What drives to buy?!? Too many options!

[LIGISTX]

I am trying to figure out what drives to purchase for my soon to be created freenas VDEV.

I was originally between WD greens and blues, knowing that greens are now branded as blues? I am still somewhat confused by this. I did plan to figure out just which drives WDIDLE3.exe would work on, but then I figure I should try and check out what other options are out there.

From some quick research, I have found HMS5C4040ALE640 (HGST MagaScale 4TB Coolspin) for a very good price on ebay, and from backblaze, they seem to be very very reliable. Has anyone here had any experience with them?

According to backblaze data, they have a .93% failure rate? That seems phenomenal.

 

[Nick2253]

I would focus on drives that are intended for NAS use, as they will have a longer warranty, and are (likely) more robust given the wear and heat. I would stay away from 7200RPM drives, unless you have a need for the added I/O (e.g. a VM datastore). If you are just using them for a file server, the extra RPM will translate to extra power and heat, and you won't see a real-world performance increase.

However, the most important thing, in my opinion, is cost. What costs the least? I wouldn't buy drives on eBay unless I knew I would get warranty service, or the drive was cheap enough for me to buy a couple extras to offset the loss of warranty.

 

[LIGISTX]

I would focus on drives that are intended for NAS use, as they will have a longer warranty, and are (likely) more robust given the wear and heat. I would stay away from 7200RPM drives, unless you have a need for the added I/O (e.g. a VM datastore). If you are just using them for a file server, the extra RPM will translate to extra power and heat, and you won't see a real-world performance increase.

However, the most important thing, in my opinion, is cost. What costs the least? I wouldn't buy drives on eBay unless I knew I would get warranty service, or the drive was cheap enough for me to buy a couple extras to offset the loss of warranty.

I found them for 99.95 a pop. They are intended for 24x7 use, and "low enterprise" workloads, which is less than 180 TB a year which I won't even come close to. They have " Enhanced Rotational Vibration Safeguard (RVS) for robust performance in multi-drive environments"

I don't know what would lead these to be a bad NAS setup. I am not even remotely at an enterprise level of usage, but why would such a class of drive be a "worse" option?

I can't see these being worse then greens for blues.

 

[LIGISTX]

This is a duplicate post from another one of my threads, sorry for the dupe but I am trying to build this thing out QUICK as my old NAS is starting to take a dumb :/

Ok. Well. This "which drive to buy" game is not an easy one!

I have narrowed it down to 2 options. First, the HMS5C4040ALE640 (HGST MagaScale 4TB Coolspin) which has a VERY VERY low fail rate on backblaze, sub 1%, and I can get them for 99 bucks, or Seagate ST4000DM000 which have a 3.06% fail rate and are 111 bucks from Amazon Prime.

The seagate is the most used drive by backblaze that I can tell, but I think that was only because they can't buy the HGST in the quantity they would want them in.

Any advice or input here would be amazing. The HGST seems to be an enterprise level device, where the Seagate is just a solid performing consumer drive which affords it a 2 year warranty vs the HGST 1 year warranty.

 

[SweetAndLow]

I wouldn't buy used drivers. I would also be careful about buying 7200rpm drivers since they are way harder to cool. Find a good 5400-5900 rpm drive that fits your storage needs.

 

[LIGISTX]

I wouldn't buy used drivers. I would also be careful about buying 7200rpm drivers since they are way harder to cool. Find a good 5400-5900 rpm drive that fits your storage needs.

The HGST's I am looking at are new, and are 5700 RPM which is more than fine for my needs. The only issue now is do they support CCTL (HGST's version of TLER). I think they may, but I am not 100% sure. Is it really THAT important though? It doesn't seem to be.

 

[Stux]

With out TLER etc, a drive might hang for a long time (say a minute or two) when trying to read data, that could've been rebuilt from parity.

Many raid systems will drop the drive in this case, and then begin a rebuild.

Rebuilding is a high risk process, and if another drive then hangs...

Meanwhile with TLER, the drives are responsive, timing out in a few seconds, and not dropping out of the array, thus not causing high data loss risk scenarios.

 

[LIGISTX]

With out TLER etc, a drive might hang for a long time (say a minute or two) when trying to read data, that could've been rebuilt from parity.

Many raid systems will drop the drive in this case, and then begin a rebuild.

Rebuilding is a high risk process, and if another drive then hangs...

Meanwhile with TLER, the drives are responsive, timing out in a few seconds, and not dropping out of the array, thus not causing high data loss risk scenarios.

I agree this could be an issue, but what I am missing here, BackBlaze states they remove drives if it fails to sync or stay synced in the array and is counted as a failed drive. If that is the case, weather these drives be in software or hardware raid, a sub 1% fail rate would include this metric "The drive will not sync, or stay synced, in a RAID array".

I really need to get in contact with someone there. If they have thousands of these deployed, they would probably know how they act in their arrays. From what I have read it seems they act well, but there isn't that much detail in the wording. They do show their SMART stats though I believe, but man that would be hard to sift through.


Sent from my jailbroke iPhone using Tapatalk

 

[Stux]

Backblaze is not you ;)

Their systems are built to use crappy drives in vast quantities, and most of their data is dormant, since their product that they supply is "unlimited backup"