To RAID Or Not To RAID

Discussion in 'General Hardware' started by ThePermster, Oct 13, 2009.

  1. ThePermster

    ThePermster OSNN One Post Wonder

    Messages:
    4
    So I'm an absolute newcomer to RAID and I think I may have ended up down the wrong path in pursuing it in the first place. I've been looking into RAID 5 for my home desktop and I've come up against a multitude of conflicting information.
    On the one hand RAID 5 sounds awesome as I can stick my three 1tb HDDs together into one big 2tb HDD with added performance increase and fully recoverable data should one HDD fail but on the other hand I hear that actually RAID 5 makes a detriment to performance, that chances are it'll make 2 HDDs fail at once somehow and backing up using more manual methods would be better for me or I've even heard that RAID isn't a backup solution, that I don't really understand as as far as I understand it RAID 5 should be a simple case of replacing single failed HDDs to have everything working fine again, right? Is RAID the solution for me in the first place?
    Not that I really get as far as considering replacing failed HDDs as I don't even really know where to start with setting up RAID.

    Essentially I'm a bit lost as to what to do so I'll just give you the vision:
    The idea is that I would like a hard-drive failure to be a non-issue. That I can multi-boot in either Ubuntu or Windows (from a HDD that wouldn't be part of any array) and have a vast amount of storage space available (I'm not too fussed about performance, as long as I'm not waiting a day to copy 100k or sumpthin' rediculous) and should a HDD fail it's a simple case of buying a new one, plugging it in and having everything just as it was before.

    Any ideas?

    Also two impending factors for me is firstly: money, I'm a student so I don't have much of it, I've seen various references to dedicated RAID controller hardware but they always seem to go for £200+ and that's just not feasible by my budget. Would it be a crazy choice for me to control RAID through software (bearing in mind that I multi-boot) or through my motherboard's inbuilt RAID controller?
    And secondly: availability of hardware, due to a number of factors, lack of space in my desktop tower, again lack of money to buy more and suchlike I'm not really into the idea of buying more hardware for the sake of nothing other than keeping the current hardware in check. Meaning I don't much like the idea of buying two HDDs at a time so one can be dedicated to backing up the other, this is why RAID 5 appeals to me as for every 2 usable HDDs you only have 1 HDD dedicated to keeping everything safe and recoverable.

    Any input on this would be really appreciated, thanks.
     
  2. fitz

    fitz Just Floating Along Staff Member Political User Folding Team

    Messages:
    4,076
    Location:
    Chicagoland
    Phew.. where to begin..

    Well, both are true and untrue. Let's start with performance.. RAID5 will generally give a boost to read performance because it's able to read data from multiple drives at once. Write performance is where RAID5 becomes questionable since there is an extra write for each (n-1) blocks to write the parity information. For example: for your 3 drive array, two blocks of data, one block of parity on the first stripe, one block of data, two block of parity, one block of data on the 2nd stripe, one block of parity, two blocks of data on the 3rd stripe (rinse and repeat). So, more I/O required for writing a set amount of data. A good quality RAID controller with a large cache (most likely not an onboard RAID controller) will generally mitigate any of those performance issues so long as all disks in the array are working. When a drive fails in a RAID5 array, performance in general takes a large nose dive - even after replacing the failed drive, while the array rebuilds, performance suffers big time.

    now let's talk about failures.. as drives get bigger, RAID5 rebuilds take a LOT longer - that's a LOT longer for another drive to fail before the array is rebuilt. And here is where the question of drive reliability gets factored in. A little known fact is that SATA drives in particular often get bad blocks and the drive firmware will automatically remap blocks of data and never report errors on the drive.. while it doesn't "make" other drives in the array fail, generally, when a drive fails it is because enough of the errors cropped up in the array and there is a fairly high likelyhood that other drives in the array have errors as well. The RAID doesn't make the drives fail, drives fail because that's their nature. While I've seen cascading failures of arrays when one drive fails, they are generally fairly rare.

    RAID is not a backup solution.

    Let me repeat that.. RAID is not a backup solution.

    RAID provides for redundant data.. any and all data..
    RAID guards against one kind of hardware failure. There's lots of failure modes that it doesn't guard against.

    * File corruption
    * Human error (deleting files by mistake)
    * Catastrophic damage (someone dumps water onto the server)
    * Virus/Trojan/Worm
    * Software bugs that wipe out data
    * etc..

    Allow me to share the story of journalspace.com who relied on a RAID mirror set to "backup" their database.
    So, let me repeat: RAID is not a backup

    Given your requirements:
    1) Performance not that big an issue
    2) Using as little money as possible
    3) Provide for data redundancy in case of drive failure
    4) Allow for drive use with multiple operating systems

    I would actually recommend a RAID1 mirror using the onboard controller. Yes, you only use 50% of the RAW capacity of the drives but drives are bigger and a couple 1TB drives is fairly inexpensive. Software RAID option get thrown out because of requirement 4 and a dedicated expansion RAID controller is out because of req. 2.
     
  3. Johnny

    Johnny .. Commodore .. Political User

    Messages:
    5,015
    Location:
    Happy Valley
    Fitz pretty much cleared it up. The thing I would like to suggest is this. If you do do it and use another source of back up, then I would go with raid 0 for the performance - Remember that it is not true raid as it is not reduntant, though. What most people do is they use one seperate drive for the OS, and the raid drives for storage or to put games on. It is all based on your own needs and wants.

    You can get more info here -> RAID 0+1
    You can also go here -> http://www.raid.com/04_01_0_1.html
     
  4. fitz

    fitz Just Floating Along Staff Member Political User Folding Team

    Messages:
    4,076
    Location:
    Chicagoland
    Personally, I try to steer people away from RAID0.

    The two main reasons people go RAID0 is speed and space. My personal opinion is the performance gain for 95% of the people isn't worth the risk you take of losing all data across the array. Unless you are talking about a 4+ drive RAID0 array, performance won't be improved by any significant margin.. but that increases the probability of array failure and data loss.If space is the issue, then why not just use drives as individual drives and when (note, i say *WHEN*) a drive fails, you only need to worry about that data on the single drive instead of the data across all the drives in the array.

    Beyond that, one of his requirements was redundancy which RAID0 does not provide.

    Why do you link to a RAID0+1 link?That's a whole different ball game requiring a minimum of 4 drives.. beyond which, if you're going to 4 drives, I would push RAID 10 over RAID0+1
     
  5. LordOfLA

    LordOfLA Godlike!

    Messages:
    7,027
    Location:
    Maidenhead, Berkshire, UK
    I have all my important data backed up and stored in several places. Optical media, online backup service (dropbox), remote webserver or two. Chances of me loosing that data are very slim...
     
  6. ThePermster

    ThePermster OSNN One Post Wonder

    Messages:
    4
    Thanks for all the input guys, real useful.
    I get what is meant by RAID not being a backup now too but yeah that's essentially what I was expecting from RAID, nothing other than a contingency plan for the occasional and inevitable HDD failure and as fitz maintains that the possibility of a cascading HDD failure is possible but still unlikely then I think I might take my chances, if this was a server and not a desktop then the risk would probably be an important factor but as it's not then I think it's reasonable to implement this contingency plan and hope that the worst case scenario doesn't happen to occur.

    LordOfLA: whilst your solution does make the most sense as even in my vision's case I would still be waiting on the arrival of a replacement HDD but it just doesn't seem neat enough for me. And there is also the factor that those backup procedures have to be periodically maintained via reuploading, reburning, etc. I'm not sure it eventually evaluates to more effort or more hassle, it just doesn't seem as tidy as unplug bad HDD, plug in good HDD. Maybe that's a bad idea on my part to think of it that way though. Do you think I should just suck it up and start uploading?

    To the other recommendations it seems that my need for cost effectiveness has ended up pushing me to not be efficient with HDDs what with the RAID1 recommendation 'n'all which might not sound as nice to me what with the 50% versus 66% space efficiency but I guess it's all about results in the end.
    And I already have 3 1tb HDDs so I think what I might do is just buy another one and go for this RAID 10 thing as it sounds pretty fancy. As I understand it that involves having two pairs of HDDs set in RAID 1 which you then combine together into one big RAID 0, correct? Or do I have it the wrong way round? Is that still feasible with my on board controller?
     
  7. LordOfLA

    LordOfLA Godlike!

    Messages:
    7,027
    Location:
    Maidenhead, Berkshire, UK
    Just suck it up and start uploading and burning disks.

    You could also get a couple external usb/firewire disks as well that you only plugin to do periodic backups.
     
  8. fitz

    fitz Just Floating Along Staff Member Political User Folding Team

    Messages:
    4,076
    Location:
    Chicagoland
    If you are going to do RAID10, be aware that most on-board controllers won't do RAID10 but RAID0+1 really isn't quite the same thing.

    If you're going down the road of trying to setup RAID10, you probably are looking at needing to buy a quality 3rd party controller.. not sure if you want to go with that extra cost.
     
  9. Perris Calderon

    Perris Calderon Moderator Staff Member Political User

    Messages:
    12,332
    Location:
    new york
    this is a terrific thread, great questions and answers, fitz, you were great

    this might answer some questions around the web and if you guys think this should get a sticky go ahead and do it

    one thing

    xp also does that with the nt file system, haven't been keeping up with os theory but I can't imagine vista or 7 will do anything less

    but there's a great idea for an app, one that recognizes remapping bad secotrs and reports with a pop up
     
  10. LordOfLA

    LordOfLA Godlike!

    Messages:
    7,027
    Location:
    Maidenhead, Berkshire, UK
    SMART does that when it runs out of available space to remap :)
     
  11. fitz

    fitz Just Floating Along Staff Member Political User Folding Team

    Messages:
    4,076
    Location:
    Chicagoland
  12. Perris Calderon

    Perris Calderon Moderator Staff Member Political User

    Messages:
    12,332
    Location:
    new york
    so, anyone here think they can make an app that informs you when a bad sector is being remapped?

    you'd become famous, that would be an app every system admin would install so they can monitor drive health, most of us techies would use it too
     
  13. LordOfLA

    LordOfLA Godlike!

    Messages:
    7,027
    Location:
    Maidenhead, Berkshire, UK
    SATA disks are always error correcting. The program would never leave you alone...
     
  14. Perris Calderon

    Perris Calderon Moderator Staff Member Political User

    Messages:
    12,332
    Location:
    new york
    you can log percentages and watch for increase?
     
  15. Johnny

    Johnny .. Commodore .. Political User

    Messages:
    5,015
    Location:
    Happy Valley
    Raid10 = JBOD. Remember to back up your data.
     
  16. fitz

    fitz Just Floating Along Staff Member Political User Folding Team

    Messages:
    4,076
    Location:
    Chicagoland
    umm.. no

    JBOD as the acrynom is "Just a Bunch Of Disk" meaning things like disk spanning, volume spanning, etc..

    RAID10 is a mirror set (or a group of mirror set's) with data striped across each mirror set and is a specific variant of RAID.

    It does not obviate the need to backup data (as explained above that RAID does not equal backup) but it prevents data loss in the event of single (and in some cases) dual drive hardware failures. JBOD provides for no data redundancy in the event of a drive hardware failure.
     
  17. ThePermster

    ThePermster OSNN One Post Wonder

    Messages:
    4
    Hrmmm well to save money then I guess RAID1 it is, I'm not too bothered about a performance increase anyway so no big deal.

    Question though, whilst this might turn out more cost effective and productive, it doesn't sound like it's gonna be space efficient, i.e. what do I do when I run out of storage space again as I won't have any bays for new HDDs left by this point probably? What would be a good way of physically having a neatly kept, large amount of HDDs or other storage devices in one place? I've heard about these external cases for sticking lots of internal HDDs into, are they worth looking into or are they designed more with other needs in mind?

    I'm thinking I might just do the whole backup thing and sod RAID just so I don't have to worry about running out of storage space for a long time.
     
  18. LordOfLA

    LordOfLA Godlike!

    Messages:
    7,027
    Location:
    Maidenhead, Berkshire, UK
    you could take a look at drobo
     
  19. X-Istence

    X-Istence * Political User

    Messages:
    6,498
    Location:
    USA
    Ehm, you can use a SMART monitoring tool to monitor your drive while it is in the computer. Gives appropriate diagnosis, and information on what is going on inside the drive.

    Yeah, but error correcting is not the same as moving a sector.

    SMART monitoring tool again.
     
  20. ThePermster

    ThePermster OSNN One Post Wonder

    Messages:
    4
    Wow those drobos look cool, I'm gonna have to get me some monehs.

    Right, well now I'm considering not RAIDing, if I were to start using RAID I would just end up having to buy expansion bays and/or an enormous case in no time.

    So now I just need a backup solution. Problem is, about 750Gb of the terabyte of information I have at the moment isn't expendable, s'all uncompressed audio and video that I'm forever in the process of using.
    So how do I go about storing that kinda data in backup? I don't think an external HDD for backup will work 'cause the whole point of this upgrade is so I can eventually store up to 3Gb and I've yet to see an external HDD with that kinda capacity, or at least one within my price range.
    Is there any feasible way to do that?

    And I'm not buying DVDs, not only would I need a few thousand but I want a backup solution I can reuse realistically.