Subject: Re: seemingly dismal performance of NetBSD-1.1A/sun3 file I/O....
To: Gordon W. Ross <>
From: Michael L. VanLoon -- <>
List: current-users
Date: 04/11/1996 09:40:36
>I have to wonder wonder what these I/O tests are really measuring.
>Is it really sustained transfer rate of the disk?

It depends on what test you're talking about, and how large.  Iozone,
in auto mode, tests sequential transfers in varying block sizes from
512 bytes to 8K, of a file varying from 1MB to 16MB.  For most people,
the large file sizes are going to be big enough that they'll not be
completely cacheable by the buffer cache.  It is a sequential
read/write test, through the filesystem.

Bonnie tests character-at-a-time transfers (saturates the CPU in the
tty code), and block transfers, giving figures for sequential output
and sequential input.  It also tests random seeks by spawning three
daemons that constantly send seek calls to the filesystem.  It tests
through the filesystem as well, and by default, uses a 100MB file that
should be not cacheable on most systems.  In addition, Bonnie measures
the CPU overhead of each of these tasks.

Both programs have decent documentation with the source that describes
what they're trying to do.

Another program that Justin T. Gibbs posted on this list about a month
ago tests the command overhead between the drive and the OS, by
measuring small continuous transfers of two different block sizes, and
comparing the overhead difference.  It effectively negates the
filesystem by using such small transfers.

  Michael L. VanLoon                       
        --<  Free your mind and your machine -- NetBSD free un*x  >--
    NetBSD working ports: 386+PC, Mac 68k, Amiga, Atari 68k, HP300, Sun3,
               Sun4/4c/4m, DEC MIPS, DEC Alpha, PC532, VAX...
    NetBSD ports in progress: PICA, others...