Re: Need help! Oracle8 instalation......

Solution Description:

  o Overhead, defined

    The "overhead" involves opening the files, checking the file headers and so forth - fairly benign stuff.

    Opening one 1Gig file will have less overhead than opening 20 50Meg files but this "overhead" only happens at instance startup

    and has little impact on the running application(s).

  o non-RAID

    On non-RAID storage facilities, it often makes a lot of sense to use many smaller files rather than one or two large files. The reason is

    the additional flexibility in scattering data over a number of physical disks, assuming of course that you have a group of smaller disks,say

    2Gigs each, rather than one or two larger disks, say 9Gigs each. If the disk hardware consists of a few large disks then there is no

    good reason to use multiple small files - they'll all be located on the same physical disk.

  o RAID

    With the advent of RAID technology, the need to scatter data across multiple physical disks is frequently obviated especially when used

    with the hardware vendors' logical volume manager (LVM) software. By striping data across multiple small disks (which is typical of

    RAID/LVM configurations), you can create a large logical volume of say 20Gbytes which has been strategically spread across 10-2Gig disk drives.

    Hence there is no need to create multiple small files; one large "virtual file" will actually consist of many physical stripes on many physical disks.

  o Other factors

    These additional factors will affect your choice of file sizing:

1. The nature of the database objects that you plan to store in the database
2. How the db objects will be accessed
3. CPU configuration (single CPU -vs SMP)

    Suppose you have the following:

• 20Gig table that will frequently be table scanned (ie. a DDS application)
• a storage array of consisting of 30 2Gig disks with 6 controllers
• 5 disks per controller
• 10 CPU platform.

    In this case I would create a tablespace consisting of 10 2Gig data files, each mapped to a different disk in the array. I would not use

    raid or disk striping since the table is accessed sequentially and I can start at least 10 query servers (because I have 10 CPU's).

    I would not exceed 2Gig file size to take advantage of async I/O and I would probably use raw devices since buffering I/O (Unix file system)

    will not be of much help in a sequential table scan.

    Now suppose I have the same table but it is accessed randomly in a OLTP application. In this case, the 10 CPU's will probably mean that

    there will be at least 10 concurrent requests for data in a random fashion so the goal here is to spread the I/O across all available

    disks so that they do not become a bottleneck. In this case I would probably use RAID/LVM to stripe the data randomly throughout the entire

    disk array. Depending on the need for async I/O I would either create 10 2Gig files as before or create a single 20Gig file in a 20Gig

    logical volume using cooked files to take advantage of Unix file system buffering.

"The Views expressed here are my own and not necessarily those of Oracle Corporation"

<lkj741_at_my-deja.com> wrote in message news:816m03$o04$1_at_nnrp1.deja.com...
> I am tring to instal oracle 8.5 on a
> 2gig drive...never instaled it before.
> What are the things that I should do or
> any idea. Please let me know....
>
>
> Sent via Deja.com http://www.deja.com/
> Before you buy.
Received on Sat Nov 20 1999 - 22:10:30 CST