On Sun, Feb 25 2007, Roy Lee wrote:
> Hi,
>
> I was wondering if fio can add support for doing workloads that has
> finner granularity, for example, to simulate a single process/thread
> that has multiple different I/O patterns.
>
> It would be better if each I/O pattern can have it's own thinktime.
>
> [job1]
> rand_read_weight=1, 2-4ms
> seq_read_weight=5, 3ms
> seq_write_weight=5, 10ms
I would not be adverse to taking such a patch, if you feel like writing
it up. To avoid complicating/changing the option parser, you'd probably
want to do something ala:
rand_read_weight=1:2-4
to indicate a split option, similar to how bsrange is handled.
-- Jens AxboeReceived on Mon Feb 26 2007 - 13:00:08 CET
This archive was generated by hypermail 2.2.0 : Wed Feb 28 2007 - 11:47:09 CET