Partitioning a RAID 5 Setup for Highest Performance

I recently purchased an OWC Thunderbay 4 RAID 5 Edition and I'm ready to format and partition the drives with SoftRAID, which is an excellent piece of software for hard drive management. (I've already certified all of the drives, to include my cold spare.) In a quest to optimize performance, whether real-world or benchmarked numbers, I need to get some opinions from my fellow experts.

Here's my situation:
I'm a heavy-duty photographer that's also getting into a fair amount of video work.

Here's my set-up:
OWC Thunderbay 4 RAID 5 Edition w/ four 5 TB drives and one cold 5 TB spare drive.
Will back-up onsite to a Time Capsule and various other stand-alone hard drives.
I also backup to off-site drives once a month by bringing them home and then return them to their offsite location the next day.

Here are my questions/dilemma:
I want to to squeeze as much performance as I can out of my RAID 5. Digiloyd's excellent article here has given me the idea to partition my RAID 5 so that I utilize the faster sections of the drives for things like video and photo editing, and then use slower partitions on the drive for things that don't require speed, such as documents, music, etc. It's also beneficial as it keeps the partitioned volumes small enough so you can back them up to individual 5 TB drives.

Here is how I'm thinking about setting it up:

Partition 1, 5 TB, partitioned to the fastest section on the drives: Video Footage Libraries (FCPX)
Partition 2, 5 TB, partitioned to the next fastest section on the drives (only slightly slower than Partition 1): Photo Libraries (Currently Aperture, eventually migrating to Capture One Pro or Lightroom)
Partition 3, 4 TB, partitioned to the slowest section of the drives: Documents, iTunes Library, Finished movies, etc.

What do you think? I know a lot of you have vast experience in this area and I would really appreciate your opinions.

Thanks for the help/advice!
Bryan

Honestly I'd just have one big drive with folders for logical organisation.

I'm using Microsoft ReFS (a reliable file system that validates and error checks each byte on the drive) with storage spaces in mirrored RAID, but it can do any RAID level. Suggest you look into that as an option, it's similar to ZFS in the way it prevents bit rot.

Tim,

Thanks for posting your thoughts. You've been at this a long time and I'm always appreciative when I read your advice. As for how I format and potentially partition my RAID 5 setup (I am definitely going with RAID 5), do you think it is really better to just have one big drive versus reserving the outer third of the drive platters for the fastest needs, and then setting aside the slower inner rings of the drive platters for the data that doesn't require speed?

http://macperformanceg​uide.com …uNeedMoreThanYo​uNeed.html
http://macperformanceg​uide.com …rIsBetter-Fastest2TB.html
http://macperformanceg​uide.com …optimizingNewSt​orage.html

I realize that my Thunderbolt RAID 5 setup will be extremely fast no matter what, but I'm trying to optimize my setup now to future-proof against any possible need for even more speed in the future. I feel like getting this right now might be beneficial later.

Bryan

Are these going to be production drives or just storage drives?

If storage drives I wouldn't worry about any differences between write locations on the platter.

If these are production drives I would first investigate what the actual throughput would be over thunderbolt. Worrying over these small differences may be moot.

silvrr,

This RAID 5 setup will be used for storage and production, which is why I was considering partitioning it into three 5 TB sections and then utilizing the fastest 5 TB section for video production.

Utilizing the OWC Thunderbay 4 with 5 TB Toshiba 7200 RPM drives via a Thunderbolt 2 interface, read speeds measure in at 520 MB/s for the fastest partition and a read speed of 440 MB/s for the slowest partition, which is an 18% improvement in read speeds if using the faster partition.

Admittedly, I don't anticipate editing with more than 3 multicam streams, but who knows what the future will bring. So, while the improvement may not offer any real-world advantages today, I wonder if it will tomorrow. Disadvantages would be limiting myself to 5 TB volumes. Again, not a problem for now, but perhaps an inconvenience later.

Thanks for the inputs everyone -- let's keep those opinions coming!

Bryan

I'm not convinced that a 20% difference in disk speed will make enough of a difference to add this (mild) complexity and risk during any possible data recovery. Though with software RAID I strongly suggest good backups so you don't need to do data recovery, which may be virtually impossible if anything goes wrong anyway. Personally I prefer the KISS principle. Can you do a benchmark of something video related that is performance critical to you? Say a big rendering or conversion or something. Copy the source material to both virtual disks, reboot, run a test on a disk, reboot, run another test. That eliminates caching as a factor.

440MBps is WAY faster than a normal disk, faster than most SSDs. I know that for photography it matters where your cache is, but whether your images are on a 100MBps disk or 500MBps SSD makes little difference to performance.

Tim, thanks again for your feedback -- much appreciated.

If I understand you correctly, it sounds like you're a big fan of simplicity and prefer to avoid RAID 5 in case there is a malfunction in the software or hardware controller and your data becomes corrupted. Am I understanding you correctly?

There are so many different ways to skin this cat and I totally understand your approach, but I'm willing to accept the (very minor) risk levels associated with RAID 5 and parity data in exchange for what I believe are the tangible benefits of RAID 5, which is great speed and the ability to continue working and recover from one drive failure. Of course, no system is perfect, and there is always a very slight risk of something going wrong with the RAID controller. I just feel the odds of a hard drive going bad are greater, which is why I'm choosing to go the RAID 5 route and accept the remote/minor risk associated with it.

Now when it comes to running 3 different partitions on a RAID 5 array, I'm not sure if there is any additional risk when compared to just using only one big partition on the array. Does this statement seem reasonable?

Thanks,
Bryan

I'm not saying don't RAID, I'm saying I'm not sure partitioning is worthwhile. But if you can use the speed, sure, go for it, I don't think it makes things too much more complex. Do keep good backups.

Cool, thanks Tim!

I'll sort out what I'm going to do and post my observations as I learn more.

Cheers,
Bryan

if youre really worried about raid performance then use raid 0 and have done with it.

after all any raid should be fully backed up so the potential lack of redundancy risks should be minimal.

neil_g wrote in post #17432518
if youre really worried about raid performance then use raid 0 and have done with it.

after all any raid should be fully backed up so the potential lack of redundancy risks should be minimal.

Thank for the suggestion. RAID 0 is a good choice for some people, but I've already decided that RAID 5 is the right choice for me. In the event of a drive failure, I don't want production to miss a beat.

After having given this an enormous amount of thought, I still feel that partitioning a RAID 5 and utilizing the slower partitions for data that does not require high speeds is a smart idea. If you have a race car, why not make a super easy modification to maximize performance? SoftRAID makes this easy to do. It could also be beneficial if you use single hard drives in your backup solution. For example, if your partitions are no more than 4 TBs, then you can use individual 4 TB hard drives to back up each partition. Useful if you have extra hard drives laying around and don't want to invest in another RAID enclosure for your backups.

Thanks again everyone for the feedback. Keep those thoughts coming. It's great to exchange ideas!

Bryan

fair enough but even with raid5 you're still going to have a degraded performance after a drive failure while the array rebuilds. this is going to be lengthy the larger the array.

personally i echo what has been said above, i do not believe partitioning will give a noticable real world benefit. ive never come across any of the big storage players recommending this while working with business class storage.

bps wrote in post #17432759
Thank for the suggestion. RAID 0 is a good choice for some people, but I've already decided that RAID 5 is the right choice for me. In the event of a drive failure, I don't want production to miss a beat.

After having given this an enormous amount of thought, I still feel that partitioning a RAID 5 and utilizing the slower partitions for data that does not require high speeds is a smart idea. If you have a race car, why not make a super easy modification to maximize performance? SoftRAID makes this easy to do. It could also be beneficial if you use single hard drives in your backup solution. For example, if your partitions are no more than 4 TBs, then you can use individual 4 TB hard drives to back up each partition. Useful if you have extra hard drives laying around and don't want to invest in another RAID enclosure for your backups.

Thanks again everyone for the feedback. Keep those thoughts coming. It's great to exchange ideas!

Bryan

That point about < 4TB making backups easier is very valid, and possibly the most compelling reason I've read so far for partitioning.

youre going to have to plug-unplug drives either way, if youre only using single drive backups.

After reading the helpful replies here, and pulling in some info from a different set of forums, I've now realized that my original proposal has something very wrong with it. If you setup too many partitions and you end up trying to access data across those partitions at the same time, then you might possibly hurt system performance by causing longer seek times/drive head movement. (Sorry if I'm stating the obvious here, but as a newcomer to the complex world of storage, I wasn't aware of this.)

It now seems to me the better way to do this is to only setup two partitions:

Partition 1: the first 70% of the drives, which is faster than the last 30%. Use a regular folder structure within this partition and use it for nearly everything.

Partition 2: the last 30% of the drives, which is definitely the slowest. Only use this partition for things like PDFs, scanned documents (I file a lot of them), and finished videos that are rarely viewed, and when they are, you definitely view them by themselves and not in conjunction with another piece of software. By creating this partition, you keep the really slow parts of the hard drive out of the way, but at the same time, you also utilize it for something so it's not just wasted space.

Am I on the right track now? If so, I want get this puppy setup and start migrating data.

Thanks,
Bryan