Best RAID5 NAS Hard Drives
- Thread starter dpavid
- Start date
- Sort by reaction score
joema2
macrumors 68000
I've had good results with the 7200 rpm 8TB HGST Ultrastar "Helium" drives. They are a bit loud for desktop use but they work fine in a NAS or server: http://a.co/fXUhnuD
As an Amazon Associate, MacRumors earns a commission from qualifying purchases made through links in this post.
There are bad batches, I have some 5 year old red and a couple red pro drives in a 24/7 OWC thunderbay that are still going. But yeah, it seems like using any drive longer than the warranty and you are on borrowed time. All of my several Seagates and older HGSTs failed within months of warranty expiration, typically two years.
You may want to check in the peripheral forum for more ideas https://forums.macrumors.com/forums/mac-accessories.77/
You may want to check in the peripheral forum for more ideas https://forums.macrumors.com/forums/mac-accessories.77/
Last edited by a moderator:
JamesPDX
Suspended
Even the HGST Ultrastar drives are not what they once were. How much storage do you need? Are you editing video or just storing a lot of media, etc?
MacPro via Thunderbolt 2 to Areca RAID 5 array with 8 bays. Previous it was filled with 3TB WD RED NAS drives for 24 total TB, 21 usable TB (1 parity drive). 4K video editing and storing mass amounts of media.
I never buy/trust WD for my systems, but I am an old dog. I go with Seagate, not the cheap-desktop ones, there have been terrible bunches in the past, as have been for other brands at the lower level/price. The NAS series from Seagate has been good enough for me, lots of ST4000VN000's none has failed. The newer series (IronWolf) should also be good.
As for disks I am not a fan of Western Digital as I have had a lot of failures. WD has been one of the worst performing drives on the BackBlaze disk failure reports. Here's the latest one:
https://www.backblaze.com/blog/hard-drive-benchmark-stats-2016/
I think they are no longer using them. Not sure of the reason.
https://www.backblaze.com/blog/hard-drive-benchmark-stats-2016/
I think they are no longer using them. Not sure of the reason.
JamesPDX
Suspended
Would you consider filling the Areca with SSDs? Is it hardware RAID or software RAID? FWIW, know that you can't TRIM an SSD RAID, at least not from the (Apple Macintosh Computer) Disk Utility.
[doublepost=1490073835][/doublepost]
Of course everyone knows by now that WD bought HGST (Hitachi made those little 2.5" spinners for Toshiba, now long gone, but I snapped up several for cloning.) Anyway, know that my "HGST" Deskstar 4TB drive that fails about every 18 months always gets replaced by HGST (Western Digital) with a refurbished unit sent from the Mira Loma, CA facility.
So figure that any WD/HGST drive will last about 2 years and then begin to die. When submit an RMA for warranty replacement, you will eventually receive a refurbished drive. Rinse and repeat until the warranty expires on the original drive.
My Hitachi 3TB Ultrastars are going strong, but I'm not going to let them run 24/7. I now subscribe to Backblaze, bless their hearts...
[doublepost=1490073835][/doublepost]
Of course everyone knows by now that WD bought HGST (Hitachi made those little 2.5" spinners for Toshiba, now long gone, but I snapped up several for cloning.) Anyway, know that my "HGST" Deskstar 4TB drive that fails about every 18 months always gets replaced by HGST (Western Digital) with a refurbished unit sent from the Mira Loma, CA facility.
So figure that any WD/HGST drive will last about 2 years and then begin to die. When submit an RMA for warranty replacement, you will eventually receive a refurbished drive. Rinse and repeat until the warranty expires on the original drive.
My Hitachi 3TB Ultrastars are going strong, but I'm not going to let them run 24/7. I now subscribe to Backblaze, bless their hearts...
joema2
macrumors 68000
It's not Disk Utility but a built-in terminal command enables TRIM on 3rd party SSDs, including RAID:
http://osxdaily.com/2015/10/29/use-trimforce-trim-ssd-mac-os-x/
The TRIM status can be checked with this command: system_profiler SPSerialATADataType | grep 'TRIM'
It returns Yes or No for each SSD drive.
TRIM status can also be checked using System Report in "About this Mac", clicking on the category SATA/SATA Express, then scrolling down the right pane until reaching the SSD drive(s), and below that an info pane lists whether TRIM is enabled on that drive. Even though the category is SATA, it also includes Thunderbolt devices. I don't have any USB SSDs, but they might be listed under USB in the system report.
Obviously use any such commands at your own risk. I've had no problems running TRIM on my 8TB RAID-0 array composed of 4 x 2TB Samsung EVO 850s in an OWC Thunderbay 4 Mini chassis.
I have to ask, why RAID5? Cautionary tale: I've used raid5 for years. It's never once saved me from a problem, it has instead led to data loss twice. The most recent was when, following a power problem, the raid enclosure itself failed. In talking to the manufacturer I learned they had revised the product, the newer revision uses a different striping format, so disks could not be simply moved to the newer enclosure.
Had I been using raid1, I could have taken either individual disc and inserted into a Mac directly, or an alternative external enclosure and it would have worked.
Fortunately there was a unit identical to mine available on eBay, unfortunately only one, and over-priced. I was lucky though, I purchased it, and my old drives, complete with data are working fine.
I chose raid5 originally to get the capacity I wanted. Since then, drive capacity has increased, I'm now planning on using raid1 with 5tb drives. I will also be able to backup the whole volume to another single external drive, something that wasn't possible when I purchased my original raid5 setup.
So... people, be very wary of raid5, ask yourself, do you really need it, would either a single raid1 volume, or dual raid1 volumes be better now disc capacity is what it is? Even raid1+0 is a better way to go if speed is a concern.
JamesPDX
Suspended
What's the fan and powersupply noise situation on that OWC?
joema2
macrumors 68000
It is very quiet but most of the time I have several spinning RAID arrays going so this swamps any tiny noise from the SSD array. The few times I have disconnected all spinning arrays except the OWC Thunderbay 4 Mini, I don't notice it.
[doublepost=1490181215][/doublepost]
Your problems were not caused by RAID5 but by using a proprietary hardware implementation. Had you used SoftRAID (which is just as fast) you could have put those drives in any generic enclosure and immediately used them.
However your experience shows that any type of RAID is susceptible to a common failure. The possible sources are many -- chassis, application software error, device driver error, file system error, user error, virus, etc. RAID5 (and similar) protects from one thing -- a physical hard drive failure. For this reason having good backups is important whether you use RAID or not.
If using SoftRAID, RAID5 has good performance for both reads and writes, and has the least storage overhead of all parity RAID formats. For common video I/O workloads, RAID5 works very well. However it has a write penalty for small random I/Os such as done by FCPX to build Event Browser thumbnails. OTOH I've tested a four-drive spinning RAID5 array vs a four-drive SSD RAID0 array, and the spinning RAID5 array using SoftRAID is still pretty good on most workloads.
Interesting, I'll check that out, thank you. Critical data was indeed backed up, and that backup is accessible.
JamesPDX
Suspended
But I thought that the whole point of hardware RAID was to offload the task onto dedicated hardware, while reducing risk to the RAID by RAID software malfunction.
If you want to stick with that unit, can you afford some enterprise-class SSDs? -They may be mostly SAS connections. If you give up and want to stick with SATA3, I'd go for 1-3 Black Magic Design Multidock II's. (scroll down) -and fill them up with 1TB or 2 TB SSDs.
If you want to stick with that unit, can you afford some enterprise-class SSDs? -They may be mostly SAS connections. If you give up and want to stick with SATA3, I'd go for 1-3 Black Magic Design Multidock II's. (scroll down) -and fill them up with 1TB or 2 TB SSDs.
joema2
macrumors 68000
That is the old viewpoint, back in the days when CPUs were slow and dedicated hardware was proportionately faster. Today CPUs are vastly faster and the so-called "dedicated hardware" is often the cheapest microcontroller they can find that (barely) does the job. That "hardware" microcontroller is simply running onboard software. It can have a bug just as easily as driver-level RAID software running on the CPU.
I have tested back-to-back similar drives in both a Promise Pegasus R4 (which uses hardware RAID-5) vs an OWC Thunderbay 4 (using SoftRAID), and the OWC/SoftRAID combination is faster.
The big downside to hardware RAID is it's proprietary and non-standard. You can't take drives formatted in a Promise Pegasus and put them in any other RAID box. However with SoftRAID you can interchange the drives in any generic RAID enclosure.
I like the Promise Pegasus line -- the chassis is well designed and very quiet. It is quieter when doing lots of seeks than the OWC. However after having used both it and OWC/SoftRAID, I prefer the latter's flexibility.
phrehdd
Contributor
dpavid, what is the volume you really require? Was your system fast enough using the 5400 rpm drives? I ask as there are other options available within RAID set ups with an enclosure that handles that many drives. In fact you may want to have less drives devoted to work and other drives within devoted to "storage" with possible mirroring etc. as well as using RAID 6 if you are worried about your work space.
As for drives, again depends on what you are trying to do. If 5400 rpm drives work, perhaps checking out the StorageReview site ( http://www.storagereview.com/ ) would be helpful.
Though there are SSD drives that are up to 1-2 tb, I don't think I would opt for those drives given cost. Perhaps investigating (if that size or similar) WD Raptor drives with good warranties and see if they do well in RAID for you.
You will find that some people are in the Seagate camp and others in the Western Digital camp and a few like other
makes. I have used both consumer level "NAS" drives (Red from WD and Seagates counterpart). I find they work reasonably well. Best drives I have used that were not enterprise were by chance - a bunch of Samsung drives back when that were I believe 5900 rpm rather than 5400rpm and they are many years old and operating in RAID to this day (for my friend now). Again, it might be worth time to got to StorageReview and some other sites that may inspire a direction for you in your search and set up.
As for drives, again depends on what you are trying to do. If 5400 rpm drives work, perhaps checking out the StorageReview site ( http://www.storagereview.com/ ) would be helpful.
Though there are SSD drives that are up to 1-2 tb, I don't think I would opt for those drives given cost. Perhaps investigating (if that size or similar) WD Raptor drives with good warranties and see if they do well in RAID for you.
You will find that some people are in the Seagate camp and others in the Western Digital camp and a few like other
makes. I have used both consumer level "NAS" drives (Red from WD and Seagates counterpart). I find they work reasonably well. Best drives I have used that were not enterprise were by chance - a bunch of Samsung drives back when that were I believe 5900 rpm rather than 5400rpm and they are many years old and operating in RAID to this day (for my friend now). Again, it might be worth time to got to StorageReview and some other sites that may inspire a direction for you in your search and set up.
Nice site. But in a quick look I didn't see any reliability data. Does anyone other than BackBlaze evaluate HD failure rates?
No, but it seems the best indication of reliability and durability is the warranty manufactures put on their drive. There can be flukes.
BlackBlaze's business model is to use the least expensive drives available and just replaces them when they fail. What you are seeing are the failure rate of the cheapest drives at the time. This is a strategy that others are using. Big data centers can save money, but the data does not translate well to the consumer or small centers.
Drives will fail, even SSDs. Be prepared. With the advent of large drives, most here don't bother with RAID5 unless they need the performance or up time. I stripe for performance, mirror for redundancy. Where stripes are not enough performance, I use SSD. It just takes so long for recovery more sophisticated RAID and the performance is a dog during recovery.. and then you usually need to deal with proprietary hardware enclosures that when they fail you may loose everything.
Anyway, reliability data posted publicly is unreliable. The best way to evaluate is to look at the manufacture's spec sheets (there are several kinds of from losing a bit of data to an entire drive) and the warranty offered. Then look at the current reviews being posted at outlets like newegg, B&H, Amazon for a trend in a recently manufactured model, may give you a heads up on bad batches.
BlackBlaze's business model is to use the least expensive drives available and just replaces them when they fail. What you are seeing are the failure rate of the cheapest drives at the time. This is a strategy that others are using. Big data centers can save money, but the data does not translate well to the consumer or small centers.
Drives will fail, even SSDs. Be prepared. With the advent of large drives, most here don't bother with RAID5 unless they need the performance or up time. I stripe for performance, mirror for redundancy. Where stripes are not enough performance, I use SSD. It just takes so long for recovery more sophisticated RAID and the performance is a dog during recovery.. and then you usually need to deal with proprietary hardware enclosures that when they fail you may loose everything.
Anyway, reliability data posted publicly is unreliable. The best way to evaluate is to look at the manufacture's spec sheets (there are several kinds of from losing a bit of data to an entire drive) and the warranty offered. Then look at the current reviews being posted at outlets like newegg, B&H, Amazon for a trend in a recently manufactured model, may give you a heads up on bad batches.
I just ordered 6 Toshiba MD04ACA600 6 TB from serversupply.com which had the lowest price ($214) I could find for the drives. It is the only 6 TB drive on the Promise R6 compatibility list so I didn't have any choice. [Interesting, I just checked and they are quoting $210 now]. It's some $20-$30 cheaper than any other site that is rated.
MD04ACA600
So if they are using the least expensive drives from all vendors, then I assume one would expect better performance from a NAS or enterprise drive? And in the back of my mind I'm thinking that if one manufacturer has better reliability in a cheaper drive than another manufacturer in the same price range, then there is a good chance that will hold true for a more expensive drive that is built for durability. Corporate culture thing.
Don't understand this statement. I looked at some of the statistical models that Backblaze used and it looked as if they are being as objective as possible. But then I'm not a statistician.
Certainly a starting point. And a 5 year vs a 2 year or less warranty certainly says something. And you probably get what you pay for.
But I guess I'm a skeptic. I don't generally believe what a manufacturer says about a product unless I've seen it verified by a 3rd party.
Storagerview.com certainly looks good to evaluate the performance part of their claim, but my main concern is reliability.
MD04ACA600
So if they are using the least expensive drives from all vendors, then I assume one would expect better performance from a NAS or enterprise drive? And in the back of my mind I'm thinking that if one manufacturer has better reliability in a cheaper drive than another manufacturer in the same price range, then there is a good chance that will hold true for a more expensive drive that is built for durability. Corporate culture thing.
Don't understand this statement. I looked at some of the statistical models that Backblaze used and it looked as if they are being as objective as possible. But then I'm not a statistician.
Certainly a starting point. And a 5 year vs a 2 year or less warranty certainly says something. And you probably get what you pay for.
But I guess I'm a skeptic. I don't generally believe what a manufacturer says about a product unless I've seen it verified by a 3rd party.
Storagerview.com certainly looks good to evaluate the performance part of their claim, but my main concern is reliability.
joema2
macrumors 68000
I don't understand the statement about publicly posted drive durability tests being unreliable; I think you mean Backblaze in particular. It is simply conveying reality -- actual real world experience with a vast number of drives. The results can be misleading or misinterpreted, but that doesn't mean the data itself cannot be relied on for the specific case.
E.g, one Backblaze study showed many failures on a particular model of 3TB Seagate drives. There is no basis to doubt the reliability of this study, as if you can't trust Backblaze. However the results of that study were often misinterpreted to mean Seagate drives in general were unreliable. That was obviously incorrect since subsequent Backblaze studies showed other Seagate drive models were very reliable. It was a bad batch of drives. The inability of many people to think logically and rationally doesn't mean the study itself was unreliable or flawed.
Yet another example are manufacturer spec sheets. There is no reason to think these are unreliable -- provided you understand what they are saying. However people often misunderstand the unrecoverable error rate in spec sheets.
A widely-publicized example was where ZDNet columnist Robin Harris wrote several incorrect articles based on a misunderstanding of error rates in HDD spec sheets. He thought it was the *expected* error rate. However the HDD "non-recoverable read error" spec is NOT a typical or average number. It is NOT the *expected* error rate. Rather it is a worst-case number. E.g, on the WD Caviar line, they specify "< 1 in 10^14". They are essentially guaranteeing it will fail no more frequently than 1 in 10^14 bits. It's somewhat like an auto manufacturer guaranteeing your vehicle for 60,000 miles. That doesn't mean it quits working at that point.
Just plain common sense should tell anyone this is not a typical number. 10^14 bits is about 12 terabytes. I have rebuilt my 8TB Pegasus R4 array dozens of times in tests, which required reading every bit on the entire array multiple times. I *never* got an unrecoverable read error, despite doing far more I/O than 10^14 bits.
This was also observed by Dr. Jim Gray (a founding father of the relational database) in this paper:
Empirical Measurements of Disk Failure Rates and Error Rates:
http://research.microsoft.com/apps/pubs/default.aspx?id=64599
This large meta study by Tom's Hardware found that so-called "enterprise" drives did not necessarily have better reliability than consumer-grade drives at the same capacity level:
http://www.tomshardware.com/reviews/ssd-reliability-failure-rate,2923-2.html
Plex sent out an email today to their PlexPass members which gives a 20% discount on the PR2100 and PR4100 at the Western Digital Store. I haven't found them discounted anywhere. Discount available for about 30 days. Last time they did this they went out of stock quickly.
Last edited:
purpletalon55
macrumors member
Seagate Ironwolf nas drives are a good option they are 5900rpm so faster than reds and come with a 3 year warranty I've had 4 of them in my Synology DS1515 for a while now with no hiccups.