Could hard drives get to big for a ZFS type filesystem

[leenux_tux]

A topic for consideration/discussion...for the home/small businesses user....

I am in the middle of a server rebuild at the moment, sourcing new hardware, cases, rack mount cabinet, more hard drives etc. I have also bought another IBM m1015 controller so now I have two to play with, however, a thought has occured to me whilst going through this exercise... Will hard drives get to big for a ZFS type filesystem ?

I currently have a mixture of 1TB and 2TB hard drives in my two pools. I have purchased a bunch more 2TB drives to extend one of my pools (currently 4x2TB, raidz1...bad choice but this procedure is the fix.)

Looking at the choice of hard drives out there now, 3TB, 4TB, 6TB, I'm sure I read 8TB is on the horizon? how much of a strain on hard drives is as ZFS rebuild? Replacing a 2TB drive in a raidz1 took over 5 hours per disk for me when I moved one of my pools from 4x1TB drives to 2TB. If you had a (for example) 80% full raidz2 containing 4x6TB drives and had to replace a drive, how long would it take and how much strain would it put on the other drives ?

I am wondering where the "sweet spot" is. Will it be that SSD drives become so cheap that we use multiples of these type of drives instead of small numbers of huge capacity drives ?

I hope this makes sense and invokes some healthy debate.

Leenux_tux

 

[cyberjock]

Hard drives won't get "too big" in your lifetime. Rebuilds may be a problem with RAIDZ3 if hard drive manufacturers don't get the reliability up though. But that's not a ZFS problem so much as a hardware problem.

 

[SirMaster]

You could always stick to pools of mirrors which is what I've seen many companies that are concerned with rebuild times on large arrays do.

Larger drives ARE getting faster. The new 6TB Seagate actually rebuilds faster than my 4TB WD Reds. We will have to wait to see how fast the new 8TB Seagates are.

 

[Pharfar]

https://blogs.oracle.com/bonwick/entry/128_bit_storage_are_you
No, ZFS isn't the limiting factor. Waiting on harddrives is the real killer.

 

[cyberjock]

Larger drives ARE getting faster. The new 6TB Seagate actually rebuilds faster than my 4TB WD Reds. We will have to wait to see how fast the new 8TB Seagates are.

Yes but they are getting faster at a MUCH slower rate than they are getting bigger. So reading sector 0 to sector "end of the disk" is taking more time than ever before. Back in the day my 20MB hard drive peaked at like 80KB/sec. You could still zero out the disk in like 2 minutes though. There is no chance on God's green Earth you could buy any TB-sized platter based hard drive on the market and zero out the disk in 20 minutes. ;)

 

[SirMaster]

That's not my experience exactly though. My 4TB drive takes 10 hours to read end-to-end (~110MB/s average speed) and my 6TB drive takes 9 hours (~185MB/s average speed). I've seen my 6TB drive read at 220MB/s even.

So I've actually seen a reduction in disk resilver time with this latest generation.

I don't see why an 8TB drive couldn't be read in 8-9 hours if the density is even greater.

 

[cyberjock]

Yep.. one data point. And if you took apart your 6TB drive you'd be disappointed to find that the improvement is only because there's more platters and more heads. I own a 6TB drive and they're very nice, but not the great thing you want them to be in performance.

Overall, as time marches on they are getting slower and slower. I guarantee you that you aren't comparing a 4TB and 6TB with the same number of platters and heads, so your comparison really isn't a fair comparison at all. If you look at the Helium filled drives they use the shingle technology, and those perform about on-par with 4TB disks. The shingle technology actually slows them down somewhat because the disk is overwriting data all over the place and has to correct for it. In those cases the 6TB drive is, on a MB/sec, about the same speed, but with 50% more disk capacity. Yikes!

 

[SirMaster]

Another one of your famous misinformed "guarantees" huh? Making guarantees on data you don't have is pretty foolish.

My 4TB RED has 4 1TB platters and my 6TB Seagate has 4 slightly short stroked 1.66TB platters if you really want to know so no, you are incorrect as they do have the same number of platters.

I can add a second data point at least and point out that my scrub speeds also doubled when I replaced a 6-disk vdev of 2TB disks with the 4TB Reds.

These 2TB disks were Samsung HD203UI which were Samsungs first 2TB disks and also used 4 platters.

These disks also took about the same time as the double capacity reds to resilver.

While I do expect the shingle technology to slow things down in Seagates first gen 8TB disks luckily shingles aren't the only new HDD tech that is going to bring us increased capacities and I expect greater performance from the others.

Only time will really tell though right now.

 

[cyberjock]

Your 6TB Seagate has 6 platters.. as noted on this review: http://www.storagereview.com/seagate_enterprise_capacity_6tb_35_hdd_review_v4

If you looked closely you'd have noticed that something is "not quite right" about the drive. I know, I own one. :)

You just don't get it.... a single data point, EVEN if you were correct, doesn't disprove the decades long trend in hard drives.

You know, I think I'm done talking to you. You seem to be so interested in arguing things instead of learning a thing or two I don't think I'm even going to bother responding to you anymore.

If your posts start being deleted, it's because I'm tired of you spreading incorrect information. Sorry but its a better use of my time to delete your posts than spend a day arguing with you over things that aren't up for debate. The facts are very clear.

 

[SirMaster]

Sorry, i'm not trying to argue.

I'm simply stating my experiences. In this case performance numbers from my hard disks.

If you want to delete my posts fine with me.

I only point out and post information of which I have first hand experience with.

For what it's worth in my experience at least from 2TB disks in 2009 to 6TB disks in 2014 a span of at least the past 6 years it has not gotten longer to read a top-capacity disk end-to-end. If the flatlined performance per capacity will stick around in in the continued future only time will really tell.

Just because in the the past the time to read a HDD has increased it doesn't automatically mean that that trend has to continue.

We will have to wait to see if the various models of 8TB disks really take longer to read than the 6TB disks of now and then keep going from there.

Thanks.