How to backup using RAID 1 without software

Hi Everyone,

This is an idea I've been playing with but haven't implemented yet.
What do you think?

Backup procedure using RAID 1 (mirroring) without using a separate backup software:

The computer has a bootable RAID 1 controller either on the motherboard or a separate card. Three identical hard drives would be used. Two of the hard drives would be in mobile racks mounted in the computer. The third hard drive would be in an extra mobile rack tray stored offsite for swapping during the backup (synchronize/rebuild) procedure. The OS/Programs/Data would be all on the RAID 1 array. Most modern RAID controllers allow for rebuilding in the background and even have little windows array monitoring utilities.

When you want to backup your computer, you shut down and remove one of the hard drives and swap it with the one stored offsite(containing data from your last backup).
You boot up and enter the RAID BIOS. You rebuild and synchronize the data FROM the new data drive TO the old data drive. You can select to rebuild the array immediately right there in the RAID BIOS or in the background as you work.

When you want to restore data, you follow the same procedure, except you rebuild and synchronize the data FROM the backup old data drive TO the new data drive.

From what I can tell, the pros are:
No extra software. The backup is done by the hardware.
Simple to do.
Backup (rebuild) while you work.

The only cons I see are:
Poor Hard Drive performance during the rebuild which may take a while and slow down your work. But would this performance be any worse than any other imaging software that makes a snapshot in the background while you work?
If you make changes to the Hard Drive (like creating or editing a file) during the background rebuild, does it handle the new updated data properly and synchronize it to the mirror drive?

So, what do you think?

 

So, any RAIDers want to comment?

 

Any particular reason you are looking into a RAID setup?
Where do your concerns lie?
Why three hard drives and not just two?

From the sounds of it, you are looking into hardware mirroring with your RAID 1 setup. I ask what are your concerns, because you may also consider looking into duplexing (where one controller is connected to one drive and a second controller is connected to the second drive). This however (will most likely) require the use of the OS to run the RAID implementation (software). I believe Windows 2000 Server can run a RAID 1 setup with duplexing. The advantages are that you are also protecting yourself from a controller failure by using duplexing. The disadvantages are that performance will obviously be slower since it will rely heavily on the OS.

I think it is also important to note that while your data will be protected in a RAID 1 setup if one drive physically dies, you will NOT be protected if you encounter data loss or file system corruption,

Also in a RAID 1 setup your storage efficiency and storage capacity is usually about half.

Might consider those cons....

Overall you seem to have already put a lot of thought into the setup but if you (or anyone else) is/are interested to read more about it... PCguide offers some very good info... you can pretty much search anything related to RAID and you will get some good hits.

http://www.pcguide.com/ref/hdd/perf/raid/

http://www.pcguide.com/ref/hdd/perf/raid/levels/

heh you sure "simple to do" is a pro?

 

Hi rerun2,

Thank you for replying. My main concern is in regards to a controller that offers background rebuild.
What happens if you change a file or create a new file DURING a background rebuild process?
Will the controller correctly update or create the new file on both drives?
Or is this a vulnerable period where changes or additions to data should not be made?
Will Windows XP Pro with its page file and file locking cause problems for the rebuild.
The answers to these questions will directly determine if this idea is feasible for "backing up while you work in the background".

This would be for a workstation environment for data redundancy (in case of HD failure) and backup purposes. It is not a high availability environment (ie web server), so duplexing is not a concern. Besides, if a controller fails (on a card), one can just replace the card with an identical model. Software RAID is fine, but I would really rather have it done all in hardware.

This is exactly the reason why you still need to backup a RAID 1 array. This is my main concern why I started this thread. An offsite (or at least disconnected from the system) backup is still vital. So why not combine the great data redundancy of RAID 1 with the ability of RAID controllers to rebuild an array in the background and use it for backup purposes. This rebuilding process would be the "backup process". Why use a separate backup software when the process can be all done using RAID?

Yes, in a RAID 1 array, two 100GB drives would still leave you with a 100GB array. The write performance is also slightly lower than a single drive. However, the read performance of a RAID 1 array is better than a single drive because the controller is able to read data from one while searching for data on the other. I chose RAID 1 because it is the simplest, cheapest way to acheive data redundancy.

The third drive would be the offsite backup that would be updated (rebuilt) once a week. This is not the same as a "Hot Spare" which is used in a high availability system. The "Hot Spare" is a third drive that is always hooked up to the controller ready to instantly kick in should one of the other drives fail.
I guess what I want could be called a "Cold Spare" that would be swapped out with one of the 2 active drives, but used for backup purposes.

Here's the idea:
Drive A contains all data for week 1.
Drive B is a mirror of A and also contains all data for week 1.
Drive C is blank and not connected to computer.

1. Power down remove Drive B and store in a safe place. It will be the offsite backup for all data for week 1.
2. Put Drive C in it's place and boot up.
3. The RAID controller will recognize the 2 don't match and will warn during boot. Some controllers will halt the boot completely and ask to go into the RAID BIOS.
4. In the RAID BIOS, you select to rebuild the array (i.e. copy data from Drive A to Drive C).
5. The RAID BIOS will ask you if you want to rebuild immediately (remaining within the RAID BIOS) or in the background. Select background to keep working while you rebuild.
6. Once the rebuild is complete, you have a working RAID array (Drive A and C) plus an offsite snapshot of your data at that moment in time (Drive B).

When week 2 comes around:
Drive A contains all data for week 2.
Drive B is not connected to computer and contains all data for week 1.
Drive C contains all data for week 2.
You basically repeat steps 1 thru 6 and overwrite the old data(Drive B) with the new data (Drive C).
To restore data, you merely rebuild the array from the drive with the old data.

Those are good sites. I read a lot of info there about RAID when I was first learning. Very useful. Unfortunately, there isn't anything on this specific topic.

Okay fine, it's not simple.
But neither is learning how to setup and run an imaging software backup procedure that backs up to an external drive. The RAID BIOS of most controllers are very simple to use with few commands and options.

 

Hopefully I can answer some of your questions. Although i do not have experience with your "cold spare" setup, i imagine much of the process works in the same way.

As you know, when the raid controller receives a read/write from the OS it will perform two writes. So if the raid controller is doing its job it will keep the drives mirrored continuously in the background. With a hot spare... when a failure occurs you will most likely not even notice the rebuilding process as it will be instantaneous.

No, because these are functions of the OS. The raid controller will know what addresses are in use.

I agree. Since some RAID info is stored in the OS registry of a software setup; if the registry some how got corrupted, you will likely not be able to recover your array. Another trade-off to duplexing i guess, unless you have a higher end raid controller that supports duplexing.

Also be sure that each drive is set to Master on their own channel. This will prevent some lockups that i hear some people have experienced when they put their drives on the same channel.

 

Great. Then this backup procedure should work well if the RAID controller manufacturer implemented the background rebuild correctly.
This seems like a practical solution that will work. (at least for me)

Thanks rerun2!