Re: designing fast, streaming disk i/o with mmap: help wanted

From: Bill Huey (billh@burn.ucsd.edu)
Date: Fri Mar 31 2000 - 22:46:45 EST

  • Next message: david parsons: "Am I missing something (2.3.99-pre4-2) ?"

    Roger,

    > Assuming you always try to read tree blocks
    > you get a worse result:
    >
    > layout: A1 A2 B1 B2 A3 C1 B3 C2 C3
    > 1 1+
    > 2
    > 3 3+
    > 4
    > 5
    > --- first blocks from all channels read ---
    > 6 6+
    > --- all blocks read, but are they still around? ---

    > Would it be viable to limit read ahead to continuos blocks?

    Possibly, the hueristics for this kind of thing is fairly ad hoc as I know
    it.

    He's probably also going do stuff like sample streaming, which is fairly
    demanding on the disk head because of random seeks if I understand the
    technical problems clearly.

    Maybe sorting the the block address to minimize disk seeking
    would help, amougnst a grab bag of traditional things.

    There's also design trade offs that are unknown for this kind of
    technology that normal file system technology wouldn't find
    acceptable.

    Maybe having the streaming volume defragmented and disipline to
    a easily manageable data structure before use could be an option.

    It's complicated and largely out of my focus currently.

    I'll shut up now. ;-)

    > /RogerL

    bill

    -
    To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
    the body of a message to majordomo@vger.rutgers.edu
    Please read the FAQ at http://www.tux.org/lkml/



    This archive was generated by hypermail 2b29 : Fri Mar 31 2000 - 22:48:44 EST