What are the advantages of RAID setup?

  • Thread starter Thread starter Rich
  • Start date Start date
R

Rich

Please help educate me.

I am considering setting up a RAID system on my P4 computer. I am
anticipating that the advantage is mostly speed of access increase for
a striping setup.

What are the pros/cons of such a system? Can anyone give me a target
figure on access speed increase? I am currently running ATA100 with a
7200rpm, 8Mb HD.

TIA, Rich
 
Rich said:
Please help educate me.

I am considering setting up a RAID system on my P4 computer. I am
anticipating that the advantage is mostly speed of access increase for
a striping setup.

What are the pros/cons of such a system? Can anyone give me a target
figure on access speed increase? I am currently running ATA100 with a
7200rpm, 8Mb HD.

TIA, Rich

Pro: can be somewhat faster. Con: far less reliable: one slight error on
either drive and you've effectively lost the content on both drives.

Personally, given the speed and capacity and price of modern IDE drives I'd
far rather have mirrored RAID and forego the possible speed increase in
favor of the reliability. Face it, with the size of modern drives it is
becoming increasingly impossible to do decent backups. In the old days (10
years ago?) I could plug a tape into my machine every night, start a backup
and be done with it. Now, how does one affordably back up a 250Gb drive so
easily and cheaply?

Just my $0.02 of course...
--
John McGaw
[Knoxville, TN, USA]

Return address will not work. Please
reply in group or through my website:
http://johnmcgaw.com
 
John said:
Pro: can be somewhat faster. Con: far less reliable: one slight
error on either drive and you've effectively lost the content on
both drives.

Personally, given the speed and capacity and price of modern IDE
drives I'd far rather have mirrored RAID and forego the possible
speed increase in favor of the reliability. Face it, with the size
of modern drives it is becoming increasingly impossible to do
decent backups. In the old days (10 years ago?) I could plug a
tape into my machine every night, start a backup and be done with
it. Now, how does one affordably back up a 250Gb drive so easily
and cheaply?

One installs a second drive and clones the first with XXCOPY (See
xxcopy.com). Any further backups are automatically incremental,
and go much faster. Then, WHEN the first drive fails, you have a
copy of everything. Your backup is probably done within 15
minutes, and may even be done in the background. For Linux, you
will have to build the equivalent yourself as far as I know.
 
"John McGaw" said in news:[email protected]:
Pro: can be somewhat faster.

I suppose if you went to all the bother of setting up RAID but then
foolishly put the 2 hard drives on the same controller then, yes, the
bandwidth available for both drives in aggregate would be only nominally
better. However, if you're going to go to the effort of using RAID, you'll
want to put the drives on separate controllers (i.e., on different IDE or
SATA ports). Then you'll get a near doubling of bandwidth (having the 8MB
buffered drives will also help).
Con: far less reliable: one slight error
on either drive and you've effectively lost the content on both
drives.

Which proves that for a server you don't want to use this configuration.
For home use, and if you have another drive (internal or external) or use
CD-R[W] (although you'll probably want to get a DVD-+RW drive) then striping
is great.
... I'd far rather have mirrored RAID and forego the possible
speed increase in favor of the reliability. Face it, with the size of
modern drives it is becoming increasingly impossible to do decent
backups.

Mirroring a drive is *not* backing it up. You are providing disaster
recovery in case the primary drive goes bad. The mirrored drive has the
same contents as the primary drive, so where are your backups? They aren't
on your mirrored drive. You are "backing up" your hardware when using a
mirrored drive. You are NOT backup up your files in the sense that you have
historical copies from which you can restore.

Rich,

If you really want to go to the bother of using RAID to get a lot better
bandwidth (i.e., effective speed) for your drives, and if this is for
personal use where the box is not critical and there is no requirement that
it always be up (or be able to be brought back up under an hour), then RAID
0 is fine. Just be sure the 2 drives are very similar. Best to get 2 of
the same brand and model. However, as John mentions, you are interlacing
(striping) the data for a file across the 2 drives (unless the file is
smaller than the stripe size) so any file error, surface defect, or other
error in the file renders that file corrupted. Corrupting a file does not
corrupt the entire striped volume anymore than a corrupted file would screw
over your file system now.

If you RAID, you *must* consider how you are going to backup your data ...
unless you consider your data absolutely worthless (and also your time
absolutely worthless to reinstall the OS or applications and redo all the
configurations and customizations). With a DVD-+RW drive, you get a
reasonably capacity on each disc. You get get pretty good storage capacity
on some tape formats, like DLT, but tape is excrutiatingly slow and all
commands are serial. You don't mention what is the size of your current
hard drive (the 8MB you mention is the buffer size). If you currently have
a 120GB drive, get another 120GB drive (same brand and model if possible).
That will give you 240GB striped RAID at twice the bandwidth of a single
drive. But consider that you'll have to buy another 120 GB, or bigger, on
which to store the compressed backups or to save disk images unless you want
to swap a lot of CD/DVD discs or tapes.

For personal use, and although it sounds great to have the potential of
double the bandwidth for drive throughput, few end-user applications would
sustain such a load on the drives. Unless for some reason your "personal"
computer is running an SQL database with huge-sized multiple db files that
are being concurrently accessed by a hundred users, or more, or some such
high disk usage setup, you won't get hardly any noticeable speedup from
using RAID when writing your letter in Word, browsing with IE, or posting in
newsgroups. You must actually have something that will put such a huge I/O
load on your system to make RAID worthwhile.
 
*Vanguard* said:
"John McGaw" said in news:[email protected]: snip...

Mirroring a drive is *not* backing it up. You are providing disaster
recovery in case the primary drive goes bad. The mirrored drive has the
same contents as the primary drive, so where are your backups? They aren't
on your mirrored drive. You are "backing up" your hardware when using a
mirrored drive. You are NOT backup up your files in the sense that you have
historical copies from which you can restore.

Rich,

If you really want to go to the bother of using RAID to get a lot better
bandwidth (i.e., effective speed) for your drives, and if this is for
personal use where the box is not critical and there is no requirement that
it always be up (or be able to be brought back up under an hour), then RAID
0 is fine. Just be sure the 2 drives are very similar. Best to get 2 of
the same brand and model. However, as John mentions, you are interlacing
(striping) the data for a file across the 2 drives (unless the file is
smaller than the stripe size) so any file error, surface defect, or other
error in the file renders that file corrupted. Corrupting a file does not
corrupt the entire striped volume anymore than a corrupted file would screw
over your file system now.

If you RAID, you *must* consider how you are going to backup your data ...
unless you consider your data absolutely worthless (and also your time
absolutely worthless to reinstall the OS or applications and redo all the
configurations and customizations). With a DVD-+RW drive, you get a
reasonably capacity on each disc. You get get pretty good storage capacity
on some tape formats, like DLT, but tape is excrutiatingly slow and all
commands are serial. You don't mention what is the size of your current
hard drive (the 8MB you mention is the buffer size). If you currently have
a 120GB drive, get another 120GB drive (same brand and model if possible).
That will give you 240GB striped RAID at twice the bandwidth of a single
drive. But consider that you'll have to buy another 120 GB, or bigger, on
which to store the compressed backups or to save disk images unless you want
to swap a lot of CD/DVD discs or tapes.

For personal use, and although it sounds great to have the potential of
double the bandwidth for drive throughput, few end-user applications would
sustain such a load on the drives. Unless for some reason your "personal"
computer is running an SQL database with huge-sized multiple db files that
are being concurrently accessed by a hundred users, or more, or some such
high disk usage setup, you won't get hardly any noticeable speedup from
using RAID when writing your letter in Word, browsing with IE, or posting in
newsgroups. You must actually have something that will put such a huge I/O
load on your system to make RAID worthwhile.
You will notice that I didn't write that having mirrored raid is the same as
having a backup. It is simply more reliable than having a single drive and
FAR more reliable than using a striped array. The difficulty of doing a
backup is the reason I would prefer mirrored over striped when the data
involved has any value at all. If a backup is to be of any real use it must
be current. And the backup ideally should be placed off-site. And there
needs to be a historical record going back at least three (in my experience
anyway) generations if not further. And the backup has to be dead reliable.
And it must be essentially hands off once it starts unless it is impossibly
quick. This makes DVD-RWs seem impossibly small for the task and juggling
even a handful of discs is anything but quick or hands off. And having three
sets of portable hard disks seems impossibly expensive. What the world needs
is a good cheap reliable tape system that can hold 1/2 terabyte!
--
John McGaw
[Knoxville, TN, USA]

Return address will not work. Please
reply in group or through my website:
http://johnmcgaw.com
 
Rather than get into why data backups are good and why hardware backup is
good, we first need to find out if Rich even does any backups now. If he
doesn't backup now, he doesn't care about his data or time. His data is of
no value to him, or it can be rebuilt, or recovered somewhere from
elsewhere. His time is of no value to him to reinstall the operating
system, reinstall his applications, and to configure and customize it all
again. Talking about RAID 1 to someone who doesn't do backups now is a
fruitless argument. They aren't backing up now. They won't be backing up
later.

Your point is that adding more drives in a striped RAID 0 set will reduce
reliability. Yes, it will. If the MTBF (mean time between failures) for a
hard drive was 3 years and you add another in the same RAID 0 set then you
get 1.5 years for MTBF. But a non-RAIDed drive with 3 years MTBF could
still tomorrow. If Rich isn't backing up now, he doesn't care about data or
hardware recovery. So RAID 1 won't be an option for him.

I also suspect if Rich is asking about RAID 0 and never mentioned RAID 5
then his RAID controller only supports RAID 0 and 1 (and maybe 0+1). It's
likely he is looking at using an onboard RAID controller on his motherboard.
If Rich had the availability of RAID 5, and if he could afford 2 more drives
instead of just one (since he isn't going to spend any money on backup
drives and media), then he should go with RAID 5 with 3 drives.

It comes down to waiting for a response from Rich to find out if he does
backups now and, if so, what backup device he has or will get. If he
doesn't backup now, does he want to spend the money on yet another drive to
only have it used as a backup with mirroring in RAID 1? Not likely. He
doesn't do backups! If he is willing to spend the extra money to
incorporate backup (data or hardware), will he spend it on the second drive
and use mirroring in RAID 1 (and still not have the speed advantage of
striping), or would he rather spend it on a tape or DVD-+RW drive so he has
the advantage of removable media that can be used in another drive (so he
doesn't have to worry about the drive going bad and becoming unusable for
restoring his system) and even restor to another host? RAID 1 is used to
keep the host running or get it back up quick, not for backing up your
files. RAID 1 is for hardware recovery, not data recovery. Or does Rich
still not bother with backing up his computer and use that same money that
would otherwise be for a backup device (mirrored drive, tape drive, CD-RW
drive, DVD-+RW drive, or external hard drive) and use it for a 3rd drive so
he has 3 drives and can use RAID 5 which offers the speed advantage of
striping along with hardware recovery? Instead of spending money on backup
hardware whether for data or hardware recovery which he does not do
presently, he instead spends it on the 3rd drive which gives him the speedup
plus covers his butt with some hardware recovery. He still has no data
backups (but he didn't have them before, either), he gets his need for speed
satisfied, and he gets some hardware protection despite his lack of
enthusiasm for backing up his data.

I can promote data backups until I'm blue in the face, but if the user
doesn't do it now then it is very unlikely they will employ it later. In
that case, RAID 1 is not an option to that user. RAID 0 is a more
appropriate choice *if* they really have a need for the increased
*potential* bandwidth it can offer. Yes, RAID 0 has reduced reliability but
then the user doesn't care because they're not doing backups now. You don't
rely on a higher MTBF to protect your data - if it's worth protecting. A
RAID 1 setup that has an MTBF of 1.5 years could run a decade before a
hardware fault killed that RAID set. A non-RAID single-channel drive (his
current setup) with an MTBF of 10 years could die tomorrow. A RAID 1 set
with 2 drives each having an MTBF of 10 years could die tomorrow because
static or a surge fried both drives, or the single controller got fried and
has to be replaced (unless you really get pricey to employ duplexing), or
the CPU fried, or whatever occurs within the same host in which the mirrored
drive coexists with the primary drive. The point is that backups which are
on separate media that is removable from the host and usable on replacement
drives or even in substitute hosts is how you backup your data. RAID 1 is
how you provide disaster recovery of your hardware. Unless Rich is using
his computer as a server (and used in a network where other users need it)
then there's little point in providing disaster recovery except based on
your own personal needs and budget.

Alas, when talking to newsgroup posters that inquire how to recover a file
that has been thoroughly deleted and telling them to use their backups, the
response is akin to deer caught in headlights: you get a vacant stare
implying "Backups?", a roll of the eyes, and then "Um, any other
suggestions?". No RAID setup supplants the need for a backup scheme. Don't
confuse hardware disaster recovery with data recovery. RAID 0 gives you a
*potential* speed increase with the reduction in *probable* failure time.
RAID 1 gives you disaster recovery provided only the drive went bad and not
the controller, motherboard, CPU, memory, or other common hardware, but you
get no speedup. RAID 5 gives you the speedup along with the hardware
disaster recovery. None provide you with backups. It is unfortunate that
"backup" can be used to discuss saving a historical image of your files on
removable media or to discuss how to maintain your hardware presence.
Backing up your data is not the same as backing up your hardware.
 
Back
Top