Decasing Seagate 8TB "Archive" (and shingled drives in NAS's??)

[Twice_Shy]

Hello peoples, you probably havent seen me for about six months because i've been completely swamped with RL requirements and had no money to spend on even the greatest deals until certain other problems were solved first... well i'm finally in film school proper and so now it's back to what I originally came here for.


My question of the moment concerns using shingled drives in general within a NAS, and specifically the Seagate 8TB external USB Archive drives going for like $170... trying to do some research but I don't fully understand what i'm reading or why things are a given way: Seagate Archive HDD Review (8TB) | StorageReview.com - Storage Reviews Seagate Archive HDD Review (8TB) Discussion

The NAS I have planned would be mixing shingled and nonshingled drives, I plan to use the nonshingled older 3tb drives for things which may involve a bit of drive thrashing or smaller files. I'd like to use the shingled drives for primarily read only material and backed up monolithic data. (think a 30gig virtual machine or a 40gig video file instead of things like pictures or documents) Or to write in a single stripe a temporary backup from the nonshingled drives once in awhile. It's not clear to me how the drive would perform under these conditions. I can understand rewriting being problematic, but was wondering how a freshly formatted drive should perform with sustained large file writes. It's not clear if that was accounted for in the tests or/and i'm confused by a backup going at 30MB/sec while sustained 128k blocks are apparently near 200MB/sec...

I'm also wondering about which specific drive model or models are found inside Seagate's external 8TB case here, and whether any reliability statistics are apparent from being around for 2+ years by now. (similar to how Backblaze had to stop using a certain 3tb seagate due to a 40% annual failure rate one year, i'm curious how these are looking and where I could find that out)

Finally if these drives are otherwise usable, is there a best guide for how to decase them? (and preferably reuse the remaining USB adapter afterwards - like could I simply throw an old 250gig SATA drive in there and use it as a spare external?)

--
PS if it matters, I plan to use SnapRAID for the time being for if the NAS ever needs data restored to something.

 

[BlueFox]

Why don't you just get the WD 8TB externals from Best Buy instead? No dealing with SMR then. They're regularly on sale for ~$180 (like now). See here: [DEAD] WD reds 4/8TB $90/$200 via shucking - update, 8TB reds now $180, 4TB deal looks dead.

 

[Twice_Shy]

Wasn't aware they were available for the same price, I thought nearly everything that size and price was going to be shingled.

So yes those will be an option next time they come up again - basically any drive 6TB and up that is winning the cost per gigabyte battle.

That said, still interested in all related information including if I were to end up with shingled drives and how to decase them, and the reliability of different models I might find inside those cases. I need to know if certain ones are "must avoids" or certain ones are diamonds in the rough really worth getting. Put another way, what information do I need to know what to get/be in the buying pool for the next deals that show up and how can I make best use of even crappy shingled archival drives?

 

[Terry Kennedy]

That said, still interested in all related information including if I were to end up with shingled drives and how to decase them, and the reliability of different models I might find inside those cases. I need to know if certain ones are "must avoids" or certain ones are diamonds in the rough really worth getting. Put another way, what information do I need to know what to get/be in the buying pool for the next deals that show up and how can I make best use of even crappy shingled archival drives?

You want a filesystem that is aware of shingled drives, otherwise the filesystem will try writing its metadata on various rows of shingles and performance will be even worse than you expect. There are 2 types of shingled drives, ones that require host assistance and ones that do it all in the drive - sort of like 4Kn vs. 512e drives. All of the shingled drives I know of (not that I've looked lately) are the kind that require host assistance.

There has been at least some support for SMR drives in FreeBSD since r300207.

 

[Twice_Shy]

Does Windows 10 have awareness? I was just planning on running SnapRAID. I'm still trying to understand under what conditions those drives run poorly vs acting pretty normal.

Even if I will have shingled drives I don't plan to ONLY have shingled drives, think more of that being a media fileserver side (primarily read oriented) or same-system backup disk for more active drives on which work is done. To me that's a very archival like use not much different than if I left it plugged into USB to back up files already on primary drives I mean. I'm aware of RAID rebuilds taking longer but i'm not planning to use them for RAID or to work the drives all that heavily. Would my use case still make them problematic?

 

[ecosse]

I haven't used shingled drives in my main media repository although I have a few in my smaller file server (Windows 2012). I don't use them in a snapraid config, but if I did I wouldn't use them as a parity drive of course. I find the performance acceptable for my use case -which is a basic file repository.

Always interesting to understand what poor performance actually means e.g. do you want to stream UHD videos?

 

[Twice_Shy]

I haven't used shingled drives in my main media repository although I have a few in my smaller file server (Windows 2012). I don't use them in a snapraid config, but if I did I wouldn't use them as a parity drive of course. I find the performance acceptable for my use case -which is a basic file repository.

Always interesting to understand what poor performance actually means e.g. do you want to stream UHD videos?

Well my biggest curiosity is to get a better clearer picture of what kind of sustained writes are possible. Like in some tests it seems to be nonstop 200MB/sec. Others go like 30MB/sec for awhile, then drop to 1MB/sec, then back up to 30MB, then down to 1MB. If it's doing that, is there some nonstop level, maybe 7MB or 10MB or something that it could just keep up with endlessly or does it have no choice but to "take bites" and do this juggling act to most efficiently write the data? (ie binge away and let the drive just handle it)

I am curious if any shingled drive could maintain a nonstop 55MB/sec reading level for instance for well organized data - perhaps the minimum LTO6 writing could continue at without shoeshining the tape to death.

I plan to use SnapRAID but not necessarily on every drive - I might do a simple manual backup/mirror or I think MD-raid it's called under linux/software mirror on shingled drives - or not mirror them at all.

Poor performance at this point is not even able to keep up with 25mbit writing rates nonstop (3MB/sec) from the internet despite being chunked up data (usenet, torrent, www mirroring, anything challenging). If it's going to choke on even that rate, that's pretty bad.

Reading needs pretty minimal - my sad poor man's home theater is still at 720p right now for the next few years. But my drive uses are still less about streaming movies (that's also a use, but a lower importance use) and more about storing personally shot video files right now, and even if some of that may be 4k HDR or even more - it's a backup NAS not a performance NAS. It's for work file overflow - migrate a project off for a month, and when I queue it to copy I can go get a pizza - I don't care if it takes 5min or 50min. I just need to cheaply store the data until I move it back to a workstation HD and SSD where it will all be hit locally.

 

[Evan]

Unless you have a bunch of SMR drives already in your possession then I wound never think the extra 2tb space is worth it over the PMR drives even if they cost a few $$ more.
Just not worth the hastle I think.

 

[ecosse]

Well my biggest curiosity is to get a better clearer picture of what kind of sustained writes are possible. Like in some tests it seems to be nonstop 200MB/sec. Others go like 30MB/sec for awhile, then drop to 1MB/sec, then back up to 30MB, then down to 1MB. If it's doing that, is there some nonstop level, maybe 7MB or 10MB or something that it could just keep up with endlessly or does it have no choice but to "take bites" and do this juggling act to most efficiently write the data? (ie binge away and let the drive just handle it)

So I've been watching a very unscientific test after I'd freed some space up on a singled drive. A bunch of small PDF files were doing between 10-40MB/s - as you say fluctuating at times between the two. Some larger video MKV files were doing 35-70MB/s. I'm in the same camp as you - for this kind of repository I don't really care too much how fast the response is as long as its within a "reasonable" boundary and for the most part my data is on these devices are static.
I personally wouldn't put these in the primary internet downloading space. I'd use a cheap pair of Hitachi 3TB drives and use the shingled as a backend. Very unscientific again tho.

 

[fractal]

My experience with shingled drives is - they read fine but they write slower than cold molasses at the south pole in winter.

I would not hesitate to use them for bulk storage on a media server which is for all intents and purposes write once - read many (or never if you are like most of us data hoarders).

I would never use one for content that is going to change.

 

[manfri]

Keep in mind that these Seagate Archive have some of drives not shingled, that are used as buffer, especially for writes, and in the background move'm to the shingled zone of drive.
If your writes do not overwhelm the buffer no big deal, but if you do the performance drop dramatically.
given the price difference i vold not eventhing about using SHR drives in a raid... let alone mix'em with PMR.
Maybe in some fringe kind of application, but the fact that these are developed for write-once no raid applied application mean, for me, use for this kind of application and forget about RAID.