File systems and hard drives have always been fascinating to me, ever
since I invested way too much money in a large (at the time) 80MB drive
for the computer I was using in the early 90s. There's just something
neat about blasting big chunks of data back and forth across your bus,
reading and writing thousands of files, and doing benchmarks.
Maybe
you don't share my infatuation of hard drives and the software that
keeps track of files and directories, but you're probably at least a bit
interested in keeping your data safe, using your drives efficiently,
and squeezing as much performance as possible from your system's meager
I/O subsystem.
Yeah, I called it meager. Unless you've got an
enormous budget for exotic hardware, your disk I/O subsystem has
advanced very little compared to things like your CPU, RAM, and video
card. Moore's famous law doesn't apply here, just micro magnetism and
advanced manufacturing processes.
Of the commonly used operating
systems, Linux® has the most support for different file systems. In
this respect, Linux diverges from other UNIX® systems, which have
traditionally supported their native file system and the ISO-9660 file
system used on standard CD-ROMs. My Fedora Core 4 system has loadable
kernel modules for about two dozen kinds of file systems. They exist
primarily to provide compatibility. You can plug in a disk from nearly
any other system and manipulate it from within Linux. But what if you're
adding another disk to your Linux system and don't need to let it speak
with Windows®, QNX, Mac OS X, or Minix?
You'll need to know at
least a little about some of the common Linux file systems, such as ext2
(second extended file system), ext3 (third extended file system), and
ReiserFS 4 (an interesting advanced file system with many exotic
features that can improve your system's file handling).
Before you start
If
you're not lucky enough to be running a Linux distribution that comes
with support for Reiser4 (such as Arch, Linspire, or SUSE), you're going
to have to do something that might be painful: rebuild your kernel.
Since rebuilding your kernel would be a tutorial of its own, you'll have
to find a how-to document for your distribution that leads you through
the steps required to rebuild the kernel.
Before you actually
start compiling your brand new Linux kernel, you'll also have to visit
the Reiser4 home page at Namesys (see
Resources)
and download the Reiser4 patches appropriate for your kernel. These
patches come with instructions for applying them before you configure
and compile the kernel.
So you'll be able to create and manipulate
Reiser4 file systems, make sure the reiser4progs package is installed.
Again, if your distribution doesn't have the resier4progs, see the
Namesys site for downloads (see
Resources).
If you want to experiment with Reiser4 easily, an excellent Gentoo Linux Live CD is available with Reiser4 support. See the
Resources section for a link.
Linux file systems
Back
in the early days of Linux (late 1993, in fact), only one file system
was supported by the kernel, a port of the rather minimal Minix file
system. This had several shortcomings, including 14-character filenames
and a mere 64MB as the maximum file system size. It didn't even support
the full set of UNIX file system attributes, specifically the full set
of creation, modification, and access time stamps required for a
conforming POSIX (Portable Operating System Interface) file system.
Because
of the limitations of the Minix file system, work on a replacement file
system began and resulted in the addition of a Virtual Filesystem (VFS)
abstraction layer, which would make it easier to write file systems for
Linux. Using the new VFS, the Minix file system was extended to allow
longer filenames and larger file systems (up to two gigabytes). This
became the extended file system (ext), but it still had several
limitations.
The solution was the ext2, which is still used on
many systems, and was the default Linux file system for quite a while.
The ext2 file system was improved by the addition of journalling, which
created the ext3 file system as its successor.
Before any of the
other journaled file systems were usable under Linux, ReiserFS (also
known as the Reiser3 file system) was the first advanced file system for
Linux that had support for journalling and more efficient disk usage.
Its successor, Reiser4, is a complete redesign and rewrite, focusing on
extensibility, security, and performance, while maintaining your data
safely and efficiently. Reiser4 isn't integrated with the Linux 2.6
kernel yet, which is usually a sign of possible instability or other
reasons to tread carefully. Remember to always back up your important
data, regardless of which file system you're using.
Let's take a quick look at the ext2, ext3, and Reiser4 file systems.
The traditional: ext2
The
ext2 file system, Linux's default file system, is a traditional UNIX
file system (that is, based on Berkeley's Fast Filesystem, FFS). It has a
maximum filename length of 255 characters and a theoretical maximum
file system size of four terabytes. (Linux's block device driver limits
this to "just" 2047 gigabytes; let me know when I can buy a single hard
drive with that much storage space).
Because of its ubiquity, ext2
drivers have been created for Windows and Mac OS X, letting you read
and write an ext2 file system directly from those operating systems,
making it an excellent format for shared devices, such as portable hard
drives.
The ext2 file system supports all of the standard UNIX features:
- Owner user ID and group ID.
- Mode bits controlling user, group, and other permissions and operating system flags.
- Storage
for entry creation, modification, and access times (although most
systems run with access time tracking turned off to improve performance
at the cost of pure POSIX 1003.1 compatibility).
The major
drawback for ext2 file systems is that hard drives have become much
larger since it was originally designed. A system crash or power failure
requires a file system check with
fsck, which takes ages on modern disks with lots of directories and files.
Traditional, but journaled: ext3
The ext3 file system for Linux is the new default file system for most popular distributions. Compared to ext2, it adds:
- A
metadata journal, to ensure that your file system structures are always
in a valid state. This (mostly) removes the need for a lengthy
fsck session after a system crash or power failure.
- Hashed tree directory indexes, to speed up access times for large directories.
- Online file system resizing and the ability to upgrade to ext3 from ext2 without reformatting the drive.
- Larger maximum file and file system sizes (two and 32 terabytes, respectively).
Although
ext3 isn't as fast, or scalable, as its competitors (such as Reiser3 or
SGI's excellent XFS), its compatibility with ext2 makes it attractive
because of the large number of mature ext2 maintenance and
administration utilities.
The Batmobile option: Reiser4
While
the Reiser3 file system gained some popularity because of its speed and
journalling support (it's the default file system for a number of Linux
distributions, even today), its designer wasn't satisfied. The Reiser4
rewrite grew out of this and added some additional interesting features:
- Efficient journalling through wandering logs.
- Efficient support of small files, in terms of disk space and speed.
- Fast
handling of very large directories with hundreds of millions of files
(yes, millions of files in a single directory without affecting
performance).
- Flexible plug-in infrastructure at the source
level, which allows easy addition of compression and encryption at some
point in the future.
- Atomic file system modification, ensuring that the on-disk file system is always in a valid state.
- Dynamically optimized disk layout through allocate-on-flush.
- Support for database-like file system transactions.
But
wait, why did I call it the Batmobile option? Reiser4 supports lots of
interesting features that you might never need to take advantage of and
some that you can't even access, because Linux's VFS layer doesn't
expose the functionality -- much like how the Batmobile has all kinds of
features you're not going to need on your way to the office.
Getting started with Reiser4
Before
you can do anything interesting with the Reiser4 file system, you'll
need to format a partition with it. As you can see from
Figure 1, I have a spare partition just waiting to be used for this purpose:
Figure 1. A partition waiting to be formatted
You need to create a new Reiser4 file system on this partition as well as get it mounted.
Creating a Reiser4 file system
To create the file system, you'll log in as root (or use the
sudo command to masquerade as root) and use the
mkfs.reiser4 command:
/sbin/mkfs.reiser4 -L myLabel /dev/hda1
|
This command creates a Reiser4
file system on the specified partition (/dev/hda1 in my case) with a
label of "myLabel" and a random unique identifier, as seen in
Figure 2.
Figure 2. Creating the Reiser4 file system
A fresh, clean Reiser4 file system awaits us! Let's mount it, so you can actually use it.
Getting it mounted
To mount the new file system, you'll log in as root (or use the
sudo command to masquerade as root) and use the
mkdir and
mount commands:
mkdir /mnt/reiser4
mount /dev/hda1 /mnt/reiser4
|
Use
mkdir first to create a mount point, then use
mount to mount the device containing the file system at the mount point you just created.
You can use the
mount
command without any arguments to get a list of the currently mounted
file systems, which should now include your new Reiser4 file system (see
Figure 3).
Figure 3. Mounted file systems, including Reiser4
Now that you've got your new file system mounted, you should make sure the system can mount this file system automatically.
Automatically mounting the file system
To let the system automatically mount your new Reiser4 volume, you need to describe it in the /etc/fstab file.
As root (or using
sudo to masquerade as root), edit /etc/fstab using your favorite text editor and add the text from the following code at the bottom:
/dev/hda1 /mnt/reiser4 reiser4 defaults 0 0
|
In /etc/fstab, you'll want to use
the appropriate device and mount point for your file system. After the
device and mount point, you specify the file system type and file system
options (defaults is usually best, unless you really know what you're
doing and you can find good documentation for your file system). The
last two options are for the "dump frequency" and "pass number on
parallel fsck", which are needed for historical reasons.
You can test to see if you've edited /etc/fstab correctly by unmounting the file system and telling the
mount command to mount everything automatically:
umount /mnt/resier4
mount -a
|
Now when you use the
mount command without any arguments, it should look exactly the same as it did before (
Figure 3). Your new Reiser4 file system will automatically be mounted at boot time, along with the rest of your file systems.
Tweaking performance and behavior
Like
most other Linux file systems, Reiser4 responds to some options that
can improve its overall performance and modify its behavior. These
options can be passed on the
mount command line using the
-o option, as seen in the following code:
mount -t reiser4 -o option1,option2,...,optionn /dev/hda1 /mnt/reiser4
|
Several file system options can be included in the command by separating them with commas.
So they're picked up at boot time, file system options can also be included in the /etc/fstab file:
/dev/hda1 /mnt/reiser4 reiser4 option1,option2,...,optionn 0 0
|
The most common options include:
defaults -- The default Linux file system settings. This is the same as specifying rw,suid,dev,exec,auto,nouser,async
directly. The file system will be mounted in read-write mode, set-UID
bits will be honored, character or block special devices on the file
system are treated normally, binaries can be executed, the file system
can be auto-mounted, and all I/O is done asynchronously, respectively.noatime
-- Don't update the access time field when files or directories are
read. This isn't strictly POSIX behavior, but it can speed up file
system operations significantly, especially on file systems with a large
number of directories and files that are mostly read from instead of
being written to.noexec -- Don't run binaries from
this file system. It's treated as a data-only file system, which can be
handy if you don't entirely trust the source of the files or
executables found there.nosuid -- Ignore the
set-user-ID and set-group-ID bits on files stored in this directory;
again, a safety feature if you don't trust the source of the files.ro -- The file system is mounted read-only. Attempts to write to a file or create a new file or directory will fail.data=journal
-- Instead of writing only file system metadata to the journal, all
data is journaled before being written out to the file system. This
ensures data integrity in the event of a disaster, but seriously reduces
write performance.
Using the defaults is generally what you want, but adding
noatime to the mix can safely help speed things up. The
data=journal
option might be a good idea for an important CVS server or a file
system used for making backups, because having valid data is much more
important than raw throughput.
Summary
Adding
a file system to your Linux setup can be daunting, especially when you
want to try out one of the many different file systems supported by
Linux. Knowing about the features and design limitations of the popular
file systems helps you to choose wisely.
Reiser4 should still be
considered experimental (even though many people are using it without
problems), because it hasn't been accepted into the Linux kernel yet.
The folks at Namesys are working hard to make their code part of the
kernel, so it's only a matter of time before distributions start
appearing based on Reiser4.
After creating your new file system with the appropriate
mkfs command, mounting it with the
mount
command (and adding it to the /etc/fstab file) allows you to begin
working with the file system. After that, remember to always back up
your important data regularly instead of when a hard drive fails.
Resources
Learn
http://www.ibm.com/developerworks/aix/library/au-unix-reiserFS/
Nama Domain Murah
