In article <200305061823.OAA29775@_stups.com>,
Stephan Uphoff <ups@stups.com> wrote:
>Paul Ripke wrote:
>> OK, what concerns me is those "random seeks/sec" rates. There's a huge
>> difference there, and that could hurt other stats. It shouldn't be CPU
>> bound (obviously), the only thing that comes to mind is in disk
>> I/O scheduling - disksort. I wonder what the difference is.
>> 
>> [alternately, maybe I'm spending too much time looking at DBMS at work]
>
>
>I would suspect the read ahead to be the culprit, since it reads in an
>extra 128 KByte, when a random read (rounded up by file the file system
>block size) ends at a 64KByte boundary. (see genfs_getpages())
>
>Block-sized random reads on a filesystem with 8K blocks would read
>an average of three blocks for every block requested. 

maybe we need an fadvise(2) syscall :-)

christos