wd red pro / gold vs seagate exos & ultrastar hc550

[phier]

hello,
these drives are almost same by spec... i think exos or ultrastar even better by MTBR better, even redpro and gold costs maybe by 100-150 euro more, any idea why?

from datasheets

thanks

 

[phier]

i plan to build raidz1 3 drives ... ie 1x exos, 1x ultrastar and was thinking to add 1x gold/red but as price is higher maybe one more ultrastar/exos instead?

even its raidz1 - should be enough as there will be off-site mirror

 

[Jessep]

WD Ultrastar DC 5xx would be the best choice if they are comparable on price

 

[phier]

@jesseph they are cheaper then gold/red ... so i dont get it. better to say if gold/red is expenseive they have to be better - but how?

@jeseeph how do you run ESXi 7.0 U3 with FreeNAS 11.1 U7 VM ? is esxi separate HW/Box?

 

[Jessep]

ESXi on bare metal with TrueNAS as a VM with passthrough HBA.

"Absolutely must virtualize FreeNAS!" ... a guide to not completely losing your data.

[---- 2018/02/27: This is still as relevant as ever. As PCIe-Passthru has matured, fewer problems are reported. I've updated some specific things known to be problematic ----] [---- 2014/12/24: Note, there is another post discussing how to deploy a small FreeNAS VM instance for basic file...

www.truenas.com

You've reminded me to update my signature as I'm running 12.0 U8.1 now.

 

[Etorix]

Different markets, different pricing strategies: Red/Ironwolf = consumers (sold by the unit); Ultrastar/Exos = enterprise (sold in bulk, with build-in bulk discount). As far as I know, since WD bought HGST, WD Gold are merely Ultrastar with a WD sticker (and the same 3.3V_PWRDISABL behaviour…).
Get the cheapest, and don't bother mixing and matching unless you're willing to spend more just to try and avoid that the drives come from the same factory batch.

Beware, tough, that raidz1 is not a recommended configuration with 16-18 TB drives. Raidz2 (or 3-way mirror) is the minimum safe option if the pool is intended to actually sustain the loss of one drive with a low risk of having to restore from backup…

 

[phier]

@Etorix thanks!
Well so if red/ronwolf is for consumers sold by the unit - the final price for a customer is higher, weird. OR you mean enterprise (sold in bulk to dealer - thats why final price for a customer is cheaper?)
I saw ppl mentioned that gold has flying height, is filled with some different gas etc ... so thats why its better. (https://www.truenas.com/community/threads/hard-drives-wd-gold-vs-wd-red-pro.44008/post-509003)

The exos and Ultrastar costs the same ... so even the price is the same its not a good idea to mix ie 2x ultrastar + 1x exos? (sure i want buy each one from different seller - ie maybe different factory batch at the end)

Well thats why i opened other thread regardin that discussion, why raidz1 is not recco? I plan to have other nas where i will mirror (daily) raidz1 ... so it will be fully backed up, even in that scenario raidz1 is not good?

I am currently running raidz1 for 6y and still (luckily no issue ... ) maybe the issue will appear once i start to migrate.

thanks!

 

[phier]

@Jessep well ... i wanted to install truenas natively / directly on hw ... but what u mentioned opened my thinking ... isnt better to virtualize it and use that HW for other VM also...

but question would be - is such a setup fully supported by truenas?
thanks

 

[Jessep]

Fully supported in what way?

• For business production with a support contract? No.
• For a home lab that has backups of critical data? Yes.

Running fully virtualized is very stable if you follow the rules, and not if you don't. Just like anything else.
The basics IMO, and not a complete list, just off the top of my head.

• Supported hardware
• High quality fully tested hardware
• Passthrough HBA
• Enough fully reserved RAM
• Using a production ready software platform (example VMware)
• Backups of critical data
• UPS with auto VM shutdown/start-up
• Not f*cking with things all the time, i.e. treat it like it's production

Otherwise run various forms of bare metal and VM using non critical data until you can maintain uptimes of months. If you don't know what you are doing, or even worse think you DO know what you are doing, you will lose data.

Ask how important is your data/uptime and act accordingly, then double it at least.

Plenty of people here who do this professionally and we ALL still make mistakes. Honestly what shows a professional is being able to fully recover from mistakes with minimal downtime and no data loss, not never making them in the first place.

 

[Etorix]

Well so if red/ronwolf is for consumers sold by the unit - the final price for a customer is higher, weird. OR you mean enterprise (sold in bulk to dealer - thats why final price for a customer is cheaper?)

Yes, I guess that's the idea…

I saw ppl mentioned that gold has flying height, is filled with some different gas etc ... so thats why its better. (https://www.truenas.com/community/threads/hard-drives-wd-gold-vs-wd-red-pro.44008/post-509003)

Helium is helium, and the hardware platform for all WD drives is probably the same by now but there may be subtle optimisations and/or firmware tweaks.

The exos and Ultrastar costs the same ... so even the price is the same its not a good idea to mix ie 2x ultrastar + 1x exos? (sure i want buy each one from different seller - ie maybe different factory batch at the end)

Fair enough, I shouldn't hurt to mix a bit.

Well thats why i opened other thread regardin that discussion, why raidz1 is not recco? I plan to have other nas where i will mirror (daily) raidz1 ... so it will be fully backed up, even in that scenario raidz1 is not good?

It's down to this:

Storage | ZDNET

Everywhere we use data - in the cloud, near the edge, on your desktop or around your wrist -- we need storage.

www.zdnet.com

HDDs return Unrecoverable Read Errors at a rate of one for 1E14 to 1E15 bits. 12 terabytes is 0.94E14 bits…
More precisely, for N bytes, the probability of NOT having an URE (rate u) is p(N) = exp(-8Nu)

Let's then assume that one drive fails in a 3-wide raidz1 of 18TB drives, which was 2/3 full, i.e. 2*18T*2/3 = 24 TB to be read for resilver.
For u=1E-14, p(24T) = 14.7% ouch! Get that backup!
For u=1E-15, p(24T) = 82.5% better but still a 17.5% chance of having an URE and having to restore the affected file from backup.

This is "just" the basic scenario where one drive accidentally fails, and does not consider that remaining drives may then fail under the strain of resilvering (as could well happen if the first drive actually wore out).

As long as you have a backup, you will eventually recover. It's up to you to decide on the acceptable level of risk and level of hassle to restore. Personally, I expect a redundant array to cope with the most basic damage scenario—namely, a single drive failure. The backup is intended solely for multiple failures and other major damage scenarios.

 

[phier]

just wondering why they suggest WD RED https://youtu.be/Wya16ef1G-E
:)

 

[phier]

anyone herE?

 

[Redcoat]

just wondering why they suggest WD RED https://youtu.be/Wya16ef1G-E
:)

They don't - they suggest WD RED PLUS - which is consistent with the non-SMR, CMR only, messaging:

and just like the header:

See this post https://www.truenas.com/community/threads/WD-SMR-iX-Statement/#post-584079 for background info on the issue and iXsystems/WD relationship.

 

[phier]

yes wd red pro, but no clue why as there are other as mentioned here
ultrastar,
exos...

 

[Redcoat]

es wd red pro, but no clue why as there are other as mentioned here
ultrastar,
exos...

Likely because they don't have a sponsorship $$ feed from any other hdd mfg besides WD.

 

[phier]

thats odd :/