Advice for using a pair of 8TB Archive Drives in a Synology Hybrid Raid

harryoui

Reputable
Jun 25, 2015
93
0
4,660
I'm about to purchase a pair of 8TB Archive Drives that I can use for raid in a XPEnology Plex Media Server. These drives should be fine for this purpose (despite labelled as not good for raid) because they will not be writing very often, and when doing so will be in fairly steady blocks.

Does anybody have any advice/tips for maintaining decent speeds/long life with these drives?

For example, I have read that it is best to have a separate non-archive hard drive to download torrents to, then have them copy across to the archive drives to prevent random write or something rather
 
Solution
Ah - I guess it does have an 'on drive' cache. That is stored in the data area itself, and I would imagine then that it (the drive controller and firmware) manages that to ensure that the cache is correctly written to the drive and accounted for. They (Seagate) specifically do NOT recommend using it in raid mode though because the sustained write throughput is rather abysmal - rebuilding an 8TB mirror took 19 hours for a WD Helium set, and 57 hours for the Seagate SMR. That's not good. I rebuilt an 8TB RAID5 set in 30 hours...

If you're only using two of them (as in 2x8TB RAID1), then they should be ok. Just be aware of the potential rebuild time if you do get a failure, and I would not recommend their use in a RAID5 as rebuilding...

JustSomeJoe

Distinguished
Oct 8, 2008
489
0
19,060
Archive Drives are quick to read and write but painfully slow to re-write, this could cause big problems for RAID (depending on what RAID); best option would be to contact the Server/Device manufacturers and ask.
 

harryoui

Reputable
Jun 25, 2015
93
0
4,660


I've read a lot of people's experiences using these drives in RAID and they all recount that they've had no problems for Movie streaming/storage.
Just curious if there are any solutions that people have to problems that could arise
e.g. extended stress on a new drive during rebuild period
 

Rookie_MIB

Distinguished
The problem with using these drives in RAID is that they can expose your data to corruption in the event of a power loss during write. If these are the SMR based drives (which I believe they might be) then the problem is that they pack data tighter by layering the writes. Thus, writing one block of data means that it has to read adjacent layers, then re-write the tracks down layering the new data in the correct 'shingle'.

Here's an article describing it better:
http://www.storagereview.com/what_is_shingled_magnetic_recording_smr

The problem is if you lose power during a write, the data (and other data in the layers around it) can be lost because you're not just doing a quick write to a single sector, you're writing MULTIPLE sectors, exposing the data in transit for a much longer time frame. In a RAID situation, this can make things worse as if you were to be using (for example) RAID5 while having two sectors on two different drives go bad, you've just lost your entire array. In a RAID0 stripe, same problem. In a RAID1 mirror, it's less of a problem as data one one drive doesn't rely on data from another drive. But the extended write times still provide potential problems.

In many ways, it's much like garbage collection on the SSDs, but SSD's are much faster (less exposure time) and usually have backup capacitors to protect data in flight.
 

harryoui

Reputable
Jun 25, 2015
93
0
4,660


Ah, alright. I hadn't thought of the possibility of data corruption if my NAS was turned off mid write, but I'm pretty sure that that data is stored in the cache. So then when the hard drive is turned back on it will display as corrupted for a short while until it finishes copying it back over from the cache; so I don't think this will be a large problem
 

PhysX_HW

Distinguished
They should be fine, and you are quite right about using a single drive as a buffer for downloads, because it's especially important if you are using HDD with SMR to store the data afterwards. So, you can definitely go for 2 of these and then a 1tb buffer which should be something faster, like a WD Black to have fast random read/write times if you are seeding multiple torrents and downloading at once. You can use a 2Tb drive, it's up to you how much space you want to have as your buffer. Also, if you are worried about data corruption in the event of a power failure, you can buy a UPS to protect your server, but I don't think it would be a problem, since you would be copying from one drive to another and if the file can't be written out from the buffer to the array, you'd still have it on the buffer and you can start copying again. So I wouldn't worry much about data loss.
 

harryoui

Reputable
Jun 25, 2015
93
0
4,660


Some delicious 20GB worth of volatile Cache? I've seen a couple sources saying that it's persistent memory
 

Rookie_MIB

Distinguished


That generally is NOT the case. Here's a great article:

https://www.thomas-krenn.com/en/wiki/RAID_Controller_and_Hard_Disk_Cache_Settings

The short of it is:

If the drives are connected to a RAID card which has battery backup AND the HDD cache is disabled, data is safe.

If the drives are connected to a RAID card and there is no battery backup, data can be lost.

If the drives are connected via SATA (ie: regular onboard connectors) and the HDD cache is disabled, data is safe

If the drives are connected via SATA and the HDD cache is enabled, data can be lost.

The clear point of it is simple enough - in a SATA based RAID card, with the HDD cache enabled, if there is a power loss you can lose data. If there is corrupted data in the RAID set, then the RAID could fail and ALL data could be lost.
 

harryoui

Reputable
Jun 25, 2015
93
0
4,660


I completely understand that the loss of power to a hard drive would remove the cache, but the hard drive has a 128MB cache which isn't that unique, and a 20GB* 'on-drive' cache (maybe buffer is a better word?), which I'm not sure if it's stored on a platter or somewhere else, but I am fairly certain that it is not volatile storage (meaning that it won't wipe without power). It is also uses this 20GB section for meta data tables and some background processes e.g. garbage collection


http://www.storagereview.com/seagate_archive_hdd_review_8tb
 

Rookie_MIB

Distinguished
Ah - I guess it does have an 'on drive' cache. That is stored in the data area itself, and I would imagine then that it (the drive controller and firmware) manages that to ensure that the cache is correctly written to the drive and accounted for. They (Seagate) specifically do NOT recommend using it in raid mode though because the sustained write throughput is rather abysmal - rebuilding an 8TB mirror took 19 hours for a WD Helium set, and 57 hours for the Seagate SMR. That's not good. I rebuilt an 8TB RAID5 set in 30 hours...

If you're only using two of them (as in 2x8TB RAID1), then they should be ok. Just be aware of the potential rebuild time if you do get a failure, and I would not recommend their use in a RAID5 as rebuilding even a 3x8TB 16TB set would probably take a week with these drives. Let alone something crazy like a 5x8TB 32TB set. I shudder at how long THAT would require.

Good luck with it. BTW have you thought of just buying the regular 8TB Seagate drives? They're only marginally more expensive, $40 more per drive but I believe they won't have the write penalty.
 
Solution

harryoui

Reputable
Jun 25, 2015
93
0
4,660


Judging from the two stores that are nearish to me, buying a 8tb archive drive instead of a regular 8tb drive would save me between 79$-174$ (mwave.com.au and kogan.com.au)

If one of them breaks, it may be a wise idea to replace it with a regular NAS drive to save it the stress of such a long rebuild. Although, a bit of downtown isn't the most end-of-the-world thing in this situation