8TB Drive Buying Decisions

[Mirabis]

Hello,

I see a lot of people buying refurbs/used disks from eBay and recently started doing the same, waaaay cheaper.
But warranty is sometimes out of the question (oem, outside warranty period etc)... which makes me unsure about my purchase decisions.

So, right now I'm swapping out my 3 & 4 TB disks for 8 TB disks, and bought:

- Ultrastar He8 8TB SaS (0F23693) 3x for 700eu tot ( those are dells... so oem..., but the price brought me over).

So on to the next, I can pickup:

• HGST Ultrastar He8 SAS 12 Gbps (OF23268), 8TB for €360,- each [NEW] Oem / sealed (€519 retail);
• OEM Seagate ST8000NM0055 8TB 24/7 V5 Enterprise for €260,- each (€429 retail)
  
• Seagate NAS HDD, 8 TB New + Warranty for €310,- (€444,50- retail)
  
• WD Red 8 TB Helium for €337,90 NEW + warranty
• Seagate Archive HDD v2 -- 8TB for €238,90 each NEW + warranty

So the Oem drives are pretty cheap.... but that means no warranty xD.... I have never had a drive fail on me (jinx) but worry that i'll have to replace one and the costs are higher than the savings...

Anyone experience with this?

Second thought: I could buy cheaper 4 TB / 6TB drives... as I soon receive a 60bay jbod... but that means i'll have to swap them out later anyway

 

[nephri]

It's hard to answer.

Admit you buy 8x OEM HGST Ultrastar He8 SAS 12 Gbps
- 8*360 = 2 880 €
- If you bought retail one, you would paid 8 * 519€ = 4 152 €
- At the immediate buy, you won 1 272 €

With the gain of 1 272 € you can have a 2 drives fails where you can bought yourself a retail one for 1 038€ and you always won 234 € on the global cost.

So if you have 3 or more drive fails, this choice is a loosed bet !!

If you save a lot of money, this money can be your "self warranty".

But in this case your "self warranty" cost to you only when a disk fails and your warranty may be longer than 5 years !
With the retail solution , the warranty cost to you even if you will never have a problem.

So even if my words suggest it's a good idea, i would admit it's hard to go this route myself without if it's a really pretty hot deal and if i'm not really confident with the seller !!

 

[marv]

I am thinking about getting WD mybook duo 16TB which based on info in other thread here contains two 8TB REDs and is about 50 eur less per drive than buying REDs separately- although comes just with 2y warranty and who knows if it will be RMA-able after you extract drives. So add some cashback after selling the box

 

[Deslok]

What about the 8tb mybook that Patrick took apart here recently?

 

[pricklypunter]

Having 8TB of data on any one disk would make me nervous never mind having an array of them full of data. That being said, the Ultrastar's are the only game in town in my book. Providing of course they have been handled and packed properly prior to shipping, they should be excellent performers. Either way, whatever you choose to go with, test, test, test and test them again before putting them to use

 

[SomeGuyInTexas]

Having 8TB of data on any one disk would make me nervous never mind having an array of them full of data. That being said, the Ultrastar's are the only game in town in my book. Providing of course they have been handled and packed properly prior to shipping, they should be excellent performers. Either way, whatever you choose to go with, test, test, test and test them again before putting them to use

I have been thinking about this for a while. At what point is enough for storage capacity on a rotational HDD? The interface can go faster as far as link speed (SAS-4 will be 24 GBS) but the throughput isn't going to increase. Heck, with SMR drives it will likely go down. Rebuild times in a RAID volume will be ludicrous. At some point things will have to change in our way of using these things. I just am wondering when it will be common place that distributed filesystems and/or clusters are the way everyone sets up their storage.

Sent via Tapatalk

 

[Mirabis]

I am thinking about getting WD mybook duo 16TB which based on info in other thread here contains two 8TB REDs and is about 50 eur less per drive than buying REDs separately- although comes just with 2y warranty and who knows if it will be RMA-able after you extract drives. So add some cashback after selling the box

My Cloud ex2 Ultra 16TB costs €759.99 here (with student discount)

My Cloud Mirror 16TB (Gen 2) costs €699.99 (student discount)
But single wd red 8tb costs €335,- new so minimal cost savings here


Sent from my iPhone using Tapatalk

 

[Deslok]

My Cloud ex2 Ultra 16TB costs €759.99 here (with student discount)

My Cloud Mirror 16TB (Gen 2) costs €699.99 (student discount)
But single wd red 8tb costs €335,- new so minimal cost savings here


Sent from my iPhone using Tapatalk

Oh man that's brutal they're ~250 usd here, it would almost vs worth someone buying and shipping them overseas at that disparity

 

[pricklypunter]

I have been thinking about this for a while. At what point is enough for storage capacity on a rotational HDD? The interface can go faster as far as link speed (SAS-4 will be 24 GBS) but the throughput isn't going to increase. Heck, with SMR drives it will likely go down. Rebuild times in a RAID volume will be ludicrous. At some point things will have to change in our way of using these things. I just am wondering when it will be common place that distributed filesystems and/or clusters are the way everyone sets up their storage.

Sent via Tapatalk

From an engineering viewpoint, I think we're already there. Disk manufacturers are pushing the boundaries now of what's possible/ reliable with traditional mechanical drives I reckon. The individual disk capacity is not so much of an issue, with properly implemented redundancy schemes, but getting data on and off magnetic media in a timely fashion is becoming much more of a choke point for modern computing as is ensuring data integrity over the longer term. I'm sure that manufacturers still have a few tricks left in the bag to eek out that last bit of useful life from the technology, but the writing is on the wall already for mechanical drives. Clever file systems and software may well extend this golden period, but even that is not going to solve the issue for very long. A whole new technology is going to be needed in the future. I'm sure ssd will play some part, at least in the short term, but I think whatever comes next needs to go beyond even ssd to be able to really move things forward.

Thinking random thoughts out loud here in the twilight zone, some form of bio or real optical storage is what I reckon we are going to need, and the means to access it. Basically something practically limitless that is instantly recyclable and available immediately at point of use. This will have a ripple effect on everything as more and more things make use of it. If we ever get there, it will be truly life changing

 

[BackupProphet]

It is much better to go the CephFS route when dealing with large drives. But CephFS benefits starts with 50+ hard drives. As Ceph structure data in shards, it means that when you lose a drive, it can rebuild the shards on different drives. So you never have to rebuild the new drive you're replacing. You still have to read all the lost data (for example 6TB stored on a 8TB drive), but with 100 drives and plenty of shards this is no longer a bottleneck.

 

[Mirabis]

It is much better to go the CephFS route when dealing with large drives. But CephFS benefits starts with 50+ hard drives. As Ceph structure data in shards, it means that when you lose a drive, it can rebuild the shards on different drives. So you never have to rebuild the new drive you're replacing. You still have to read all the lost data (for example 6TB stored on a 8TB drive), but with 100 drives and plenty of shards this is no longer a bottleneck.

Hmm moved from Storage Spaces to, Stablebit Drive pool + Snapraid as I extend x disks at a time and zfs wouldn't be feasible for me atm. Heard of ceph for docker... But will look into that


Sent from my iPhone using Tapatalk

 

[CyberSkulls]

From an engineering viewpoint, I think we're already there. Disk manufacturers are pushing the boundaries now of what's possible/ reliable with traditional mechanical drives I reckon. The individual disk capacity is not so much of an issue, with properly implemented redundancy schemes, but getting data on and off magnetic media in a timely fashion is becoming much more of a choke point for modern computing as is ensuring data integrity over the longer term. I'm sure that manufacturers still have a few tricks left in the bag to eek out that last bit of useful life from the technology, but the writing is on the wall already for mechanical drives. Clever file systems and software may well extend this golden period, but even that is not going to solve the issue for very long. A whole new technology is going to be needed in the future. I'm sure ssd will play some part, at least in the short term, but I think whatever comes next needs to go beyond even ssd to be able to really move things forward.

Thinking random thoughts out loud here in the twilight zone, some form of bio or real optical storage is what I reckon we are going to need, and the means to access it. Basically something practically limitless that is instantly recyclable and available immediately at point of use. This will have a ripple effect on everything as more and more things make use of it. If we ever get there, it will be truly life changing

I often wonder how large they will get. On the flip side I remember reading these same things about they can't get much bigger when 1TB drives came out, then 2TB and 4TB.. And here we sit with 8-12TB drives. I think we will see them continue to get larger and cheaper per TB till SSD price/TB comes down substantially. I look forward to the day of affordable 8TB SSD's but I think we're many many years away from that.


Sent from my iPhone using Tapatalk

 

[pricklypunter]

Hehe, I remember similar stories when ATA (IDE) drives were beginning to replace RLL Drives, never mind the actual size of them. I still remember being told that the capacity of my new 40MB disk (being twice the "normal" size of 20MB) in my new AT computer with an upgrade to 2MB of RAM and the new 386DX chip, would never be needed

Most expensive PC I ever owned

 

[SomeGuyInTexas]

Hehe, I remember similar stories when ATA (IDE) drives were beginning to replace RLL Drives, never mind the actual size of them. I still remember being told that the capacity of my new 40MB disk (being twice the "normal" size of 20MB) in my new AT computer with an upgrade to 2MB of RAM and the new 386DX chip, would never be needed

Most expensive PC I ever owned

Ah, the good old days... I remember having to use DOS debug to run the RLL controller (or was it MFM, 25 years makes it hard to recall exactly which one) configuration utility to format the drive. My first AT computer was definitely the most expensive one I ever owned - splurged on not only a 32MB HDD but ALSO one of those new fangled VGA cards and not the old school EGA ones.

But, for this topic, we really are at an inflection point when it comes to single point of failure storage devices. Even the ever increasing in size SSDs will have other side effects - FTL table rebuilds on power loss, secure erase initialization times, etc. are growing as these capacities get larger and larger.

 

[Mirabis]

Ah, the good old days... I remember having to use DOS debug to run the RLL controller (or was it MFM, 25 years makes it hard to recall exactly which one) configuration utility to format the drive. My first AT computer was definitely the most expensive one I ever owned - splurged on not only a 32MB HDD but ALSO one of those new fangled VGA cards and not the old school EGA ones.

But, for this topic, we really are at an inflection point when it comes to single point of failure storage devices. Even the ever increasing in size SSDs will have other side effects - FTL table rebuilds on power loss, secure erase initialization times, etc. are growing as these capacities get larger and larger.

Yep... so many decisions. Swapping out old hardware for new, have 8 consumer SSDS... but sometimes my VM's still feel sluggish... then i look at IOPS spec.. and it's even more confusing. 15K Enterprise Write IOPS / consumer lists 90K xD... been reading a lot on the forum... and see a few deals, but have yet to decide.

Just need the 8TB disks for mass storage & a few SSD's for cache/VM storage

 

[T_Minus]

Yep... so many decisions. Swapping out old hardware for new, have 8 consumer SSDS... but sometimes my VM's still feel sluggish... then i look at IOPS spec.. and it's even more confusing. 15K Enterprise Write IOPS / consumer lists 90K xD... been reading a lot on the forum... and see a few deals, but have yet to decide.

Just need the 8TB disks for mass storage & a few SSD's for cache/VM storage

I urge you to do more reading. Consumer SSD are not tested the same as Enterprise SSD and thus their rating are not apples to apples. Enterprise SSD are as fast as consumer doing the 'same test' if not faster yet consumer SSD are not nearly as fast as enterprise SSD when doing enterprise tests.

 

[Mirabis]

I urge you to do more reading. Consumer SSD are not tested the same as Enterprise SSD and thus their rating are not apples to apples. Enterprise SSD are as fast as consumer doing the 'same test' if not faster yet consumer SSD are not nearly as fast as enterprise SSD when doing enterprise tests.

I did read that ;p, just makes it hard to buy stuff hehe. Couple Enterprise ssds for vm storage or 1 nvme price wise... Will wait for a good deal


Sent from my iPhone using Tapatalk

 

[T_Minus]

I did read that ;p, just makes it hard to buy stuff hehe. Couple Enterprise ssds for vm storage or 1 nvme price wise... Will wait for a good deal


Sent from my iPhone using Tapatalk

Exactly, I go back and forth! It's tough when so many gotchas exist for NVME right now.
IE: Price, OmniOS/Napp-IT not working with NVME in pass-through, limited capacity/price, etc...

If you're using ZFS I would start a high-performance VM pool with 4x Enterprise SSD and expand from there

 

[Mirabis]

Exactly, I go back and forth! It's tough when so many gotchas exist for NVME right now.
IE: Price, OmniOS/Napp-IT not working with NVME in pass-through, limited capacity/price, etc...

If you're using ZFS I would start a high-performance VM pool with 4x Enterprise SSD and expand from there

I slowly add more disks, so ZFS doesn't really work for me

Been using Storage Pools / StableBit on/off , but think I will run Storage Pool's version of Raid 10 with 4 Enterprise SSDs OR buy 2 Enterprise SSDs / and do Storage Tiering, to have cold data on a cheaper disk.

 

[marv]

My Cloud ex2 Ultra 16TB costs €759.99 here (with student discount)

My Cloud Mirror 16TB (Gen 2) costs €699.99 (student discount)
But single wd red 8tb costs €335,- new so minimal cost savings here


Sent from my iPhone using Tapatalk

My cloud is NAS, so you are paying more because of it. But check My Book DUO, which is DAS. I am finding these for 530 EUR incl. VAT.