The Linux Letter for July 12, 1999
RAID...it's not just for bugs anymore! Instead of tracking down
those pesky cockroaches with a can of Raid, this week we're
going to look at computer RAID systems. RAID is a mechanism that
can allow you to create faster, more reliable storage systems
without spending a ton of money.
RAID is an acronym for Redundant Array of Inexpensive Disks. And
with the price of today's hard drives plunging like my stock
market picks, inexpensive is the word! RAID allows you to
configure sets of disks to appear in several different ways.
RAID 0 is also called "striping". With this method,
several disks are combined to appear as one, by dividing the
single logical disk into "stripes" and interleaving
them together.
RAID 1 is "mirroring". With this method, your system
maintains two complete copies of the same data on two separate
hard drives. If one drive fails, you still have an identical
copy to fall back upon.
RAID 4 is similar to RAID 0, but adds an extra partition for
parity information. If a drive fails in a RAID 4 array, the data
can be reconstructed from the parity data and from the data
contained on the other hard drives.
RAID 5 is similar to RAID 4, but uses striping to store the
parity information.
There are other RAID levels that have been defined by industry
groups, but they are either obsolete or are proprietary. Linux
supports RAID 0, 1,4 and 5, as well as RAID Linear which simply
adds one partition to the next, creating a large logical drive
without striping.
Now that you know what it is, how do you use it? With Linux, you
have two choices. You can purchase a hard drive controller that
supports RAID naively, or you can use Linux's software RAID
tools. Since my goal in managing computers is to do it as
cheaply and reliably as possible, I opt to use the software
tools.
Of course, the first thing to do before you attempt to configure
your computer to use RAID is to determine if you need it. RAID
is a tool that allows you to potentially increase the
reliability and performance of your system. But it can also
potentially cause disasters. Since RAID 0 and RAID Linear do not
perform any kind of parity functions, if one disk dies, all of
the data on the other drives could be rendered useless. This
should serve as a warning that if you use RAID, you must have a
frequent, reliable backup program in place.
Now that I've said that, if you're still ready to configure your
drives as a RAID system, you're in luck because much of the
support for RAID is part of the Linux kernel. And even more
fortunately, there is an excellent document on the web to assist
you: the Software RAID HOWTO.
Consider a couple of other issues. While it may seem obvious,
you do need to have more than one physical hard drive. A single
drive RAID array doesn't make much sense. Also, while RAID will
work just fine with IDE hard drives, there are some drawbacks.
IDE uses quite a bit of processor power to work, especially
compared to SCSI. But help may be on the way. Abit has released
a new motherboard that contains two complete IDE interfaces,
allowing up to 8 devices to be installed. The controller has its
own processor that is supposed to reduce the load on the CPU,
allowing performance similar to that of SCSI drives. Although
Linux drivers for this controller are not available just yet,
they are under development, and you can be sure that within a
few weeks you will be able to use this system under Linux.
Perhaps the most important part of creating a RAID array under
Linux is to have a plan. The HOWTO provides a lot of guidance on
configuring the tools that you'll need and issuing warnings
about what terrible things will happen to you if you don't heed
them, but perhaps the most important thing to pay attention to
is the section on planning your array. You really must create
the proper partitions in the proper sizes and mount them in the
proper order or instead of having a well-performing, reliable
array of disks, you'll have a big mess. Don't let that happen to
you!
Finally, as a word of advice, if you are considering installing
a RAID array on a Linux system that uses multiple processors
(Symmetrical Multiprocessing, or SMP), you absolutely must be
running the 2.2.xkernel. In fact, if you're running any SMP
system, you should run that kernel because of the huge
improvements made to the SMP subsystems.
So, if you're ready to learn more, try these first:
Software RAID HOWTO: http://metalab.unc.edu/LDP/HOWTO/mini-Software-RAID.html
Software RAID: http://tunelinux.com/bin/page?hardware/ide/software-raid.html
