Re: 2Gb Exp problems 8.1.7

From: DA Morgan <damorgan_at_exesolutions.com>
Date: Fri, 14 Feb 2003 13:21:23 -0800
Message-ID: <3E4D5DD3.6308099F@exesolutions.com>

Frank wrote:

> DA Morgan wrote:
> > Billy Verreynne wrote:
> >
> >
> >>Frank wrote:
> >>
> >>
> >>>Could some OS guru elaborate on why some OS's only handle 2GB?
> >>>Seems to me these must be 31 bit then. (AIX 4.x had a special
> >>>'add-on' to go beyond 2GB, looked like the MSB was enabled, so to speak)
> >>
> >>Won't call myself an OS guru, but I will have a bash (don't like ksh or csh
> >>that much ;-).
> >>
> >>Firstly, it has _nothing_ to do with operating system and _everything_ to do
> >>with the file system.
> >>
> >>And this type of problem has been around for a long while. Back in my
> >>mainframe days, the filesystem used on the Siemens BS2000 o/s (still IMO
> >>one of the best mainframe o/s's ever) used a single byte for storing extent
> >>number. Thus you were limited to 255 extents. (wasn't this also the case
> >>with Oracle extents a few versions ago?)
> >>
> >>Using DOS old FAT16 filesystem is no different. The File Allocation Table
> >>has also size restrictions. Thus, a completed path of a subdirectory can
> >>only be (if I recall correctly) also 255-256 characters in total. When you
> >>reached that limit, you can not create any more subdirectories.
> >>
> >>Okay, now for ionodes and Unix. A file is represented by a structure called
> >>an inode (similar to a FAT entry).
> >>
> >>Each inode contains stuff like file type, access rights, timestamps, size,
> >>pointers to data blocks, and so on. A bunch of data blocks make up a file.
> >>Their addresses are stored in that file's inode.
> >>
> >>An ionode is of a certain size. Thus, you are limited to the number of data
> >>block pointers (or addresses) you can store in the ionode of a file.
> >>
> >>Here are a table listing Unix filesystems used by Linux through the years:
> >> Minix ext ext2
> >>Max fileystem size 64MB 2GB 4TB
> >>Max file size 64MB 2GB 2GB
> >>
> >>Because operating systems are closely tied to specific filesystems (often
> >>only running on a single specific filesystem), filesystem limitation such
> >>as the 2GB filesize limit is seen as an operating system issue. Which is
> >>not really the case as that limitation belongs solely to the filesystem.
> >>
> >>--
> >>Billy
> >
> >
> > You are correct but I don't see the separation as clearly as you do. If an O/S
> > doesn't support a file system it is the O/S that is the issue.
> >
> > In the case of some O/Ss they default to 2GB files and default parameters in the
> > O/S can be changed to allow larger files. The makes it look, again, like the O/S
> > to me.
> >
> > Daniel Morgan
> >

>

> First of all - Billy is right in stating it is a file system issue, not
> OS. I must become more precise...

>

> And Daniel: what about Microsoft's 536MB "barrier", 2GB, 4GB, ...?
> Doesn't MS have DOS (in two flavors, if I may believe the Linux
> Installer/Disk Druid: FAT12/FAT16), NTFS, FAT32? Afaik, you can
> still install Win2K on a DOS16 partition - it will effectively
> limit your file (and partition!) sizes, but you can. You can also
> migrate to NTFS of FAT32, allowing for much larger files/partitions.

>

> OS/2 could use HPFS (superior to NTFS, in my opinion), or FAT.

>

> I remember burning a special CD off the standard NT Server CD, with
> drivers of sp4 (iirc), just because that allowed you to install
> NT in a larger-than-2GB partition. Until sp4, you could not create
> a boot partition, larger than 2GB.

>

> Don't many of today's Linux distro's give you a choice of ext2,
> ext3, reiserfs, journalled rfs, or a combination of all of the above?

>

> If you still want to call it an OS issue, fine with me, but I hope
> I have made it clear with the above examples, that it is more of
> a file system issue.
> OS's (even MS!) support multiple file systems; it's up to the user
> to decide what's best.
> --
> Regards, Frank van Bortel

If you limit the discussion to Intel boxes ... but how about Sun, HP, and IBM (AIX) which is where most of the serious Oracle work is done?

Daniel Morgan Received on Fri Feb 14 2003 - 15:21:23 CST