How to Use the Disk Performance Benchmark Tools

The following are some tips on how to use the benchmark tools and what you need take care of before using the tools:

• Cache Effect
  Cache can play a very bad role when doing the disk performance tests, especially to the write performance. You may want to get rid of it when doing the performance tests. RAM is the first to blame. My own observation shows swap only plays a secondary role in the cache effects.
  
  To remove the cache effects in Liunx, you may following the follow steps:
  
  • Reboot the machine with 16 or 32MB:
    1. Append this line to the /etc/lilo.conf:
    append="mem=32M"
    2. Run the linux loader:
    /sbin/lilo -v
    3. Re-boot the machine.
  • Reduce or turn the swap off:
    swap is critical when running large applications and/or supporting multiple processes/threads (except you have really large physical RAM). It seems to me it does NOT play a role when writing disk, unlike the physical RAM. Maybe I am wrong. Please e-mail me if you have a good point on this.
  There are several other ways to do this. You may look at the linux manual or ask your system administrator to help you.
  
  
• Always use a set of tools:
  Different tools may give you slightly different results. They may also measure different aspects of the disk performance. You may want to try several tools to get a better understanding of your overall disk performance. One good starting point is the scripts under tiobench benchmark tool. I will write some scripts when I have time.