HP-UX Direct vs Buffered I/O - HP UX

This is a discussion on HP-UX Direct vs Buffered I/O - HP UX ; Oracle 9.2.0.4 EE on HP-UX 11i I have a question relating to HP-UX file I/O: As far as I am aware HP-UX is capable of Direct I/O (i.e. bypassing the Unix buffer cache) when using VXFS and the following mount ...

+ Reply to Thread
Results 1 to 4 of 4

Thread: HP-UX Direct vs Buffered I/O

  1. HP-UX Direct vs Buffered I/O

    Oracle 9.2.0.4 EE on HP-UX 11i

    I have a question relating to HP-UX file I/O:

    As far as I am aware HP-UX is capable of Direct I/O (i.e. bypassing the
    Unix buffer cache) when using VXFS and the following mount parameter:

    convosync=direct

    This parameter is not used in the mount options in /etc/fstab so I am
    assuming that I am double buffering on I/O operations.

    However is there any way of 'proving' that a particular operation is
    going via the OS cache or not. If I tusc/truss the process I will just
    see pread() calls which may or may not turn into an OS cache hit.

    Matt


  2. Re: HP-UX Direct vs Buffered I/O

    Hi Matt:

    Why not just use raw logical volume instead of vxfs? That way you will
    bypass fs cache.

    Vince
    wrote in message
    news:1121440226.216536.307710@g43g2000cwa.googlegr oups.com...
    > Oracle 9.2.0.4 EE on HP-UX 11i
    >
    > I have a question relating to HP-UX file I/O:
    >
    > As far as I am aware HP-UX is capable of Direct I/O (i.e. bypassing the
    > Unix buffer cache) when using VXFS and the following mount parameter:
    >
    > convosync=direct
    >
    > This parameter is not used in the mount options in /etc/fstab so I am
    > assuming that I am double buffering on I/O operations.
    >
    > However is there any way of 'proving' that a particular operation is
    > going via the OS cache or not. If I tusc/truss the process I will just
    > see pread() calls which may or may not turn into an OS cache hit.
    >
    > Matt
    >




  3. Re: HP-UX Direct vs Buffered I/O

    I'm not interested in bypassing the fs cache at this point - I just
    need confirmation that I am using it..

    Cheers

    Matt


  4. Re: HP-UX Direct vs Buffered I/O

    mccmx@hotmail.com wrote:
    > I'm not interested in bypassing the fs cache at this point - I just
    > need confirmation that I am using it..


    Well, there may be some sort of an "O_DIRECT" flag in a system call.
    Otherwise, for a vxfs filesystem, I believe there is an I/O size that
    will automagically trigger direct I/O, the vxtunefs command may be the
    one to see/set that value. Armed with that knowledge you could I
    suppose ass-u-me that any IO you see in the tusc trace that was at
    least that large bypassed the buffer cache, and presumeably any IO
    smaller than that did not.

    You might also look at buffer cache and disc I/O stats from Glance,
    might be a way to check there.

    rick jones
    --
    oxymoron n, Hummer H2 with California Save Our Coasts and Oceans plates
    these opinions are mine, all mine; HP might not want them anyway...
    feel free to post, OR email to rick.jones2 in hp.com but NOT BOTH...

+ Reply to Thread