB free diskspace on one partition. We advise to use ext3 as filesystem. No further requirements are needed.

CMS benchmark

Make sure the machine is connected to the internet. Run as root:

 wget -q -O;chmod +x;./ 4

The benchmark consists of a script that will

  1. download the necessary software in one single tarball (approximately 650MB. It might take a while to download.)
  2. extract the tarball and install the software
  3. run all benchmarks
  4. printout the numbers
  5. create a tarball with detailed results.

Running the application benchmark itself takes at most 6 hours and requires no manual intervention. At the end, a number will be printed on screen indicating the relative performance of the system. Also a file with a detailed description of the system used and detailed results of all the tests will be provided.

When the script ran successfully, it will print out something like

Running test from directory /benchmark
  Now timing 1 single Root stress test:
  Now timing 1 single Root stressgeom test:
  Now timing 1 single Root stressLinear test:
  Now timing 1 single Root bench test:
  Now timing 4 simultaneous root stress runs:
 Exit code 0
  Now timing 1 single Pythia stress test:
  Now timing 4 simultaneous pythia stress runs:
 Exit code 0
  ATTENTION !!! All environment variables 
  matching pattern  *_PATH will be unset !!!
  Now timing 1 single oscar run:
... , 0
  Gold1 1288.3 , 0
  Gold2 1291.4 , 0
  Gold3 1285.2 , 0
  Gold4 1292.8
     Res64 0: 1.0901462631045 , 1
  Gold1 135.488503800453 , 1
  Gold2 135.50135501355 , 1
  Gold3 135.699940292026 , 1
  Gold4 135.587704906919
     Res64 1: 1.10156314295309
  Relative strength of this box: Res= 4.78483211009748
  Relative 32 to 64= 1.03331347491351
  32 to 64= 1.2876476914783

The number indicating the relative performance of the system is found at the line biginning by: "Relative strength of this box: Res=". The file with the detailed results is the zipped file created in the directory where the script was executed.