Jump to content
RESET Forums (homeservershow.com)

WHS vs. raid; 10+ disk systems - Theory


stalni
 Share

Recommended Posts

I was thinking after my system drive failed, since I have 10 drives in my WHS and have seen/read post of 20 or more drives, using WHS. What type of backup/drive redundancy WHS actually offers. My thinking is that if I have more than 10 drives in a computer/server a redundancy system that provides more then a single disk failure policy would be wise to implement.

My thinking is that a RAID 1 array provides failure of 1 disk with 100% certainty of 0% data loss, WHS offers the same, but at 2 drives WHS provides a more random recovery process(you are still not sure that you will lose data, but could expect some data loss) than RAID 1 were if you were so unlucky as to loose the to drives that have been duplicated one would loose all data on that drive or more depending on the raid array policy(amd sb710 I know stores data in a readable format in a raid 1 array, I do however not know if it is readable for windows in ahci or ide mode on a sb 710/750 system).

This made me think that a RAID 6 or multiple RAID 5 arrays sounds like a reasonable/smart idea to implement on a 10+ drive system.

I can read from the 52 drive build that he was planning on 24 drives in RAID 6 x 2 or 16 drives in RAID 5 x 3 + 2 spare, I don't know why he thinks a single drive failure policy for 16 drives is good or if a 2 drive failure policy for 24 drives offers lower likelihood of array failure.

As I have very little experience in this field(age 20), what are your thoughts?
If you have any statistics that would be great, have read googles paper, that there was linked to on another post her in HSS forums.

I cannot remember googles specific findings though, apart from that the drives fail within 6 months or last 4-5 years, and that keeping drives below 45 degress celcius does not extend drive life, I guess this would be the most relevant data to make such conclusions, as to say you need a 2+ drive redundancy policy for a 10+ drive array.

Link to comment
Share on other sites

Hi Stalni, you have some good points, and I like discussing the big OEM WHS systems, that's where the fun is at! One thing that I differentiate between in discussing storage are the system drives vs data drives. The only reason I would recommend a raid 1 for WHSv1 is that there isn't an easy way to re-image/restore the system drive with the whole toombstone scenario. In WHSv2, I'd be fine with running a single disk system drive if all you're concerned with is data protection and not high availability/uptime because they provide the means to do easy system drive backups and restores.

The other thing to consider is if this is your single storage location for your data. So, if you don't have any backups of your data outside of your WHS, then protecting against a 2 drive failure may be more important than if you have your data on a backup server which could be accessed if you did have a 2 drive failure on a typical WHS system.

As always, the $$ is the big factor with performance coming in 2nd. If you're going to run raid 5 or 6, hardware cards are best...but they cost a pretty penny. You could go software raid, but that sucks for performance and I don't believe it works so well w/ WHSv1. Anyway, you're on the right track reading over at the Data Storage Systems sub-forum on HardForum, great stuff over there.

If you want to get more interesting, you can start adding some storage virtualization to the mix which can add some good features by separating the physical storage from the actual OS's running on top. Say create a big 20TB raid 6 array and carve out 10 2tb volumes to add to WHS. That is one interesting way to deal with it.

Link to comment
Share on other sites

wodysweb, would you like to virtualize just for the sake(having flexibility to add new systems), and then have multiple smaller vurtual disk(files) in case of file corruption? My thinking is that the raid array would provide up time, but doing a double raided system(raid 6 means loosing 4 TB in you array + up to 50 % with WHS), sounds like one of the most expensive ways to set the system up
and thanks for the response.

I was thinking more in the line of what would be a good drive count to drive failure ratio, for personal data(data I do not want to lose but have no financial loss if I do loose).

Maybe the more relevant question would be what you could expect from different policies. Meaning what failure rate could be expected from a 5,7,9,11,13 and so on disk array.

and what policies do you deploy.

Of course the whole reason for not doing cloud or offsite backup is $$ or in my case dkk.

I do currently have a offsite backup option with my brother living 10 miles away, but that is limited, as he is running a WHS as well, and I have a substantially higher data capacity on my server. but we both have good data connections 40/40 Mb/s for me and 100/100 Mb/s for him, and than of course the option to just spend the 20-40 minutes it takes to travel from his place to mine or vice versa.

Link to comment
Share on other sites

Regarding dual drive failures, the chance does increase with array size. Another thing that increases the risk is letting an array run degraded for a long time (this happens very often on consumer RAID setups which have no spares). The risk is also higher for some types of RAID which place a lot of stress on the array during rebuild, like RAID5/6.

Regarding details of different RAID types, although RAID1 can sometimes be readable when directly connected, depending on the implementation, the same cannot be said of RAID10. Essentially that means that if you were to lose the two drives in a set, you would lose the whole array, just like if you lost a drive in a RAID0. There are other ways like triple redundant RAID10 which can take even two failures in a set and is a true dual redundant system.

RAID5/6 will require hardware. It is possible to implement these in software, and they will operate reasonably well during regular operation. During a failure however, the rebuild cycle will bring the system to its knees and will last a very long time, increasing the risk of dual or even triple failure. If you do want a RAID6 solution, you will want hardware and it will be expensive.

Bear in mind that RAID solutions do not offer any easy way to expand. There are two ways to upgrade a RAID array. One is to backup to a temporary storage (big enough to hold all data), then build the new array and restore from backup. The other way is if you have a system that is large enough to host the old array alongside the new one, you can set it up that way and directly migrate across. Either case requires significantly more hardware than what you would actually run with. WHS of course just requires you to add a disk.

Link to comment
Share on other sites

Guest no-control

Not to point out the obvious, as I realize this is an exercise in theory as opposed to actual application. But RAID is primarily for either performance or uptime not data backup. Not to mention most people do not have a need for 10tb of storage.

IMO the reason for not using raid for anything other than a system drive is that WHS is a bit more flexible and granular that what a basic RAID solution provide. Not all data needs a backup, but lets suppose it does. In my experiance I would prefer to have a backup off site. If we're talking pure theory and logitics and budget isn't an issue. Simple 1:1 in two locations would be the ideal. My RAID 10 array was waaaay more unstable than my previous RAID 5, array which was waaaay less flexible and predicatable and cost efficient than my WHS when measured by cost to own per GB stored/minute.

When adding multiple layers WHS over RAID5 or worse WHS over RAID 1 of a RAID 0 (RAID10) your also adding multiple points of failure.

Virtualization works well because the system drive isn't physical per-se It's a virtual image that you can easily backup. It adds a layer of software and flexiblity, which could be perceived as better than a layer of hardware. Treating your system drive as just another file that gets backedup with your data is powerful stuff.

Link to comment
Share on other sites

no-control, what does uptime imply, as far as I know you need backup to provide uptime. I do realize that RAID i a very expensive solution, and like all backup solutions not perfect, but when you reach a certain array size you would loose more data to the WHS method and be at an increased risk of loosing that data. The offsite backup should provide a offline restore option i you are so unlucky as to loose 2 non system drives on WHS.

The whole reason for my pondering as to failure/recovery options and RAID, is that I feel somewhat weary of expanding my WHS beyond 10 drives, simply because of the backup solution provide. I consider my duplicated data as secure, and legal stuff I do offsite backup to my brother, which should provide 3 physical drive data locations(duplication + offsite).

And than there of course diehards solution that uses the shares backup option in WHS, that does however not provide auto rebuild, like the WHS duplication option.

Link to comment
Share on other sites

Guest no-control

Yes Technically you are correct the RAID provides uptime via a Backup. But the intent isn't to provide backup the intent is to provide continuous access. I think the intent is what is key here. There are much cheaper, simpler solutions for backup. RAID is expensive because of the intent to keep all data accessible 24/7, which we both know makes sense in a business environment. If my wife needs to wait a day to access wedding photos or grey's anatomy so be it. Althoguh RAID may save me a headache in that situation ;P

I hear ya I have 10TB of storage right now and I only have maybe 3TB in duplication as I've been reconsidering what the real impact of catatrophic failure is. I no longer back up photos, home videos, or the music catalog. These don't change often enough and I have 2 backups of these items offsite + cloud. I have never considered backing up my movie/tv catalog. I have the backups in the form the original disks. But it would be some serious heartburn to have to re-rip it all. So the only daya that gets the backup treatment is my User/Public/Software shares User holds alot of my business and hobby info, public just because its the temp dumping ground for the family, and software mainly to preserve my PC Tools and Technet files.

I have been considering using diehard's backup too, but never got around to it...I can be lazy sometimes ;P

If you do decide to go down this road anything more than RAID5 is going to be expensive.

Link to comment
Share on other sites

no-control, thanks for your answer. Have you thought about the disks(bluray/DVD) may become unreadable simply because of storage degradation. I guess it would probably be cheaper just do repurchase the movie than provide duplication for your videos.

The whole RAID 5 is expensive thing, makes me think that 2 x 10 drive RAID 5 array, would be just as secure as a 1x 20 drive RAID 6 array?

And I guess being a student I do not or cannot always get/make 2 hours spare time to use on a server error, I guess with a good backup option(more $$), I could still get access to the files, within a day.

Link to comment
Share on other sites

@stalni: Just for reference, the recommended maximum drive count for RAID5 is considered to be 8. Also, it is a good idea to have at least one hot spare per array.

If you have not read this yet, it might be helpful to you.

Link to comment
Share on other sites

will take a look, have only read parts of that article.

thanks roddy

and would the recommended array size than be 16 drives for RAID 6 + 2 spare?

and again, would it not be more interesting to use a RAID 5E or 5EE, if you were to spend that type of money.

Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
 Share

×
×
  • Create New...