Storage Options for iMac 5K

Discussion in 'iMac' started by Techtrender, Nov 14, 2014.

  1. Techtrender macrumors newbie

    Joined:
    Nov 7, 2014
    #1
    I'm about to make a huge transition from the PC world to the Mac world.
    I have always worked with both systems in my work field. But this will be the first time ever that a Mac will be my main workstation.

    I'm in the process of ordering the new iMac :apple: 5k with i7, 16gb ram, M295x, 512 GB SSD.

    I currently have about 12TB on an NTFS RAID 1 file server that is half full. I would like to copy the data onto a new storage device to be available at all times via Thunderbolt 2.0 on the new iMac. I plan on keeping the NTFS file server for weekly backups.

    I would like to purchase a 4 bay storage device that has thunderbolt 2.0 with support for at least RAID 0 and 1.
    I'm thinking of adding 4x6TB in it for a total of 24TB to be on RAID 1 for an actual size of 12TB.

    Any recommendations for a storage device that fits my needs will be greatly appreciated.

    I'm considering these two options:
    AKiTiO Thunder2 Quad
    OWC ThunderBay 4

    Thank you very much in advance.
     
  2. markyr17 macrumors 65816

    markyr17

    Joined:
    Apr 8, 2010
    #2
    I personally bought the Thunderbay IV

    http://eshop.macsales.com/shop/Thunderbolt/External-Drive/OWC/ThunderBay-4

    I bought it without any storage, and put my existing hard drives in it. It's great, I love it!!
     
  3. Alesc macrumors 6502

    Alesc

    Joined:
    Nov 11, 2014
    Location:
    France
    #3
    Just one piece of advice: don't buy your RAM on the Apple store. Take the minimum (8 GB) and buy 2x 8 GB elsewhere: you will have 24 GB and it will cost you less than the 16 GB on the Apple store...
     
  4. lewisd25 macrumors 6502a

    Joined:
    Jul 6, 2007
    #4
    I can also vouch for the OWC Thunderbay IV. I swapped all of my hard drives from my old Mac Pro 1,1 to the Thunderbay without issue. The fan noise isn't as bad as people claim.
     
  5. thetoness macrumors member

    Joined:
    Jan 16, 2008
    #5
    Get the Thunderbay 4 not Thunderbay IV, as the former is thunderbolt 2 and the latter is thunderbolt 1
     
  6. bbbrisco macrumors newbie

    Joined:
    Nov 13, 2014
    #6
    I'm thinking of adding 4x6TB in it for a total of 24TB to be on RAID 1 for an actual size of 12TB.

    Any particular reason for going RAID 1 instead of RAID 5 (assuming you buy an enclosure with HW support for RAID 5)? Seems to me like RAID 5 would be substantially faster than RAID 1, give you 18GB instead of 12 and still allow for a drive to fail without issues.
     
  7. 5iMacs macrumors regular

    Joined:
    Oct 25, 2014
    #7
    I'm not familiar with this enclosure in particular, but I have not seen cases where RAID5 is faster than RAID1, all things being equal. Especially in single-user scenarios without a lot of concurrent activity.
     
  8. bbbrisco macrumors newbie

    Joined:
    Nov 13, 2014
    #8
    I'm no RAID expert, but I was basing this off http://www.larryjordan.biz/product-review-owc-thunderbay-4-part-2/ and
    http://macperformanceguide.com/SoftwareRAID-why-RAID5.html which claims that RAID5 can approach the speed of RAID0 (I would think that performance would get closer as number of drives in the RAID gets larger).

    It seems to me that using RAID 1 over 4 drives (2 sets of mirrored drives I'm assuming), has few advantages over RAID 5 and the drawback of lower capicity and (likely but not willing to bet huge $$$ on it) lower performance. With two sets of mirrored drives I guess you have a scenario where 2 HD's can fail, as long as they are one from each mirror, as opposed to the 1 failure allowed with RAID 5, but that seems like a pretty weak benefit.

    Of course, there are the ancillary issues which can make an enormous difference - ie time to rebuild a RAID, software management issues. I've not used a RAID setup on OSX yet, so I may be missing some practical issues which make RAID 1 simpler or more appealing than RAID 5.
     
  9. Techtrender, Nov 14, 2014
    Last edited: Nov 14, 2014

    Techtrender thread starter macrumors newbie

    Joined:
    Nov 7, 2014
    #9
    Thanks!

    Great tip! Thanks! I sure will do that.


    Thanks!

    Great! Thanks!

    It will be great to have RAID 5 if it's available on any of these devices. It will actually be more practical since I'll be using the RAID 1 NTFS file server as a backup anyway.

    Thanks!
     
  10. 5iMacs macrumors regular

    Joined:
    Oct 25, 2014
    #10
    This is great context, I see where the recommendation is coming from, especially in a video editing application.

    With this many drives I think RAID5 will win on read speed and capacity and lose on write speed, compared with RAID 1+0. So depending on your use case there are definitely some advantages to RAID5.
     
  11. Techtrender thread starter macrumors newbie

    Joined:
    Nov 7, 2014
    #11
    Update:

    I just ordered the :apple: iMac :apple: 5K 27" with i7, 8GB, M295x, 512 GB :apple: with delivery date Dec 2nd to Dec 9th

    Also, after days of going back and forth between the OWC Thunderbay 4 and the Akitio Thunder2 Quad, I finally decided to go with the Akitio. I found it at a great price and order it as well.

    In my search for a good thunderbolt 2 storage device, I found that both devices have great user reviews and similar benchmarks.

    If anyone is interested in my configuration or the Akitio, let me know and I'll post some photos and benchmarks once I have everything up and running.
     
  12. joema2 macrumors 65816

    joema2

    Joined:
    Sep 3, 2013
    #12
    Thanks for the update. Will you be using the built-in OS X RAID 0/1, as you originally posted, or a 3rd party product like SoftRAID for RAID5?
     
  13. Techtrender thread starter macrumors newbie

    Joined:
    Nov 7, 2014
    #13
    You're welcome joema2. As soon as I receive the Thunder2 Quad, I'm going to test the performance across all available RAID types (RAID 0, 1 and 5) to see how well the Akitio Thunder2 Quad performs.

    As of now, I'm leaning more towards the RAID 5 option because it will give me both the increased disk performance as well as the extra disk space and of course, the extra protection layer that allows to lose one drive to failure without affecting the data.

    As mentioned earlier, I do have a RAID 1 backup server. So I'm still open to all available options. The only thing that makes me hesitate when it comes to RAID 0 is the thought of having to move Terabytes of data from the server back to the Thunder2 Quad and the downtime that will occur. Also with RAID 0, I'll have to setup daily backups to the server instead of weekly backups.

    What would you go for if you're in my situation? RAID 0, 1 or 5?
     
  14. jdphoto macrumors 6502

    jdphoto

    Joined:
    Jan 13, 2014
    #14
    I think you're right on with RAID 5. Performance will be pretty good, and it will definitely be worth it to have the tolerance of 1 drive failure. That drive failure will naturally happen at the most inopportune time so it'll be good to still have access.
     
  15. joema2 macrumors 65816

    joema2

    Joined:
    Sep 3, 2013
    #15
    I use a 4-drive Pegasus R4 8TB RAID5 array, and backup to two different 8TB 2-drive RAID0 arrays. I use RAID0 on the backup drives because it gives faster backup and restore performance, plus they are redundant. I have yet other backups I take periodically on 4TB single drives.

    Classically, software-based RAID5 has slow write performance but SoftRAID-5 on the OWC Thunderbay4 is pretty fast. Since that's just a JBOD array, I assume the speed comes from the software, so it should be about the same as the Akitio Thunder2 Quad.

    I'd be afraid of a 4-drive RAID0 array since if any drive fails you lose the whole thing. Also well-done RAID5 isn't much slower than RAID0, at least to the extent it makes a difference in application workflow.

    Remember: it's not just about benchmarks, but whether that makes a material difference in your application workflow. If RAID0 gives you 650 MB/sec and RAID5 only 500 MB/sec, you probably won't even notice that in most applications. However if you lose your data, you'll definitely notice that.

    Even if you have excellent backup and can restore everything in one day, how long would it take for each little saved sliver of time from RAID0 to equal a full day of lost work?
     
  16. Bryan Bowler macrumors 68040

    Joined:
    Sep 27, 2008
    #16
    This is one of the most well thought out and informative replies I've ever seen. Thanks for sharing such great advice!

    Bryan
     
  17. Techtrender thread starter macrumors newbie

    Joined:
    Nov 7, 2014
    #17
    Thanks jdphoto, agree! One can't jeopardize losing work on a major project with a looming deadline due to RAID0 failure.

    Thanks joema2, as Bryan had said, your reply is very informative and well thought out. It's greatly appreciated.
    I'm now convinced that RAID5 is the best solution for my needs.

    Agree %100
     
  18. WinstonRumfoord macrumors 6502

    Joined:
    Mar 27, 2014
    #18
    I'd love to know your r/w Raid 5 performance with 7200RPM drives.

    I am about to buy the akitio quad and put in 4 x 3TB 7200 in a raid 5.

    Thanks!
     
  19. bjmoose macrumors member

    Joined:
    Oct 4, 2011
    #19
    RAID 5 is not as reliable as you might think as we get bigger and more dense devices: see this article for details.
     
  20. Melodeath macrumors 6502a

    Joined:
    Dec 9, 2009
    #20
    What is your procedure for weekly backups using an NTFS RAID?
     
  21. joema2 macrumors 65816

    joema2

    Joined:
    Sep 3, 2013
    #21
    I don't think that article is correct. It implies you can't even scan a large RAID array once without a significant probability of encountering a non-recoverable error. If that was correct, errors would be popping up all the time for any user of heavy I/O on *any* large RAID, whether RAID0, RAID1 or RAID5. We don't see that happening in the real world.

    Instead of RAID5 for home users, he suggests instead disk-to-disk backups: http://www.zdnet.com/article/why-raid-6-stops-working-in-2019/

    If the limitation on RAID5 is likelihood of hitting an unrecoverable read error during a rebuild, why is doing a disk-to-disk copy of the same data any different? They would both have about the same chance of encountering an error. At least with RAID5 you can keep working during the rebuild, but with a disk-to-disk copy that may not be possible.

    He took a drive reliability spec, apparently did no personal testing or sanity checking, then wrote multiple articles on that.

    Here is possibly his error: he looked at a typical HDD unrecoverable read error rate of 1 in 10^14 reads, equated this to bytes, then divided by the number of HDD in the array (8 in his case), which gives his number of 1 URE per 12.5 TB.

    HOWEVER: the URE rate is 1 in 10^14 *reads* -- NOT bytes. One read is a sector read. Modern drives use 4k sectors, not 512-byte sectors.

    10^14 *sectors* is 10^14*4096 bytes per sector = 4.096e17 bytes, or 409,600 terabytes. So on average you'd expect one unrecoverable read error per 409,600 terabytes (for a single drive). I don't think an n-drive array would be any different than a single drive of the same size, since you're doing 1/nth the reads off each drive. A multi-drive array has a greater failure chance per unit TIME, not per unit data transferred.

    For my 4-drive 8TB RAID array, I'd have to read the whole thing 51,200 times before hitting an unrecoverable error. I haven't done that but I have scanned it multiple times using Scannerz (http://scsc-online.com/Scannerz.html) without encountering any errors, which according to the above article should be impossible.
     
  22. bjmoose macrumors member

    Joined:
    Oct 4, 2011
    #22
    Thanks - excellent rebuttal to his articles. I believe you are entirely correct and agree with your analysis. I stand corrected.
     
  23. joema2 macrumors 65816

    joema2

    Joined:
    Sep 3, 2013
    #23
    I'm still investigating this. I think my above numbers and reasoning are incorrect, because the HDD non-recoverable error spec is failed reads per total bits read, not sectors read. However the fact remains, hard drives and RAID5 is much more reliable than indicated by the manufacturer's "non-recoverable error rate".

    If the spec is 1 failure per 10^14 bits read, 10^14 bits = 12.5 terabytes. So by that spec you'd expect a failure on average every 12.5TB -- that's where Robin Harris gets his number.

    However -- we know from observation that HDDs and RAID systems do not fail anywhere near that often.

    One answer is the spec is simply a "worst case" spec -- IOW a guarantee the HDD will be no *worse* than that. It is not an average failure rate, nor a predicted failure rate. It's more like an uptime or availability guarantee. If a vendor promises 90% availability, that doesn't mean the system is unavailable 10% of the time. It may well achieve 99% availability -- it's just a worst case guarantee.

    If the HDD is really more reliable than, say, 1 error per 10^14 bits read, why don't they say that? We could just as well ask if the average engine in a Honda car will last 180,000 miles, why is the engine warranty only 60,000 miles? There are many reasons for that.
     
  24. joema2 macrumors 65816

    joema2

    Joined:
    Sep 3, 2013
    #24
    A key study (while several years old) covers this exact area: namely the disparity between the "non-recoverable error rate" spec published by HDD manufacturers and empirically-observed results. A spec of one non-recoverable error per 10^14 bits read would equate to one error every 12.5 terabytes read. This study found four non-recoverable errors per two petabytes read, which equates to one error per 4E15 bits read, or about 40 times more reliable than the HDD manufacturer spec. Empirical Measurements of Disk Failure Rates and Error Rates (2005, Jim Gray, et al): http://research.microsoft.com/apps/pubs/default.aspx?id=64599

    The bottom line is all the ZDNet RAID5 "Gloom and Doom" articles by Robin Harris are off base and without merit. You can't use the HDD manufacturer's worst-case error rate spec as a guideline for predicted real world reliability.
     
  25. bjmoose macrumors member

    Joined:
    Oct 4, 2011
    #25
    A good blog post on disk failure, and failure to fail, and how to build a large scale disk farm that is robust. Has some interesting points and useful links in it.
     

Share This Page