hard drives: WD Red vs WD Red PRO

[koifish59]

The only tangible difference I can see between these Western Digital NAS drives is that the standard Red spins at 5900k RPM, while the newer red PRO spins at 7200k RPM. I don't care for the added warranty on the pro series.

So this might be a silly question, but does disk performance matter for ZFS, since everything is cached in the memory? I plan on having 32gb of memory and 24tb of disks, so I don't imaging maxing out the cache and utilizing direct performance from the disks.

With that said, it would save me a good margin of money to get the standard red vs red Pro if disk performance does not matter much.

 

[depasseg]

My guess is that most folks will say the pro isn't worth it. In fact some will point to the higher power and heat and say it's actually worse than the standard reds.

 

[Robert Trevellyan]

Agreed.

does disk performance matter for ZFS, since everything is cached in the memory?

You can't necessarily expect that "everything is cached", but your network is more likely to be a bottleneck than your hard drives.

 

[DrKK]

I think it would be difficult to find any of the most active posters in here that would recommend the Red Pros. As has been said by the other guys, the RPM difference will literally be unnoticeable in the face of the other bottlenecks in the system, they use more energy, and they are much more expensive. You will literally just waste money if you buy them.

 

[Cybix]

One of the reasons I bought WD Red Pro over WD Red is that the drives in theory have a better chance of rebuilding the RAID if it does die. For me this is a pure backup system so I only have to recopy data back, but I am using 3 6TB drives in RAIDZ. (They do run hot compared to WD Red drives)

WD Red drives have a URE (unrecoverable read error) rate of <1 in 10^14 compared to WD Red Pro being <1 in 10^15

Specs for the WD Drives:

  WD Red Pro   http://www.wdc.com/wdproducts/library/SpecSheet/ENG/2879-800022.pdf   WD Red   http://www.wdc.com/wdproducts/library/SpecSheet/ENG/2879-800002.pdf

Although ZFS does a good job with solving error based problems before they become major, if a hard drive dies and needs replacing a rebuild is needed. From my understanding using larger hard drives have higher failure rates during rebuilds.

Hope this helps

 

[D G]

The Red Pros are also rated for higher densities of drives, but if you're not going over 6 drives, then it's not a concern.
Some people do like the appeal of the longer warranty though.

 

[jgreco]

A higher RPM drive will allow somewhat higher IOPS to and from the pool. For the average home user, this is probably a nonissue. However, in a busy environment where there's a lot of constant data access, the faster drives will definitely be a (smallish) win.

 

[Xelas]

Do the pros have shorter rebuild times? My concern is that the 4/6/8 TB Red drives can take days to rebuild, and thus have a huge window of time to blow up the remaining disks. I did do some testing when I put my system together 2 years ago, but I'm not sure that the rebuild times will be the same as when the disks were new, and the data were nicely contiguous as opposed to when the disks are a few years old and the data are heavily fragmented. I'm not willing to test with my real data at this point - restoring from online backups is a multiday (multiple week?) endeavor.
My understanding is the the pros are pretty much rebadged SE drives.
Its not the day to day performance so much as rebuilds that would concern me.

 

[Bidule0hm]

Even a full 8 TB drive will resilver in less than a day at 100 MB/s so it's not that bad. And this is a good reason to not use RAID-Z1 with drives this big.

 

[Xelas]

Even a full 8 TB drive will resilver in less than a day at 100 MB/s so it's not that bad. And this is a good reason to not use RAID-Z1 with drives this big.

That's not too bad! Thanks.

 

[jgreco]

Do the pros have shorter rebuild times? My concern is that the 4/6/8 TB Red drives can take days to rebuild, and thus have a huge window of time to blow up the remaining disks. I did do some testing when I put my system together 2 years ago, but I'm not sure that the rebuild times will be the same as when the disks were new, and the data were nicely contiguous as opposed to when the disks are a few years old and the data are heavily fragmented. I'm not willing to test with my real data at this point - restoring from online backups is a multiday (multiple week?) endeavor.
My understanding is the the pros are pretty much rebadged SE drives.
Its not the day to day performance so much as rebuilds that would concern me.

The pros might have shorter rebuild times simply because they have faster seek. Resilvering (and scrubbing) is a process of starting at the root BP and traversing all the data on the pool. For a pool with lots of small files or lots of fragmentation --- this implies a lot of seeking. You can use the scrub time as a loose indicator as to what the resilvering time might be.

A pool with lots of larger files that isn't heavily fragmented will resilver a lot faster than a heavily fragmented one.

 

[Cybix]

Xelas: On a rebuild it causes a massive spike with hard drive use, this amount of workload puts extra stress onto the drives and increases chances of failures to occur. As Bidule0hm says it should only take a day to restore at 100MB/s and this less time is a good thing.

For me: The extra URE takes the risk down by a lot, but in my case I should have said, RAIDZ1 is only used for the scrub so I can recover data and not lose 50% space like in a mirror. If I had a disk fail, I would not rebuild, but buy a new disk and make new pools and copy the data back over.

I completely agree with Bidule0hm: Relying on RAIDZ1 alone with large drives is a bad idea.

 

[Ericloewe]

WD Red Pro being <1 in 10^15

Wrong!

They're 10 in 10^15, otherwise known as 1 in 10^14.

Deceptive marketing at its best.

 

[Cybix]

They're 10 in 10^15, otherwise known as 1 in 10^14.

I stand corrected

 

[jgreco]

Wrong!

They're 10 in 10^15, otherwise known as 1 in 10^14.

Deceptive marketing at its best.

 

[cyberjock]

Also, those values are "worst case scenarios". A WD drive manufactured could be 10^17 while a WD Pro manufactured could be 10^16. Those values simply represent the "worst case" the manufacturer claims. Many here can argue for/against those numbers as being real-world (or made up). So YMMV.

I, personally, wouldn't buy them until the have several orders of magnitude improvement. At that point it is probably a safe bet that what you are buying is almost certainly improved.

 

[DrKK]

If you make a proper pool with WD Reds, choose the RAID properly, perform your smart tests, manually monitor the system, and so on, the odds that you would *EVER* lose a pool are so close to 0 that it doesn't even bear speaking about.

The main threat to almost everyone's pool is INCOMPETENCE and/or LAZINESS, not 10^15 vs 10^16 vs 10^17 uncorrectable error rates. "Awww, this guy dropped a pool, and he says he was doing everything right!" then we check on it, and he's running a 7-disk RAID-Z1, never has run a smart test, never set up the email correctly, and didn't even know that he was running degraded for the past 4 months.

:)

 

[jgreco]

The main threat to almost everyone's pool is INCOMPETENCE and/or LAZINESS, not 10^15 vs 10^16 vs 10^17 uncorrectable error rates. "Awww, this guy dropped a pool, and he says he was doing everything right!" then we check on it, and he's running a 7-disk RAID-Z1, never has run a smart test, never set up the email correctly, and didn't even know that he was running degraded for the past 4 months.

That's approximately correct. There are some other bonus point enhancers, such as building on a bad platform, failing to do proper burn-in testing, and using RMA'd drives. The whole idea of ZFS is that you need a trustable computing and storage platform, and then you need to make sure you're immediately made aware of any adverse changes in its status. All this crap we harp on about probably seems like paranoia to the newcomer, but is in fact the hard-won knowledge of long experience.

 

[Gen8 Runner]

In Germany there is just a difference of about 40€ (50$) per disk between red and red pro (6000gb).
But you get 5 and not just 2 years warranty. That is anyway a positive argument to me. For use and of course also for reselling it, when you replace the disks and want to sell them.

 

[jgreco]

The problem is that in three years, the cost for a disk has usually fallen through the floor. If you're paying top dollar for that drive today, it is likely to cost less than half that in three years.