RAID-1 for home backup: duplexing vs. mirroring; recommended harddisks

  • Thread starter Thread starter J.Clarke
  • Start date Start date
J

J.Clarke

At home my family has several computers connected to a wired (T100)
network. Since each of us produces a large amount of critical files
which we'd like to backup for an indefinite amount of time, as well as
to have some protection from unforeseen disasters such as a house
fire, we are considering setting up a RAID-1 on a special computer
dedicated for this purpose. This computer would be placed into a part
of the house so it can be better protected from fire, flood and other
disasters. Since a lot of data is personal, we really do not wish to
move the data over the network to somewhere outside of our property
(but would consider it -- our connection to the outside world is not
that fast, though.)

We envision this RAID-1 to run Windows 2000 Server.

2K server does fine, but Linux is free and will also do what you're
describing just fine. Go with one of the free alternatives and
use the thousand bucks or so you save on better hardware.
It will have three
IDE harddisks, two very large capacity disks for the RAID-1 disk
mirroring or duplexing, and the third disk to be a smaller one
sufficient to hold the OS.

Duplicate the OS disk, too, otherwise getting your data back is going to
be a major pain.
It will be a software RAID-1 system, since
the main purpose of this system is long-term storage only -- the
stored files will only be accessed on occasion so blazingly fast I/O
speed is not critical. This box will be run from a UPS with surge
protection. And it will be placed into an insulated brick/cement
enclosure in the basement so if a fire or whatever occurs, there is a
chance that at least the harddisks will survive (I'll be separately
researching how to buld this enclosure, but any recommendations you
have are welcome.)

Problem with the enclosure in the basement--in the event of fire where
do you think all the water the firemen pour on it is going to end
up? And immersion in dirty water is not good for disks either. You
could of course seal the enclosure but then you have cooling issues.

If you have a detached garage or other outbuilding with power, consider
putting a backup server there.
Now a couple questions:

1) Should I consider duplexing rather than mirroring (duplexing is
where each storage disk is run from its own controller -- mirroring
is where both disks run off the same controller). Will duplexing
even work in Win2KServer? (I'm essentially clueless on these
matters.)

This will work fine.
I prefer mirroring for now for cost reasons because it's the
integrity of the data, and not whether it is available 24-7, which
is the most critical. Of course, with mirroring I worry about the
situation when the single controller goes on the fritz and that it
damages the data on *both* harddisks. Or don't controllers damage
harddisk data when they go bad?

The situation you describe can occur, but there is another
consideration--with typical PC IDE controllers, the channels are not
well isolated, so a drive failure on one channel can corrupt data on a
drive on the other channel. Adding a completely separate controller
should address this.
2) What inexpensive, high capacity IDE drives (brand/make) should I
consider? I'm thinking in the 100 gig to 250 gig size. Which ones
are reputed to be more reliable than average? I don't necessarily
need high speed disks, just work horses. Anyway, the I/O of the
system will probably be limited by the home network speed, so I
really don't need to worry about high-speed harddisks, SCSI,
hardware raid, etc.

Pick any drive and somebody somewhere is going to tell you how he saw 90
out of a hundred of them fail. Provide clean power and adequate cooling
and any current drive should run for years.
Any other insights you can give on this issue, including web sites
which discuss how to do what I want to do, are more than welcome.

Thanks.

Bud


(p.s., a third question. What are my options for temporary media
backup of the RAID-1 system? Something that will handle 25 Gigs or
more? Or is that prohibitive in cost? I'm thinking I may wish to
periodically backup what's on the RAID-1 harddisks onto some media for
storage in a fire-proof safe or maybe even off-property somewhere.
This adds to the level of security of the data. Very critical data is
now being backed up on CD-ROM and kept in a small fire-proof safe, but
this is now getting to be unwieldy since the quantity of data is many
gigabytes now, and when I start digitizing family pictures and such,
will become a substantial fraction of a terabyte.)

Just a suggestion, but get a high-performance hardware RAID controller
such as the ones that 3ware sells, and set up a 3 drive raid with
hot-sparing and use removable drive trays. Put three drives in,
designate one a spare, and go. When you're ready, pull one of the two
active drives and put it in the safe deposit box or whatever and the
remaining drive will automatically be mirrored onto the spare. Then
when you're ready to swap again, just repeat the process, putting the
drive you removed back in and designating it the spare.

Simplest way to archive up to 5 gig these days is DVD+/-R. There are
those who say it's not stable, and if you're talking about something
that has to last over geological time they're right, but for ordinary
purposes it's fine. Above 5 gig you can go with tape or with drives
in trays. Hard disks don't cost a whole heck of a lot more than tape
cartridges with equivalent capacity these days, so which you go with is
more a matter of philosophy than anything else.
 
At home my family has several computers connected to a wired (T100)
network. Since each of us produces a large amount of critical files
which we'd like to backup for an indefinite amount of time, as well as
to have some protection from unforeseen disasters such as a house
fire, we are considering setting up a RAID-1 on a special computer
dedicated for this purpose. This computer would be placed into a part
of the house so it can be better protected from fire, flood and other
disasters. Since a lot of data is personal, we really do not wish to
move the data over the network to somewhere outside of our property
(but would consider it -- our connection to the outside world is not
that fast, though.)

We envision this RAID-1 to run Windows 2000 Server. It will have three
IDE harddisks, two very large capacity disks for the RAID-1 disk
mirroring or duplexing, and the third disk to be a smaller one
sufficient to hold the OS. It will be a software RAID-1 system, since
the main purpose of this system is long-term storage only -- the
stored files will only be accessed on occasion so blazingly fast I/O
speed is not critical. This box will be run from a UPS with surge
protection. And it will be placed into an insulated brick/cement
enclosure in the basement so if a fire or whatever occurs, there is a
chance that at least the harddisks will survive (I'll be separately
researching how to buld this enclosure, but any recommendations you
have are welcome.)

Now a couple questions:

1) Should I consider duplexing rather than mirroring (duplexing is
where each storage disk is run from its own controller -- mirroring
is where both disks run off the same controller). Will duplexing
even work in Win2KServer? (I'm essentially clueless on these
matters.)

I prefer mirroring for now for cost reasons because it's the
integrity of the data, and not whether it is available 24-7, which
is the most critical. Of course, with mirroring I worry about the
situation when the single controller goes on the fritz and that it
damages the data on *both* harddisks. Or don't controllers damage
harddisk data when they go bad?

2) What inexpensive, high capacity IDE drives (brand/make) should I
consider? I'm thinking in the 100 gig to 250 gig size. Which ones
are reputed to be more reliable than average? I don't necessarily
need high speed disks, just work horses. Anyway, the I/O of the
system will probably be limited by the home network speed, so I
really don't need to worry about high-speed harddisks, SCSI,
hardware raid, etc.


Any other insights you can give on this issue, including web sites
which discuss how to do what I want to do, are more than welcome.

Thanks.

Bud


(p.s., a third question. What are my options for temporary media
backup of the RAID-1 system? Something that will handle 25 Gigs or
more? Or is that prohibitive in cost? I'm thinking I may wish to
periodically backup what's on the RAID-1 harddisks onto some media for
storage in a fire-proof safe or maybe even off-property somewhere.
This adds to the level of security of the data. Very critical data is
now being backed up on CD-ROM and kept in a small fire-proof safe, but
this is now getting to be unwieldy since the quantity of data is many
gigabytes now, and when I start digitizing family pictures and such,
will become a substantial fraction of a terabyte.)
 
At home my family has several computers connected to a
wired (T100) network. Since each of us produces a large
amount of critical files which we'd like to backup for an indefinite
amount of time, as well as to have some protection from
unforeseen disasters such as a house fire, we are considering
setting up a RAID-1 on a special computer dedicated for this
purpose. This computer would be placed into a part of the house
so it can be better protected from fire, flood and other disasters.

Trouble with this approach is that it cant give the
ultimate in protection. Only offsite can ever do that.
Since a lot of data is personal, we really do
not wish to move the data over the network
to somewhere outside of our property

It isnt hard to encrypt it so that its very secure.
(but would consider it -- our connection to
the outside world is not that fast, though.)

It would normally be fast enough if the offsite backup
is intelligent so only changed data is moved over that.
We envision this RAID-1 to run Windows 2000 Server.

This wont protect you against the entire system getting infected
and the deliberate damage being propagated thru to that tho.
It will have three IDE harddisks, two very large capacity disks for
the RAID-1 disk mirroring or duplexing, and the third disk to be a
smaller one sufficient to hold the OS.

This doesnt really amount to the terabyte mentioned at the bottom tho.
It will be a software RAID-1 system, since the main purpose of
this system is long-term storage only -- the stored files will only
be accessed on occasion so blazingly fast I/O speed is not critical.
This box will be run from a UPS with surge protection. And it will
be placed into an insulated brick/cement enclosure in the basement
so if a fire or whatever occurs, there is a chance that at least the
harddisks will survive (I'll be separately researching how to buld
this enclosure, but any recommendations you have are welcome.)

That approach should be viable. BUT you are likely
to lose it in a flood, because its in the basement.
Now a couple questions:
1) Should I consider duplexing rather than mirroring
(duplexing is where each storage disk is run from its
own controller -- mirroring is where both disks run off
the same controller). Will duplexing even work in
Win2KServer? (I'm essentially clueless on these matters.)

That approach is more for higher uptimes and that isnt vital in
your situation. Simarly with duplexing over multiple machines.
I prefer mirroring for now for cost reasons because it's
the integrity of the data, and not whether it is available
24-7, which is the most critical. Of course, with mirroring
I worry about the situation when the single controller goes
on the fritz and that it damages the data on *both* harddisks.

Thats relatively rare. Its much more likely that its the
power supply that will blow up and kill both drives.
Or don't controllers damage harddisk data when they go bad?

Not often. It can obviously happen tho.

If you want the ultimate in security, you really need a pair of PCs.
2) What inexpensive, high capacity IDE drives (brand/make)
should I consider? I'm thinking in the 100 gig to 250 gig size.
Which ones are reputed to be more reliable than average?

No one really knows because the industry never releases
the statistics that matter. Not that important when duplicated
anyway. You can certainly make a case for a pair of PCs
and use a different brand of large drive in each one.
I don't necessarily need high speed disks, just work horses.
Anyway, the I/O of the system will probably be limited by
the home network speed, so I really don't need to worry
about high-speed harddisks, SCSI, hardware raid, etc.
Correct.

Any other insights you can give on this issue, including web sites
which discuss how to do what I want to do, are more than welcome.
(p.s., a third question. What are my options for temporary media
backup of the RAID-1 system? Something that will handle 25 Gigs
or more? Or is that prohibitive in cost? I'm thinking I may wish to
periodically backup what's on the RAID-1 harddisks onto some
media for storage in a fire-proof safe or maybe even off-property
somewhere. This adds to the level of security of the data.

Yes, that is essential for this sort of data. And
off property is the only really secure approach.
Very critical data is now being backed up on
CD-ROM and kept in a small fire-proof safe,

Offsite is MUCH more secure and isnt that hard to achieve.
but this is now getting to be unwieldy since the quantity of data
is many gigabytes now, and when I start digitizing family pictures
and such, will become a substantial fraction of a terabyte.)

The obvious way to handle the higher volume stuff like that is
to write it to more than one DVD when its produced/edited etc
and keep one copy offsite. Then the total volume isnt really that
significant because the act of writing it to the multiple DVDs is just
a small bit of time at the end of the creation of that particular stuff.

If you arent very conscientious, it might be best to automate
the reminder on whats been written to DVDs, but thats not hard.
 
bud:
actually with mirroring the disk can be on the same or different ide controllers.... duplexing is mirroring, just the disks are on separate controllers...

My experince with software raid-1 is that it basically stinks.....as far as disaster recovery is concerned....

The software raid-1 ie dynamic disks in win2k server is basically unrecoverable with normal disk tools....

The software raid built into motherboard controllers also sucks....

Hardware raid is the way to go for raid-1....and there are reasonably priced hardware raid scsi and ide cards available...

I would consider external removable firewire or usb 2 disks and use ghost or drive image to automate the imaging of the internal disk to external disk on a more or less regular automated basis...you can do this with win2k...

I have been doing automated full system backups on a daily basis using ntbackup in win2k for years now....its fast and its very recoverable in terms of disaster recovery....

shockie B)
 
shockwaveriderz said:
My experince with software raid-1 is that it basically stinks.....as far as disaster recovery is concerned....

The software raid-1 ie dynamic disks in win2k server is basically unrecoverable with normal disk tools....

I don't fully understand. Can you elaborate more?

All I care is that the data on at least one of the two backup
harddisks of the RAID-1 configuration maintains its integrity, so in
case some disaster wipes out one of the backup disks, I can pull off
the data on the good drive to another drive on another computer until
I fix the RAID-1 system. So, is hardware RAID-1 worth the extra cost
if software RAID-1 gives us this capability?

Again, the purpose of this auxiliary computer is to simply hold
important data for the family for occasional access (maybe once
every few days) and updating (maybe once every couple weeks.)
Periodically, maybe once a month or every two months, I'll copy the
data from the RAID-1 onto a third, removable harddisk for storage in
a fireproof safe (so in a worst-case scenario of losing the whole
RAID-1 data, we'll only lose the most recent.) Extremely critical
data, which probably amounts to no more than a few hundred megs, if
that, will also be burned to a CD-ROM and stored in the fireproof safe
along with the removable harddisk (and may even be stored offsite,
like a friend's house.

(Anyone have suggestions for a fireproof box with a lot of space? I
don't really need a safe capability, although I suppose that with all
the fireproofing which makes it safe-like in the end that adding a
lock to it is not a big deal.)

I've also decided, after pricing offsite storage (it's quite
expensive for the amount of storage I need), that I'm not doing that
option. I figure our family backup will be about 50 gigs in a few
months, and probably grow to about 100 gigs in a couple years.

Of course, I appreciate the comments so far, and welcome any
others. I assume I'm not the only person with these particular needs.

Bud
 
Bud Freeman said:
I don't fully understand. Can you elaborate more?
Most conventional disk tools such as Drive Image,ghost,partition magic,etc will not work on SOftware raid , dyanmic disks in win2k server.... To use software raid-1 in win2k, you convert conventional,traditional disk with master boot records,etc into dynamic disks......If they get screwed then its been my experience that you get screwed and they are baiscally unrecoverable.... Others mileage here may vary ....
http://support.microsoft.com/default.aspx?scid=kb;EN-US;175761
http://support.microsoft.com/default.aspx?scid=kb;en-us;Q298155
http://support.microsoft.com/default.aspx?scid=kb;en-us;308209&Product=win2000
http://support.microsoft.com/default.aspx?scid=kb;EN-US;222189

Now you may be considered a Pc that has a software raid 0 or 1 built into say a chipset....this is really "firmware" raid and agsin recovery can be difficult....I don't have much experience with these chipset raids.....

Again, why not just do an automated full system backup on a daily basis in the middle of the night to external firewire or usb drives? you can detach and attach them at will, place them in a fireproof safe,etc.... Get a 250GB disk and do a disk to disk backup with ntbackup or get ghost or drive image and image them with scripts to automate the process..Or get Dantz Retrospect Backup as you can automate it to do a disk clone.....from one disk to another....

Raid-1 should not be used as a backup and diaster recovery solution.....
I don't think that IDE hardware raid-1 is all that expensive......

All I care is that the data on at least one of the two backup
harddisks of the RAID-1 configuration maintains its integrity, so in
case some disaster wipes out one of the backup disks, I can pull off
the data on the good drive to another drive on another computer until
I fix the RAID-1 system. So, is hardware RAID-1 worth the extra cost
if software RAID-1 gives us this capability?

usually raid-1 is used on the operating system such that if the primary disk fails, the system will continue to run off the secondary disk in the disk mirroe set without the system failing....hence its fault tolerance.....I don't know how well software rauid can perform this function.....I do know that with hardware raid-1.....you can actually pull the power switch on the primary disk and the system will run uninterrupted........

Raid-1 should not be used as a backup and diaster recovery solution...

shockie B)
 
Back
Top