Seagate 8TB Compatibility

Butcher9_9

Limp Gawd
Joined
Mar 8, 2010
Messages
146
I'm thinking about getting some Seagate 8TB Archive drives (ST8000AS0002) but i cant seem to find any information about compatibility with Controller Cards.

Has anyone had any experience with these drives and raid cards?

I have an Adaptec 51645 so that's the card I'm most interested in (Adaptec in general would help) but I'm sure it would be useful for anyone out there if people could list their experiences with these drives in general.
 
i don;t think any of the archive series is in wide availability for a body of experience to exist yet.
 
They have just arrived in Australia so I was hoping the US would have had them for a while before that, I guess not.
 
Beware that they are 4k native drives rather that 512 or 512E drives.
And shingled, so write performance will suffer if the use scenario is not SMR-aware.
EDIT: The specs I found state it is 512e.
 
I'm not sure if my controller supports 4k native drives, there are no Whitepapers on the 5 Series and 4K support but looking at the other white papers the bigger issues with 4K is booting not array creation and none on the arrays on my card are boot drives.

Regarding Shingle drive, My Arrays are for video media storage and I only write about 1TB and Read 2TB per month to all 20 drives and each drive is rated are 180TB per year (15TB per Month) so fingers crossed its not an issue.

I'm hoping that since these are technically enterprise level drives they should have OK Raid support even if the duty cycle is very low. But then again you never really know with Controllers what drives they might take a hating to.
 
Once controllers got passed the 2TB roadblock, is there a difference between 3,4,6,8 TB?

Is there a built-in limitiation like LBA that would keep them from working with any drives in the foreseeable future?
 
Do any current controllers that support > 2TB have limits way before the 48-bit LBA limit? Some of the early 48-bit LBA controllers wouldn't allow more than 750GB or 1TB.
 
There is a limit imposed by 32-bit addressing of 4096 KiB sectors, which would be at 16 GiB/17.6 GB.
I'm not sure if there is really any hardware that will suffer from that, but in theory some external controllers that use 4K native addressing could be affected.
 
Well I can confirm that the Seagate 8TB drive work with adapter 51645 cards. The array is up and running and and been transferring data for the past 36 hours.

The drives are very heavy and the platters take up almost the complete thickness of the drive. The PCB is flush with the bottom as well is the rest of the drive no recessing as is normal.

I have done a bit of testing, details below.

HDtune Read.
yIIOm3V.jpg


HDtune write. Clearly the drive can trick HDtune.
NCmSuvX.jpg


HDtune access times
lE7WimF.jpg


I tried to use the file benchmark to see if that would show the real performance but I guess not. maybe its because its the first time data has been written the the drive (no read/rewrite required) but i could not be bothered to wait the 3 days its takes to fill the drive to test it further.
GjtNL3l.jpg


Built an array, my controller is a bit of a bottleneck for the reads. (6X8TB Raid 5)
O9EbsnA.png


Then I started a 25TB transfer to the array to properly test the writes.

This is a snippet to show the highs and lows of the transfer.
3QAT6MC.jpg


8 hours + of transfer data
o5zwTSb.jpg


Average is 130MB/s from the array so sub 30MB/s per drive (excluding the parity) not great but OK. If I upgrade the array to 8 drive s in the future I might get 180MB/s out of it. The average has also drops a bit since I finished this record (around 120MB/s now) Ill post the longer record once the transfer is close being finished.
 
Sounds amazing, I think this technology was practically made for ZFS scalabilty. 20 of these in a pool and you would get large space and great performance, 116 TB usable with 4 drives of parity :D
 
I have 10 of these in a ZFS RAID-Z2 pool off of an Intel RES2SV240 SAS Expander from a LSI 9201-16e. Works as expected.
 
Sounds amazing, I think this technology was practically made for ZFS scalabilty. 20 of these in a pool and you would get large space and great performance, 116 TB usable with 4 drives of parity :D

Um.. no. These disks are specialized in write-infrequently, read-often. So they're actually *not* ideal for striping based RAID due to the combination of write penalty inherent with striping RAID (including raidz/z2/z3)) and the slow write speed of SMR (40-50MB/s is typical).

What they are ideal for is JBOD-based/non-striping parity, such as SnapRAID, where SMR drives like these can be data disks that are separate from dedicated (and faster) parity disks. This is why Seagate recommends "single disk applications" for these. What they really mean is non-striping.
 
Last edited:
Well more info

48 Hours of transfer stats
8gwV12w.jpg


The write speed does not seem to drop off towards the end of the disk ever much which points to other bottlenecks.

Um.. no. These disks are specialized in write-infrequently, read-often. So they're actually *not* ideal for striping based RAID due to the combination of write penalty inherent with striping RAID (including raidz/z2/z3)) and the slow write speed of SMR (40-50MB/s is typical).

40-50MB/s if you are lucky.

What they are good for is home user style parity raid where most people would use them for media storage which is written infrequently and long rebuild times are a non issue.

Rebuilding or coping over large amounts of data are a bitch but (50+ hours for my 25TB transfer) but that does not happen very often and i have backups worst case.
 
I would like to know if these can be used in a mineral oil build. Most drives can't unless they are SSD's. The only mechanical drives I know of that are sealed and can be submerged in mineral oil are the Helium filled drives which are way too expensive so at $300 for an 8TB drive, I could finally build my mineral oil PC, that is if these are submersible.
 
I would like to know if these can be used in a mineral oil build. Most drives can't unless they are SSD's. The only mechanical drives I know of that are sealed and can be submerged in mineral oil are the Helium filled drives which are way too expensive so at $300 for an 8TB drive, I could finally build my mineral oil PC, that is if these are submersible.

I'm pretty sure these are open air drive so to speak.

If you really want to use normal HDDs you could build a water tight HDD enclosure and submerge that. It would be hard to do but then again oil PCs are a pain any way so whats a bit more hassle I guess.
 
I wonder how much you're being held back by that old 5-series card? Might have to see if we can order some of the 10TB SMR drives for... testing... and see how they behave on newer controllers (8-series and comparable LSIs, can never remember their models)

BTW, should've bought some of these drives. :p ;) Sounds like we could submerse them in mineral oil too!
 
(BTW, after watching the video series posted by WestSidaz, I thoroughly enjoyed watching Linus and Luke shock the heck out of each other... good times. :D )
 
I wonder how much you're being held back by that old 5-series card?

I have a feeling something in my system is holding the card back, though I m not sure what. I used to get 600MB/s off some of my arrays but now its 500MB/s . Still with these drives for reads the I would assume that an array of 6 in raid 5 should hit 700MB/s easily (based on double the minimum speed of the array, a single drive follows that curve as seen above) with a good controller + about 140MB/s per drive after that. Once Win10 comes out i'm formatting my PC and maybe getting a new Mobo/CPU/RAM so ill see if that increases the performance.

For writes its a drive limitation not controller. The drives handle all of the internal data movement and are just requesting data at a very low/random rate. I doubt that a better controller would give you a write speed increase.


I thought the whole point of oil cooling was no fans ect. Yet that build has loads.
 
Nah, the oil builds are only good for very low wattage systems and with minor current applied. They are using a full on X99 system so the oil build is mostly novelty here.
 
There is a transfer from a 8X3TB to a 8X4TB array on the same card for reference. The card is a bottleneck for this transfer it seems (no drop off as the disks fill)

But still that's 450+ MB/s both read and write for more than 900MB/s total is not terrible.

DAfcumC.jpg
 
Back
Top