• United States

How we did Sun Fire test

Dec 19, 20052 mins
Data Center

How the Clear Choice Test of Sun’s T2000 was conducted.

We tested a preproduction unit that included a preinstalled version of Solaris 10. The T2000 had an eight-core, 1.2GHz Sun CPU with 32GB of DDR memory, and internal CD/DVD drive, a 73GB serial advanced technology attachment internal hard drive with dual power supplies, three PCI-E bus slots and two PCI-X bus slots. We connected the T2000 via an LSI Logic Fibre Channel host bus adapter to our internal network. All tests, however, used the internal SATA drive.

We ran tests with LMBench3, an open source benchmarking tool. As is required by the LMBench license, full disclosure has been made to LMBench’s authors, and full results are available here.

  • The complete lmbench3 test file (compressed file)
  • Performance with Linux OS, 1 instance (text file)
  • Performance with Linux OS, 8 instances (text file)
  • Performance with Solaris OS, 1 instance (text file)
  • Performance with Solaris OS, 8 instances (text file)

We ran a single copy test instance and an eight-copy test instance.

We also compared the Sun Fire T2000 to an HP 585 (configured with four dual-core AMD64 Opteron CPUs running 2.4GHz each, 16GB of dynamic RAM, internal HP controller with two SCSI (UltraSCSI320-like drives – although only one drive was used for testing) under Novell/SuSE Linux 10 SMP CPU 2.6.7 kernel, in single copy mode, and in eight-copy mode via LMBench3.

Both operating systems were factory installed, unoptimized configurations, running in a bare configuration (no httpd, bind or other server processes), to enhance repeatability. LMBench3 was compiled with the gcc compiler, using default (unoptimized for threads, cache optimizations or other enhancements) compile-time switches. We commented out code (removed source code) in LMBench3 relating to network tests.