RAID Array for Long Term Redundant Storage across multiple systems builds

First I want to start off by saying, that I am in the planning stages of this. I have bought no hardware, and everything is a possibility to ensure the success of this project while keeping convenience and portability in mind.

I am looking for a solution that will allow me to have 2 RAID arrays comprised of 2x 3Tb drives each. These will be my LTRS drives. Referred to as LTRS0, first array, and LTRS1, second array. These drives are meant to store my work, music, videos, and documents from system build to system build. I want to ensure their long term integrity by placing them within RAID 1. Basically, LTRS0 will have 2 HDD of 3Tb in RAID 1, LTRS1 will have 2 HDDs of 3TB in RAID 1, but they can contain different data.

I want to know what my best option is based on these requirements. Should I get a PCI(e) card that I can take from system to system as I build them, or should I get a NAS and put all of my data in that, or should I use the on board motherboard controller and just make sure I have the same controller from system build to system build?

  1. The issues that come with each are somewhat daunting as well, as for example the first option PCI(e) might get in the way of a Cross Fire or SLI setup.

  2. The second option NAS might be slower then internal storage as it’s going across a 1Gbps link and not say a 3Gbps SATA, or 6Gbps SATA interconnect.

    The other option would be a light path / thunderbolt NAS but those tend to run expensive and I don’t own a Mac, and I can’t seem to find a Thunberbolt PCI(e) card on line, and that might have the same issues as the first option (getting in the way of an SLI / CX setup.

  3. The third option might not work as well because manufactures might change RAID controllers from motherboard to motherboard and that would reduce the LTRS lifetime.

So, I turn to you, of the options given, what’s the best one to meet the goal of long term redundant storage. If you could provide product details that would be great, or information on your own in house solution.

Answer

Use software RAID. You don’t say what operating system you’re using, but most modern operating systems (OSX since 10.2, Windows since at least 2000 but only for the most expensive editions, Linux since forever, *BSD since a long time ago, …) support RAID out of the box. Software RAID is more reliable than hardware RAID, precisely because you’re not dependent on some hardware controller that may be incompatible with others. There can be a performance advantage to hardware RAID (the expensive kind, not the entry-level priced controllers) for RAID modes that require computing checksums (e.g. RAID-5 or RAID-6) but not for the simple data duplication of RAID-1.

On re-reading your question, I see that you’re linking RAID to long-term storage. There isn’t any connection between the two really. RAID is for short-time redundancy, to make it more likely that you’ll survive a disk failure with zero downtime. For medium-term data storage (a few years)¹, what you need is offline redundancy, i.e. backups. There’s no major advantage in retaining the disk arrangement for a long time; moving data between disks every couple of years is not a big deal. Serious operating systems even let you do this online, though this is a moot point if you’re connecting the drives to a different motherboard anyway.

A more important issue for you, especially given that you want your data to be accessible from multiple operating systems, is what filesystem to choose. With a NAS, you can isolate the data from your operating system (any operating system should support NFS or Samba), but network filesystems are often noticeably slower than local ones. There isn’t any very good option to share data between dual-booted Windows and Linux: the only serious choice for Windows is NTFS, which has correct but slow support under Linux. The current recommended choice for common requirements under Linux is ext4, which is not easily usable in Windows. If you dual-boot Linux and Windows, consider a NAS (in the form of a Linux or *BSD small-form-factor PC) plus local replication under each operating system’s prefered operating system if the performance is insufficient (benchmark first).

¹ Long-term storage (tens of years and more) is yet another completely different issue.

Attribution
Source : Link , Question Author : Mark Tomlin , Answer Author : Gilles ‘SO- stop being evil’

Leave a Comment