By Philip Churchill on August 11th, 2008
Windows Home Server Drive Extender v RAID
Over at the Fear The Cowboy website, Garrett Serack explains why Windows Home Server Drive Extender technology is better than RAID5, with his own personal experiences thrown in. Garrett follows this with some great advice for Home Server users.
Read it here.
Share this WHS Article with Others:
I use 2 Hardware RAID 5 arrays with WHS. There are 2 reasons for this,
1. The ratio of space lost to redundancy compared to usable space is better than with Drive extender, in which case you would only have 50% usable space if you duplicated all your files.
2. At this point I just don’t trust Windows reinstall process if the System drive fails, what happens to all those backups and files?
Just remember RAID is not backup. Using two arrays is probably a good idea, but more advanced than most home users.
Take a look sometime at the Advanced Disk Management Add-in, and see where WHS stores files. You’ll see that hardly any gets stored on the physical drive with the system partition.
After watching my system for over a year, I’m becoming more confident that if my system drive fails, I’ll be able to recover all my files. I may have to rebuild my system drive (that can be backed up using nt tools, but isn’t automated yet) but my data is safer.
“Garrett Serack explains why Windows Home Server Drive Extender technology is better than RAID5”
No, actually he didn’t. What he explained were scenarios why RAID 5 is not bullet-proof. He did not eplain why DE is any better, other than some vague implication that DE is simply hetereogeneous RAID 1 (which is what it seems to be). RAID 1 suffers from the same issues in losing 2 drives that he highlights regarding RAID 5. In fact, with the right tools, it is easier to reconstruct the data in a RAID 5 array that has lost two drives than it is to reconstruct the data in a RAID 1 array that has had both drives go.
If you want to protect against multiple drive failures, then you start looking at building in additional redundancy into the array: RAID 6, RAID 5 + hot spare, etc. Until MSFT can explain which trade-offs DE is making in terms of space consumption vs. redundancy, the idea that DE is “better” than RAID is a canard; I expect there is simply re-branding of old concepts going on.
I agree with Bill. I already know somebody who has had his WHS system hard disk fail on him, rendering the supposedly amazing drive extender technology useless. With some technical assistance from me, we have recovered all his files from the remaining drives, but I would hardly call it better than raid. It seems it only protects against failure of a data disk, and not the OS disk. This seems to be as much of an obstacle to your average user as anything else.
MS Hype
Actually, from what I’ve read, DE sounds more like JBOD with selective mirroring. The disks can be of variable size, and even variable type (SATA, IDE USB, Firewire), and DE will treat them all as one big “storage pool”. Also, with the exception of the system drive, they can be added or removed without much issue.
The mirroring is only on user selected files/folders, so space can be saved compared to a RAID 5 by not mirroring certain data. WHS doesn’t mirror PC backups, for example (with the reasoning that the backup is already a mirror of the PC drive).
If the system drive fails, the other drives can be plugged into a PC and accessed like regular drives, since DE is a file-system level storage handler, so the individual drives are formatted like regular NTFS volumes, and files are saved normally, with DE creating an associated “tombstone” file to keep track of it.
There’s still the same problems with a two-drive failure, in that the data that was both stored and mirrored on those drives will be lost. Also, the loss of the single system drive is worse than losing 1 RAID 5 drive, since WHS no longer works, and would need to be reinstalled, while a RAID 5 system would continue working while the failed drive is replace
An interesting idea would be to use a RAID 1 for the “system drive” in WHS, and let DE manage all the other drives. That way, if one of the drives in the RAID fails, the system would still be operational, while DE would allow using variable size/type drives for the rest of the storage pool and selective mirroring.
I’ve been reading a lot of articles about raid 5 and WHS over the last couple of days and both camps seem to make good points about their champion. What I am most curious about is why Microsoft felt it necessary to build functionality into WHS that makes capabilities native to Server 2003 unusable. I agree that most home uses will be adequately serverd by drive extender, however, if I want to use a raid 1 volume for the system drive and raid 5 volumes as managed drives why should this be actively prevented. I am also discouraged that there is no way to create/move the “D” drive on a different physical drive than “C” – as all data written to the WHS first lands on D: and is then “balanced” out to the other managed drives this really hammers the system drive which IMHO will make the failure of the system drive more likely, to say nothing about the performance impact continually moving the heads from swap file to “D” drive for every disk write.
The ironic thing is that the various blogs and articles focus on the possibility of failure of hardware and the need for backup being more important than redundancy, but by disabling these capabilities WHS denies the user the option to beef up the hardware and make it more fault tolerant.
Raid 50 is supported on many motherboards these days and will allow for continuous operations with the failure of any two drives. Hot spares are also supported by many raid chipsets and the latest twist (RAID 6) can keep on trucking with two drives down.
I guess my point is, that allowing WHS to install on and manage RAID volumes as well as on individual IDE, SCSI or SATA drives does not take away any of the advantages of DE (use of disparate drive technologies and sizes), it would just permit additional (optional) levels of redunancy and fault tolerance.
Yes RAID is not a backup, but what exactly IS a backup?
Backups are NOT designed for a drive failure recovery where as RAID is. Backups are designed for a file by file basis recovery. Typically in accidental deletion, file corruption, or for revision history.
RAID is for disaster recovery and lets be honest, when you say drive failure we arn’t talking about individual files, we’re talking about a whole drive.
A good IT admin out there knows that you need to use both RAID and Backups.
In this case the guy at microsoft sounds very incompetent regarding the two roles of Backups/RAIDS and why both are needed/used in tandem.
Any IT admin who has tried to recover a lost drive from a regular backup has probably already been burned. The most expensive Backup Exec version was not meant for full system recovery. It’s designed to backup “**FILES**”, not drives. Meanwhile a RAID is more thorough and safe in the recovery process. However RAID cannot recover individual files or keep a revision history. You see now why both need to be used together?
His article makes the assumption that
A) Windows Home Server needs gobs of performance. Lets be realistic, we’re talking about a home file server here. How much performance do you really need? The leading Home Server appliances are running atom processors for *(@#$#@ sake.
B) So you went with raid, does that mean you stop backing up? NO! He makes it sound like if you have a RAID your all the sudden incapable of backing up. No you continue to backup.
Finally his argument that RAID 5 when broken is dreadfully slow cannot be held up in an actual lab. It may be a little bit slower on a software raid but it’s not dreadfully slow as he argues. Also, he argues that one could be in a situation where the drive could not be replaced. What situation is that? You can easily have a drive on spare or have one overnighted.
His whole post is completely ridiculous. Further proof on the types of knowledge that comes out of Microsoft. I’m glad I turned down that job offer.
Now for Windows Home Server. Whether you need RAID 5 or simple Backups thats entirely up to you and for what content your putting on your home server. If your ONLY backuping workstations to your home server it’s silly to do file backups of backups. It’s like backing up the backup.
Home Server is based on server 2003 and can run RAID 5 over software easily.
http://support.microsoft.com/kb/323434
Or you may actually use some of the cool features of Home Server and store various types of media. In this case you may want to be more specific with what gets backed up and what doesn’t get backed up. You may decide that you don’t want to deal with the complexity or expennse of raid 5.
At any rate the path you choose is up to you. In my case RAID 5 isn’t needed and the built in backup system is adequate.
I read that so called comparison by Garrett.
Well…, there is absolutely NO explanation why DE is better.
He lost his RAID-5, well, boohoo.. 🙂
Get a HOT Spare and go to RAID-6.
However that is not the answer for the lack of RAID support in WHS.
Why is it restricted from people who have capability and knowledge of running it is beyond me.
Ideal backup situation would be to to allow RAID-5(6) (over 2TB) and allow user to manually set folder duplication. That way I can add a few more junk drives and use them for a secondary copy of important data.
Well I can tell you that i had a Raid 5 setup 2TB total a while back and 2 drives failed and i lost all was lost. I knew though that this was a possibility so had a 1TB external drive that i used to backup all the data i though was most valuable. Music,pics, user created content,etc. All the other stuff like movies, tv show, applications, etc i could care less about. So in the end I wasnt too peeved because i know all i had to do was rma the drives, reconfigure my raid 5 and restore my data. Now im in the process of building a raid 6 8TB box and im debating how I should configure this thing. So ive been searching for all kinds of things whs related and thats how i came to this article. Good read becuase it was based on personal experience, but in the end no matter how reliable you think your drives are in a Raid x config you should always have another copy of the data you value most. I mean honestly storage is so cheap these days that there is no excuse. Buy a drive every two months or so and just have them lying around i say. Peace
beo,
It’s very easy to hide behind your keyboard and try and tarnish someone’s professional reputation. You state “In this case the guy at microsoft sounds very incompetent regarding the two roles of Backups/RAIDS and why both are needed/used in tandem.” I wonder if you’d be like that face to face? The OP’s post was quite useful.
If you stop thinking like a techie, and start thinking from a home user’s point of view, you’d see that drive replication is a far better way to ensure one’s abillity to get back up and running after a drive failure.
Go back to your ivory tower.
I’m not a professional by any means when it comes to raid but this Garrett character is talking out his @ss.
Quote:
“Software RAID5 is slow. Damn Slow. Faster than that… maybe pretty damn slow. Not a great solution. You won’t be happy at the end of the day”
Maybe on some ancient piece of hardware…. Any reasonably recent dual core system will run software raid 5 no problem. I’m running it on my system as I speak and it’s not slow by any means…
Quote:
“Here’s the kicker. What happens when a drive fails, and you are not there? If the system is in use, it’s going to get really really busy, and all of the drives in the array are going to get a lot of use.”
Where talking about a home server here, not some data center….
Quote:
“When that hard drive fails (and you are not planning for IF but WHEN), and the others pick up the slack, the chances of losing a second drive go thru the roof. What will you lose if a second drive goes?”
The chances of losing a second drive go through the roof? I would say he’s exaggerating just a little bit there… Now a days chances are if your drive doesn’t fail immediately it’s going to last you a very long time! I’m talking about a plain old disk failure, not about flood, fire, tornado, etc…. If your system gets whacked by lightning or some other kind of electrical surge it’s not going to matter if you had a raid array or if your using this awesome new MS tech… lol
Garrett is not wrong, but he is abbreviating a bit. The chances of losing a second drive, under the duress of a reconstruction, the way most home users (including me, by the way, for practical reasons) buy and build arrays, is very high.
Home users tend to buy the same drives, from the same lots all at once.
I had a two drive failure on separate arrays in the same machine at the same time. I’m actually inclined to believe it was a power supply spike, or a defective board. Don’t know, don’t care. Got lucky that there were 2 4xdrive arrays.
What Garrett really misses is that I am building a once-a-day backup server. I DO NOT WANT the drive space lost by doing DE. My WHS __IS__ the backup mechanism and nothing more; it can be unbelievably slow, and I don’t care.
I’ll pile on there; I wasn’t particularly impressed with that “why DE is better” post.
My solution is that I run hardware RAID-10: mirrored and striped. That way I get potentially TWO drive failures before I lose data. I have two arrays of (4-disk RAID 10) with 1TB drives, on a TowerRaid TR8M-B.
I let WHS do it’s DE fun if it wants using the two arrays, but at least I know there’s going to have to be a lot of failure goin’ on before I lose something.
Oh…and this WHS’s sole role is to backup my 4 x 1TB RAID-10 arrays that live inside the client machines!