School me on Windows software RAID

S

sap

Limp Gawd
Joined
Feb 3, 2006
Messages
372
I recently got the server in my sig up and running and configured. There is one 1TB drive holding my data and a separate 160GB drive for the OS. I'd like to move all of my stuff (music library and movies) onto the server and get rid of the laptop. As such, reliability concerns are beginning to be a factor in my thoughts and I would like to know more about using Windows software RAID if I bought a second 1TB drive.

So far, all I can really tell is that I need to convert the 1TB from a basic disk to a dynamic disk. What I have not found any information on is whether I can turn the partition on the basic disk into a mirrored volume without losing the data. I could always back it up first, then do the convert, then make the new volume, then restore... but I'm lazy and I was wondering if it could be done :p

Also, does anyone have any real experience using Windows to do mirroring? Any "gotchas" I should be aware of before deciding to do this? Do you think I would be better off to do a script that just copies over everything new each night to another drive?
 
The only 'gotcha' is that mirroring is inferior to home users. With the same overhead (double storage space required) you could have a full backup, while a RAID1 is a single volume without backup but double redundant; those two are not the same.

I would opt for two RAID0 arrays anyday rather than one RAID1.

In other words, why not focus on backup instead; put one HDD in another system that would be so much more reliable.
 
sub.mesa said:
The only 'gotcha' is that mirroring is inferior to home users. With the same overhead (double storage space required) you could have a full backup, while a RAID1 is a single volume without backup but double redundant; those two are not the same.

I would opt for two RAID0 arrays anyday rather than one RAID1.

In other words, why not focus on backup instead; put one HDD in another system that would be so much more reliable.
Click to expand...

Keep in mind this is not really for a "home" user. I am not concerned about being able to get data back if I accidentally delete it. I am concerned about keeping the server up, doing it's job.

Using a second drive as a backup would require manual intervention to change the drive letter on the backup to what the primary drive was so everything keeps going.

I don't understand why I would use two RAID 0 arrays, as now we're talking about putting in 4 drives when 2 will do - I don't need the speed, it's just a file server (these are the WD10EARS drives we talked about in the other thread).

And running two systems is off the table entirely - partly because of the manual intervention failing over to the backup would require, partly because I would like to keep power consumption down - one box is plenty.
 
@ OP, you're talking yourself in circles. One minute its "reliability concerns are beginning to be a factor in my thoughts" and then you say "I am not concerned about being able to get data back if I accidentally delete it".

What Sub suggested was to actually back up your data, not use raid. Raid isnt backup, and never will be. You very well could use a raid 1 mirror as a backup, but its not the best route. The disadvantages are the extra write overhead, it doesnt protect from a user deleted file ( you do it, you kid, your mom, a virus). Theres just too many things that even the simplest 2 drive array can screw up.

I also would suggest a modest box just for backups, and dont stress about the power bill, you only need to turn it on for a little while once a week or so and make an incremental backup. As far as the part about changing drive letters, all that being just too much work, well, thats up to you. I would think 5 minutes a week to cover 1Tb worth of data to not be too harsh. I would imagine re-ripping, re-downloading, re-torrenting etc 1 Tb of data to take a lot more time than that.

Also, you mentioned "I am concerned about keeping the server up, doing it's job." Well, a server with a failed array, or corrupt data, etc isnt gonna be worth a whole lot


If your dead set on using Windows raid, make the array, then copy data.
Unfortunately theres no quick, easy, cheap, right answer for your question.
 
I don't understand how reliability means backups. I don't want backups, I want the server to stay online. That's what I mean when I say reliability. Maybe I'm not using the right word. Or maybe I don't clearly understand what RAID 1 really means. I thought if I used RAID 1 (as a "mirrored" volume in Windows), that meant that one of the two drives in the array could fail completely and the files would still be available. Am I incorrect?
 
RAID is not a backup solution, it's never been designed for that purpose so yes, you're correct in your reasoning that RAID 1 specifically is a fallback method of reliability - if the primary drive fails, the system "falls back" immediately (at least that's the hope) on the secondary, the mirror volume, without missing a beat or a bit of info in the process.

I've done simplistic 160x2 software RAID 1 on Windows machines in the past for some situations and it never failed, never had issues, even in spite of people preaching that "software RAID is horrible in terms of performance." Hell, I ran two 80GB VelociRaptors (not Raptors - these are/were Dell OEM VelociRaptors, 80GB in size each) with Windows software "RAID 0" (since RAID 0 isn't really raid, more appropriately JABOD than anything) and got ridiculous performance just a percentage point or two below actual hardware RAID speeds, and the CPU usage was negligible on the given platform I was using.

If you have a 1TB and get another, setting up RAID 1 using Windows software-based RAID is perfectly acceptable and you shouldn't have any concerns, honestly.
 
software raid is fine for raid 1 or 0 because there is almost 0 overhead it has to do. Now raid 5, raid 10, raid 60 and so on, usually sucks when not using a dedicated raid card because of all the work it has to do, parity bits and so on.

Raid 1 will work for the needs you want, and why someone would suggest raid 0 over raid 1 is beyond me, lets go to 4 points of failure vs 1.
 
If you only want uptime protection, yes RAID1 does that for you. Don't consider your files to be safe though, but the system would continue running with one drive drop-out. If that is what you want, RAID1 is your solution.

As for RAID0 being superior to RAID1; no you look at it wrong. Two RAID0 arrays in different systems that store the same data is a backup. And RAID1 isn't. Two RAID0 arrays in different systems would protect that data against more dangers than RAID1 is capable of. RAID0 is an excellent RAID level to use if it makes sense for your application; for example when you have solid backups.

RAID0 also has the lowest chance of anything going wrong, like RAID1, as it generally works with a very simple RAID driver; unlike RAID5/6 which require a lot of complexity. The RAID driver itself adds another layer that can fail; so using RAID can both enhance and reduce reliability/data security and uptime protection.
 
sure 2 raid 0 in 2 different systems.... but why bother, sure you get the speed, but you also need twice the harddrives, twice the computers for your method.

Raid 0 has more to go wrong then raid 1, like 1 single harddrive dying and you loose everything, as you said, it reduces it really compared to raid 1

Raid anything is NOT a backup it is redundancy, even one 2 separate systems.

Why not do 2 raid 1 arrays and get even more protection then raid 0 can ever give you.
 
Mirroring or RAID1 would also need twice the harddrive to gain the same storage speed. RAID0 is simply a technique of getting more storage space (and 'free' performance gain) without applying any redundancy. So it is not RAID0 which makes the storage less safe; but using multiple disks. You could define a single disk as a RAID0 array with one disk member.

Raid 0 has more to go wrong then raid 1, like 1 single harddrive dying and you loose everything, as you said
Click to expand...
You think about drive failures. Yes, true, though two drive failues in two separate systems within 24 hours are extremely rare. But not so rare if the disks are in the same system.

But other than drive failures; please consider the failure of the RAID layer as well. Most Windows (Fake)RAIDs that caused data-loss lost data not by failed drives, but due to failures in the RAID layer itself; or the user-interaction with the RAID layer. Especially RAID5s with Silicon Image, AMD/ATi, nVidia and windows software RAID5 would have a high risk of experiencing data-loss as a result of failures in the RAID drivers and the user interaction that follows.

Example: It is not uncommon for a user to kill all his data by removing a split RAID5 array and re-creating it. Only this time, the disk order changes and a subsequent rebuild would kill valid data on the volume. So data that WAS recoverable but just temporarily inaccessible, became unrecoverable by the user interaction that followed an initial (minor) failure.

If you think it is because of the "shitRAID" - think again: also Areca hardware-RAID would experience split arrays; at least my 1.43 firmware Areca does. So by all means consider the RAID layer as something that can fail, too. And if you don't have a backup and rely solely on a low-quality RAID implementation, common on Windows, then you would have a high risk of actually losing data; which is what we were trying to prevent by using RAID; wasn't it?

So i say; do not rely on one thing - spread your risks! Two arrays still means you rely on one driver; one piece of thing that can fail anything that is 'underneath' that layer. A computer is a layer as well; storing your data as a backup in another computer is a better backup by many degrees.

Imagine a home situation like this: most households have multiple computers in home. In such a situation, a very simple and effective solution would be for both computers to be equipped with high-capacity disks. The PC in the living room would have the main RAID0 array, and the PC in the bedroom, which is usually offline, would store the backup which is done weekly or automatically when the bedroom pc is started up.

RAID is a tool, you can use it to your advantage in the right conditions. RAID is not a goal; redundancy is not a goal; data security or uptime protection, is. Data security would mean that, regardless of the required user action to perform, all stored data would keep retrievable, without any corruption originating from the storage (RAID) layer.

As proof of my conviction: you guys know how much i love ZFS, and FreeBSD. Still; i opted to build an entire clone of my main fileserver as REAL backup; only without the extra SSDs. I wouldn't want to run without a true backup. Though i can understand that, for some people, the protection of a single good RAID engine (like ZFS) could be enough for their needs and fit inside their tight budget. But with disks of 2TB/3TB capacity nowadays; why not spend a little bit extra for a real backup? You could use RAID0 without significant risk and generally it simplifies your storage setup if you have a real backup.
 
To answer your question, once you have a dynamic disk, and a blank disk in the system (with no partiton on it) then you just go to disk manager and mirror the partition. It will not erase any data and will create a software RAID1.

however if this is the boot drive, and the primary drive fails the copy will not be bootable.. It can be made bootable should the primary drive fail, but that requires editing the boot.ini file.
 
Lazn_Work said:
To answer your question, once you have a dynamic disk, and a blank disk in the system (with no partiton on it) then you just go to disk manager and mirror the partition. It will not erase any data and will create a software RAID1.

however if this is the boot drive, and the primary drive fails the copy will not be bootable.. It can be made bootable should the primary drive fail, but that requires editing the boot.ini file.
Click to expand...

Awesome, thanks for taking the time to post. I figured someone must have used this before :cool:
 
Just to be clear, right click on the partition with the data and choose "add mirror" you have to have empty space available for this option to be available.
 
The major 'gotcha' with Windows RAID1 is if you are running Hyper-V. You need to manually tweak the boot settings to enable Hyper-V if you boot off the second drive ("BCDEdit /set hypervisorlaunchtype auto") .

It's little things like this which show Hyper-V is still not a version 3 Microsoft product yet.
 
Lazn_Work said:
Just to be clear, right click on the partition with the data and choose "add mirror" you have to have empty space available for this option to be available.
Click to expand...

Noted. I will head to MicroCenter tomorrow to pick up the new drive and probably try it out on Sunday. Thanks again!
 
Just wanted to follow up on this in case anybody is looking for answers on how to do it... The built-in help doesn't say anything about how to mirror.

1) Install the new disk
2) Click the existing volume on a dynamic disk
3) Choose add mirror and select the new disk
4) Wait for Windows to sync the drives

Thanks again Lazn_Work!
 
You must log in or register to reply here.
Back
Top