comp.os.linux.hardware, comp.databases.postgresql.general - Hardware

This is a discussion on comp.os.linux.hardware, comp.databases.postgresql.general - Hardware ; totally disabling caching is neccessary in some cases, for example, some tables used as transactional checkpoints or tables used to make sure some some actions are recorded, like users loging in and out and payment confirmation records. .. using hdparm ...

+ Reply to Thread
Results 1 to 2 of 2

Thread: comp.os.linux.hardware, comp.databases.postgresql.general

  1. comp.os.linux.hardware, comp.databases.postgresql.general

    totally disabling caching is neccessary in some cases, for example,
    some tables used as transactional checkpoints or tables used to make
    sure some some actions are recorded, like users loging in and out and
    payment confirmation records.
    ..
    using hdparm a' la "hdparm -W0 /dev/hda" in order to disable internal
    drive write caching helps. Using a journaling file system such as
    ext3, XFS or reiserfs helps too and accordingly tinkering with the
    BIOS' "block device's settings" is also beneficial
    ..
    an easy and safe way to have some tables in a database with no cache
    enabled in the disk controller is just having these tables in separate
    disks
    ..
    based on your experience, what else should be taken into
    consideration?
    ..
    which of the journaling fs have you had a good experience with? I have
    heard b-tree based ones are not so reliable.
    ..
    which consecuences could have disabling the cache in some disks and
    keeping it alive in some others? would it tax the memory, I/O
    subsystem?
    ..
    any good articles on these kinds of micro hardware optimisations?
    ..
    lbrtchx


  2. disabling caching on disk controllers

    totally disabling caching is neccessary in some cases, for example,
    some tables used as transactional checkpoints or tables used to make
    sure some some actions are recorded, like users loging in and out and
    payment confirmation records.
    ..
    using hdparm a' la "hdparm -W0 /dev/hda" in order to disable internal
    drive write caching helps. Using a journaling file system such as
    ext3, XFS or reiserfs helps too and accordingly tinkering with the
    BIOS' "block device's settings" is also beneficial
    ..
    an easy and safe way to have some tables in a database with no cache
    enabled in the disk controller is just having these tables in separate
    disks
    ..
    based on your experience, what else should be taken into
    consideration?
    ..
    which of the journaling fs have you had a good experience with? I
    have heard b-tree based ones are not so reliable.
    ..
    which consecuences could have disabling the cache in some disks and
    keeping it alive in some others? would it tax the memory, I/O
    subsystem?
    ..
    any good articles on these kinds of micro hardware optimisations?
    ..
    lbrtchx


+ Reply to Thread