Jump to content
  • 0

Do I start buying 8TB archive drives or not?


Chris Downs

Question

The time has come for more space in my Plex server...

At present, it has three 4TB drives (just consumer grade). I was going to stick to 4TB drives (or maybe the 6TB WD reds), but "upgrade" to NAS versions, but I've been wondering about the 8TB archive drives from Seagate. Has anyone been using them for a decent amount of time with Drivepool?
In theory they seem ideal for a Plex server, mostly used for reads.

 

They also seem very good value in cost/TB. I reckon I'm using space at a rate of about 3.5 to 4TB per year now. A 4TB drive is about £109 at present. The 8TB is £178, which is £89/year if I assume 4TB/yr. It's clearly better value.

However, I do wonder about whether the archive drives are suited to a system that can either run continuously, or be powered up and down fairly regularly (it varies but once a day is usual). 
I welcome thoughts and advice on this!

Link to comment
Share on other sites

Recommended Posts

  • 0

the write-penalty will hit you hard, once you want to write more than 50gb in one go to them.

the write performance crumbles to about 30mb/s or less.

so if you push big backups often, this gets annoying.

 

apart from that, i guess you will only notice it when initially filling them or while balancing big amounts

Link to comment
Share on other sites

  • 0

Actually, there is a non-SMR cache of 20GB, so writing more might incur a write penalty. However, that only goes when areas of the HDD are re-written. If it is just write-once and no deletions then it should be fine AFAICS.

 

As an example, a recent Server Backup wrote 1.53TB in 3:34 hrs. That is, OTOH, about 120MB/s. I have yet to come across a use-case where it would hit you noticable, let alone "hard". In what circumstances have you experienced that?

Link to comment
Share on other sites

  • 0

I've actually seen where the writes drop off. They hit 0 bytes/sec for a few seconds and then bounce right back up. And do this repeatedly.  So there is definitely a write issue with the drives.

 

That said, I get ~180-190MB/s read from these drives when I switched the allocation unit size to 64k.  Worth doing (manually). 

 

 

And if write performance is an issue, then get a couple of SSDs and use the SSD Optimizer balancer. This way, data is written to the SSDs and then moved off. You'll never see the issue then. ;)

 

 

(I'm up to 5x 8TB Seagate Archive driver and am very happy with them)

Link to comment
Share on other sites

  • 0

 In what circumstances have you experienced that?

Pretty much every time i wrote more than 50gb. I repeated this multiple times on 2 different drives with the same results.

 

I've actually seen where the writes drop off. They hit 0 bytes/sec for a few seconds and then bounce right back up. And do this repeatedly.  So there is definitely a write issue with the drives.

This is exactly what i have seen which averaged to a little lower than 30mb/s if i recall correctly.

Link to comment
Share on other sites

  • 0

Pretty much every time i wrote more than 50gb. I repeated this multiple times on 2 different drives with the same results.

 

This is exactly what i have seen which averaged to a little lower than 30mb/s if i recall correctly.

 

I didn't really do a lot of testing. I was more concerned with reads than writes.  Namely because I knew the writes were probably going to be crap to start off with.  

 

Basically, if it's really an issue, use the SSD Optimizer Balancer Plugin. Write to a copy SSDs and let StableBit DrivePool handle the slow write speeds of these drives. :)

Link to comment
Share on other sites

  • 0

Pretty much every time i wrote more than 50gb. I repeated this multiple times on 2 different drives with the same results.

 

This is exactly what i have seen which averaged to a little lower than 30mb/s if i recall correctly.

Oh, I know that. I was more wondering about when do you, as a user, have I/O in excess of 50GB that you are actually waiting for to complete? I mean, I use them as Server Backup HDDs and I am certain I get that write penalty over and over but I don;t care because the backup process is automated and does not affect my user experience (restoring is different of course). Strangely enough, these backups run way faster on the 8TB HDD then on a 4TB WD Red...

Link to comment
Share on other sites

  • 0

As an example, a recent Server Backup wrote 1.53TB in 3:34 hrs. That is, OTOH, about 120MB/s.

I just realised this is simply false. The backup "transferred" 1.53TB but that does not mean that amount of data was written to the HDD, that would have been far far (FAR!) less. Sry.

 

I still agree with everything else I wrote here.

 

The SSD optimizer plug-in would indeed help should you encounter write-performance issues. Of course, only until writes exceeds the size of the SSDs + HDD cache and PMR cache. But I am pretty sure you could use a 4TB HDD you already own as cache as well with that plug-in. That might actually be the best setup of all.

Link to comment
Share on other sites

  • 0

The SSD optimizer plug-in would indeed help should you encounter write-performance issues. Of course, only until writes exceeds the size of the SSDs + HDD cache and PMR cache. But I am pretty sure you could use a 4TB HDD you already own as cache as well with that plug-in. That might actually be the best setup of all.

You could absolutely use a couple of HDDs for this.  And for the most part, that would be fine.  Over the network (if that's you're use case) may be the limiting factor anyways.

 

But yes, small SSDs may get filled rapidly, depending on how much data you're dumping onto the pool, and depending on your balancing setting.s

Link to comment
Share on other sites

  • 0

Oh, I know that. I was more wondering about when do you, as a user, have I/O in excess of 50GB that you are actually waiting for to complete?

 

One such Occasion was backing up 5 VMs from a laptop.

"Throw SSDs at the problem" can't be the answer :)

 

I was expecting bad write rates at some point, just not THIS bad.

For now, i have replaced the drives.

Link to comment
Share on other sites

  • 0

One such Occasion was backing up 5 VMs from a laptop.

"Throw SSDs at the problem" can't be the answer :)

I don't know. It's a pretty good solution for most issues. :P

 

 

I was expecting bad write rates at some point, just not THIS bad.

For now, i have replaced the drives.

 

Yeah, it definitely can be. It's not a great backup drive.  And yeah, the writes are pretty bad overall.  

 

But it's a great storage drive (Archive drive, even! :P) and it would make a great "sneaker-net" drive, as well). 

 

But yes, if you do plan on using these drives and writing a lot of data in the pool, then grabbing a smaller, faster drive for write caching would be a fantastic idea, and HIGHLY recommended (if not needed outright). 

Link to comment
Share on other sites

  • 0

To clarify here (and mark as answered):

 

The Seagate Archive drives get great read speeds. But beacuse of how they work, they get atrocious write speeds.  

 

They're great for long term storage, "sneaker nets" and for the pool.  

 

But if you do use them with the Pool, you don't want them to be written to first. You will want to use the SSD Optimizer to write to other drives first. Be it SSDs or spinning HDDs.  

 

This way, you get good write speeds to the pool, and StableBit DrivePool will handle the horrible writes. And considering that StableBit DrivePool uses a background priority for file transfers, and it runs in the background, invisibly, this may work out VERY well. 

 

 

You can get the SSD Optimizer Balancer here:

https://stablebit.com/DrivePool/Plugins

 

 

 

 

As for performance:

For reads, I got about 160-190MB/s from the drives, consistently. (using large, sequential files, and using 64k allocation unit sizes when formatting the disk).

For writes, closer to 80MB/s but would stall frequently during writing, decreasing the overall speed.

Link to comment
Share on other sites

  • 0

I just read through this, as I too will probably be looking into getting one of these come tax season.  However, I don't feel like the OP's question was answered, or even touched upon by the responses.  Still, this is a very informative thread. 

 

So, to re-ask the OP's question...  How well do these drives operate in a server environment that's not on 24/7, with occasional power cycles or sleep times?  Do the constant power on spinups, etc have any major detrimental effects on the performance or longevity of the drives?

Link to comment
Share on other sites

  • 0

I just read through this, as I too will probably be looking into getting one of these come tax season.  However, I don't feel like the OP's question was answered, or even touched upon by the responses.  Still, this is a very informative thread. 

 

So, to re-ask the OP's question...  How well do these drives operate in a server environment that's not on 24/7, with occasional power cycles or sleep times?  Do the constant power on spinups, etc have any major detrimental effects on the performance or longevity of the drives?

Well, which part?

 

Aside from writes, the drive is just like any other disk.  It should respond just fine, aside from the mentioned write issues.

 

And yes, you should be able to use it continually or to turn it off frequently. Both without issues.

 

The best comparison is the MTBF and warranty:

http://www.wdc.com/wdproducts/library/SpecSheet/ENG/2879-800002.pdf

http://www.seagate.com/www-content/product-content/hdd-fam/seagate-archive-hdd/en-us/docs/archive-hdd-dS1834-3-1411us.pdf

 

They're not as "good" as NAS drives, but they should be fine with always on functionality.

Link to comment
Share on other sites

  • 0

I would say yes. I can not find a "normal" 6TB HDD that is cheaper than these 8TB HDDs and they read like crazy (for spinners, certainly at 5.9Krpm). And if you write about 1 movie a day (assuming they are less than say 25GB) then you won't even suffer a write-penalty). Or per hour. Or per ten minutes I would think. And if you have a number of them in a Pool and don't use file-placement rules it gets even better still I would speculate.

Link to comment
Share on other sites

  • 0

I wonder if with SSD pluggin acting as write cache they couldn't be use for more general purpose like storing family pictures/movies that can be sometimes deleted, created.

My actual setup use 4x4TB WD Red + 1 4TB Green and 3x240GB SSD as cache. The more I think to it the more I wonder if its not a waste of money and power draw.

Maybe with 3x8TB archive drives + 3x240GB SSD I wouldn't see any difference with my actual setup... :huh:

Link to comment
Share on other sites

  • 0

So can we asume that this disk is a good hardware choice for movie libraries?

Movies are 99% of the time written once then only read from time to time.

Definitely. 

 

I would say yes. I can not find a "normal" 6TB HDD that is cheaper than these 8TB HDDs and they read like crazy (for spinners, certainly at 5.9Krpm). And if you write about 1 movie a day (assuming they are less than say 25GB) then you won't even suffer a write-penalty). Or per hour. Or per ten minutes I would think. And if you have a number of them in a Pool and don't use file-placement rules it gets even better still I would speculate.

And yea, if it's rather small-ish, then you shouldn't see the write issue at all.  I did, because I was copying a large number of large files. 

 

I wonder if with SSD pluggin acting as write cache they couldn't be use for more general purpose like storing family pictures/movies that can be sometimes deleted, created.

My actual setup use 4x4TB WD Red + 1 4TB Green and 3x240GB SSD as cache. The more I think to it the more I wonder if its not a waste of money and power draw.

Maybe with 3x8TB archive drives + 3x240GB SSD I wouldn't see any difference with my actual setup... :huh:

Absolutely. That's exactly what I do, and I see great performance from my pool.  Though, I'm using 120GB SSDs (I had them already). 

 

As for power draw, I'm not sure about in usage, but peak power draw is a bit higher for the Seagate Archive drives.   Your current sent is probably as power effecient as swapping the drives out of the Archive drives.   But I'm not 100% sure.

Link to comment
Share on other sites

  • 0

Yes Red are very power efficient, maybe the best of all HDD but if I use half number of HDs...

RED 4TB :

Peak : 1.75A

Read/Write : 4.5W

Idle : 3.3W

Sleep : 0.4W

 

Seagate Archive 8TB :

Peak : 2A

Read/Write : 7.5W, normalized to 4TB = 3.75W

Idle : 5W, normalized to 4TB = 2.5W

Sleep : 0.9W, normalized to 4TB = 0.45W

 

4 RED : 7A, 18W, 13.2W, 1.6W

2 Seagate : 4A, 15W, 10W, 1.8W

 

I know things are more complex (1 Red read/write will be better than 1 Seagate archive read/write) but....

 

 

Edit : @Christopher talking about SSD plugin by any chance, did you find the time to make it not using always the same SSD when multiple SSDs are used?

Maybe my solution is to simple to work but if the plugin could just randomly choose one of the SSDs this would help to reduce wear occuring always on the same SSD ;-)

Link to comment
Share on other sites

  • 0

You have triple duplication? Wow...

 

Anyway: "general purpose like storing family pictures/movies that can be sometimes deleted, created."... but pictures will, I assume, be "small"? In the MBs, not tens of them? Movies, say a decent .mkv rip might be what, 4GB at most? Writing those occasionally would not cause the write penalty. You would need to write at east 20GB in one go before you might experience it and I *think* that the PMR-cache gets written to the SMR-arrea pretty quickly once there is no I/O.

 

Wrt. power consumption, those are rated numbers and given how close they are, actual measurement would be nice. But yes, if you have a lot of writes then you'd need a more difficult statistic as the Archive will suffer from a kind of. I guess we can call it, write-amplification indeed.

Link to comment
Share on other sites

  • 0

I have quadruple duplication for my Folder Redirection folder (the user folders for domain user accounts). :)

 

But that's about 40-50GBs. 

 

And I have everything duplicated personally.

 

 

As for exact power consumption, well, grab a kill-a-watt meter and compare the wattage before and after powering on the drive.  If you do the math, you can identify the estimated usage (you may want to disable stuff like speedstep, or change the power management options to disable it, so you can get an accurate reading).

 

 

Edit : @Christopher talking about SSD plugin by any chance, did you find the time to make it not using always the same SSD when multiple SSDs are used?

Maybe my solution is to simple to work but if the plugin could just randomly choose one of the SSDs this would help to reduce wear occuring always on the same SSD ;-)

Did you mean this issue:

http://community.covecube.com/index.php?/topic/1523-ssd-usage-with-more-than-one-ssd-as-cache/&do=findComment&comment=10689

 

 

If so, then no, not yet. Alex has been working on CloudDrive and bug fixes, so we haven't gotten to a lot of the non-critical issues yet. We know it's been ... well, too long, and we want to get to all (or most) of the pending issues soon. But there is only so much that we can do.

 

 

As for the specific issue... since I have taken a fresh look at the issue, it may just be a lack of randomization. If there are three identical drives, I suspect that the driver will always use the first disk enumerated. (by chance is the SSD in question, the lowest numbered one?).  However, it may be possible that the System Volume Information folder is taking up more room on one disk than other, and causing it the algorithm for selecting the disks to pick that specific disk over the others.

Link to comment
Share on other sites

  • 0

I use two of the Seagate 8TB archive drives for an unduplicated backup pool from my media server.  99% of the data is write once and I will saturate my network link.  I have my pool set up to keep free space the same on both drives so usually it will alternate writes between the two drives giving them more opportunity to do their background processing and rewrite shingles if needed.  I didn't do a full disk check or write and read zeros to the drives before I put them in use.  I heard this may make the drives consider any future write as a rewrite and make the speeds slower.  For media or backup they work perfect.

 

I run a robocopy mirror script from the network shares to the pool nightly and so far it's the easiest backup solution I have found.

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Answer this question...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Loading...
×
×
  • Create New...