NAS, backups, RAID0 and RAID1

[cereusb]

Hello,

First time NAS user. I've been reading lots of reviews and going through the forum (great site!) and I'd like to ask a bit of clarification/advice.

I have two general goals: a) backup and b) network accessible data. I don't want to backup my computer's hard drive to the NAS, I want to store on the NAS all the old photos, ripped movies for storage etc.

This would be for myself and my parents, not super users and don't need 100% uptime so features like hot swapping (while cool) seems a bit excessive. I run linux, parental units run windows but are reasonably tech savy (i.e. bitorrent users, bought router and wired house for 10/100, can stick in a hard drive).

I've read in several places that RAID is not backup. Or rather it seems half backup (backup against disk failure but not the controlling hardware failure) plus possible speed improvements. Correct(ish)?

My first question is if you *do* use RAID for backup (RAID1 mirror), and the hardware dies, does it kill the data on the drive? If the RAID disk is writing in just plain old ext3 instead of some weird proprietary alteration, couldn't you read the disk by sticking it in your computer (I read some variations about booting from live CD in the forum). Or if the hardware dies it could corrupt the disk and that's why its impossible to guarantee you could get data out.

So if you want backup PLUS network availability the best way (by measures of cheap and easy) might be - two NAS's (2 disk striped RAID 0) on network, one exists only to back up the second. This setup backs up disk failure, controlling hardware failure and maybe theft failure because the NAS's are kept separate. At this point, setting up the two NAS's for RAID1 seems a bit ... redundant.

Am I missing the reason for having the NAS and backup NAS set for RAID1 (RAID5)?

Also, as a bonus just asking, if anyone has hardware suggestions that would be great. What I'd love most is just make my own NAS, but I'm willing to give up control for a nice, little, quiet, low power, BYOD box.

Thanks in advance,
Cereusb

 

[thiggins]

My first question is if you *do* use RAID for backup (RAID1 mirror), and the hardware dies, does it kill the data on the drive? If the RAID disk is writing in just plain old ext3 instead of some weird proprietary alteration, couldn't you read the disk by sticking it in your computer (I read some variations about booting from live CD in the forum). Or if the hardware dies it could corrupt the disk and that's why its impossible to guarantee you could get data out.

It all depends on the failure mode. If you lose one drive of a RAID 1 set and the filesystem isn't proprietary, you can do as you propose. But if a power supply or controller board dies at the wrong time, then the whole array could be toast.

So if you want backup PLUS network availability the best way (by measures of cheap and easy) might be - two NAS's (2 disk striped RAID 0) on network, one exists only to back up the second. This setup backs up disk failure, controlling hardware failure and maybe theft failure because the NAS's are kept separate. At this point, setting up the two NAS's for RAID1 seems a bit ... redundant.

That is essentially what I advise in Smart SOHOs Don't Do RAID.
And, you're correct, RAID 1 is like belt and suspenders.

Also, as a bonus just asking, if anyone has hardware suggestions that would be great. What I'd love most is just make my own NAS, but I'm willing to give up control for a nice, little, quiet, low power, BYOD box.

All depends on the performance that you want. Take a look at the NAS Charts as a start for performance ranking. A pair of Buffalo LinkStations could work fine if you don't need high performance. The D-Link DNS-321's are inexpensive for a RAID 1 box and also can do NAS to NAS scheduled backup. Pretty speedy too.

 

[cereusb]

Thanks for the reply

Thanks Tim for the quick reply.

As for my system, I've been thinking about D-Link DNS-323 for the 1st main NAS and then something not so full featured for the backup second NAS. Thinking about backup choices will take some more reading though.

Cereusb

 

[thiggins]

Consider a DNS-321. Price is low and you only have to use one drive (same on the
DNS-323).

 

[valentin]

I have got a related question.

What do you think is a good backup device/strategy these days?
It is not possible to use DVDs, tapes are outdated and still not sufficient.
It seems the only option is hard disks.

But is a hard disk really a backup? I have had recently an incident. I started a USB-attached hard disk for backups by hand - copying files and folders. Two weeks ago I made mistake - copied about 250 GB of files but apparently clicked Move instead of Copy with the mouse. Few minutes later I decided I delete the files there, convert the backup drive to NTFS (as it would not accept big files) and copy again. I clicked Select All, Delete and ... all the files were gone from both places.

Now I see it seems stupid, and it is, but my point it is just too easy to delete the data n hard disk, and I learned it hard way.

What would you recommend? May be to access the backup drive only indirectly via backup software? Would it be enough really?
I wonder what backup do the web sites use, like SNB?

Valentin
________
Vaporizer Volcano

 

[thiggins]

Two weeks ago I made mistake - copied about 250 GB of files but apparently clicked Move instead of Copy with the mouse. Few minutes later I decided I delete the files there, convert the backup drive to NTFS (as it would not accept big files) and copy again. I clicked Select All, Delete and ... all the files were gone from both places.

You could have made the same mistake with any media and the result would have been the same. The weakness is the use of one backup.

Backup is a form of security and has many layers. What you do also depends on the time/$ you wish to spend vs. the value of what you are trying to preserve.

Most pros use a backup method that involves multiple backups on different physical media. You can choose daily, weekly, monthly, or any combination, but the trick is to have multiple physical media that are rotated.

One small business I'm familiar with uses two USB drives. Every night the working directory on their server is copied to a drive and the drive goes home with the owner, after is replaced with the other drive.

 

[Dennis Wood]

Tim is 100% on point with that post. What you described is absolutely the number one source of data loss in my experience...accidental deletion. In the enterprise, normally there is a rather large tape collection depending on how far back you need to go. In my past environment security was a major concern meaning retrieval of backups was possible going back 2 years. This is done with a daily, monthly and yearly backup tape rotation. At the time, there was a collection of something like 75 LTO tapes to do this, daily off-site data exchanges, and daily tape rotations/logging. In this environment every server room had at least 3 servers, each with the identical LTO drive.

What we're experimenting with for SOHO is a combination of remote replication, external drives, and workstation backups using Storagecraft. Storagecraft supports (and it works) a bare-metal restore of a workstation including restores to different hardware, and the addition of drivers after its restore DVD boots. Of all the options I tested, it's the only one that worked with both XP and Vista and also supported the loading of all of the various RAID drivers we use. In other words, your old workstation image can be restored to a box with completely different hardware set..normally a huge headache requiring re-installation of the OS, and if your machine doesn't have a floppy drive, a slip-stream modified OS disk with mass storage drivers integrated. Storagecraft was the only product that we tested that worked flawlessly in this regard. The strategy works something like this:

Part1: The QNAP remote replication (and I believe every major player offers similar options) takes care of replication on a daily basis, but does not archive anything. This is great, providing what you're sync'ing is OK. You can configure the option to delete or leave alone deleted files as part of your backup, which can take care of accidental deletions..they'd be sitting on the backup NAS. Data is replicated at whatever frequency you decide....likely every night. Rsync does delta binary backups so only the binary changes in a file are actually sent over the WAN. This makes it very efficient for backup up things like email, modified spreadsheets etc. Your first sync should be a local one, otherwise a full sync over the WAN may take weeks.

Part2: Storagecraft creates encrypted full workstation images, incremental backups and automatically manages archive sets so you can decide how far back you want to go. These backups are scheduled to the NAS and live there...although you'd likely not replicate these using rsync. However, if a workstation goes down, you could be back in business on a new box in 1 to 2 hours. Boot using the Vista pre-installation environment DVD, add in your RAID, or controller drivers from a USB drive, activate the network feature and restore right from the NAS at a rate of 50 to 80 MB/s (depending on hardware)

Part3: Enter the external eSATA drives. An eSATA enclosure and 1TB drive is all of about $160. We use several of these in rotation. The enclosures can be brought on site and used to backup things like the workstation images mentioned above. At 40MB/s, you can transfer 1TB of data in about 7 hours. This means that the external drive would never be left on-site providing the work day is over 7 hours. This is also why I've been pestering the QNAP folks to improve the write read/write performance of their products....if the unit is backing up during the day, then you still need acceptable performance for LAN users.

Using the three strategies, you can nest them as appropriate to get the file history and disaster recovery strategy that works for you. Our own strategy is that if Site A is utterly destroyed by fire, theft or otherwise, then Site B can be up and running fully functional in 8 hours or less, losing no more than 1 day's worth of data. This includes the time for one person to purchase, set up, and re-image 4 workstations . Total cost for the strategy, including 4 TB of working data storage, is about $3000 with minimal maintenance and zero storage maintenance costs. What's very important is that other than the external drive jockeying, the process is completely automated, and any problems are reported via email...including failure of the eSATA transfers.

There's some free advice for you

 

[edwaa]

...But if a power supply or controller board dies at the wrong time, then the whole array could be toast.

Could you elaborate on this a bit more? If the RAID1 drives are in the middle of a write operation (I assume this is worst case) when the power supply or controller board dies that means the entire array is ruined? I know very little about RAIDs but I do know I've lost power to my personal computer in the middle of a write operation and only the files that were being accessed were corrupted, not the entire drive.

 

[bigclaw]

Am I missing the reason for having the NAS and backup NAS set for RAID1 (RAID5)?

RAID1 offers the following benefits:
1) high availability of data. If one drive fails, the user's work is not interrupted; s/he is not even aware of the failure and can continue accessing the NAS as usual. The admin can come in later to replace the drive and resync the RAID array.
2) prevention of data loss (only) in the event of a single drive failure.

RAID1 does not prevent any other types of data loss, which arguably may be more common than a hard drive failure--virus, accidental deletion, file corruption resulting from power failure, etc.

So you can see RAID1 is not a complete backup solution at all. If you don't care about high availability of your data, you don't even need RAID1 as long as you have a backup plan.

 

[thiggins]

but I do know I've lost power to my personal computer in the middle of a write operation and only the files that were being accessed were corrupted, not the entire drive.

Consider yourself lucky. You dodged a bullet.

 

[germ]

RAID1 failures

Could you elaborate on this a bit more? If the RAID1 drives are in the middle of a write operation (I assume this is worst case) when the power supply or controller board dies that means the entire array is ruined? I know very little about RAIDs but I do know I've lost power to my personal computer in the middle of a write operation and only the files that were being accessed were corrupted, not the entire drive.

It is possible to have a catastrophic hardware failure that will affect both drives, for example if the power supply fails or there is a power surge (storms). So always also use a small UPS with your NAS if you care about your data. Its cost is a small fraction of the cost of the NAS, and it will enhance the reliability of your data.

Also, the file system on one drive could become corrupted, and in that case the second drive may also become corrupted.