SMR 8TB Vs Iron Wolf 8TB - which would you choose ?

[Jim Beam]

 

I have to buy my drives in Europe because it makes more sense then buying in the US and shipping to the DC to go into a server.

 

So these are the prices i have found in Europe.

 

Seagate IronWolf, 8TB - 8TB 3.5" HDD (SATA-600) @ 210/MB/s : € 265,-

 

Seagate Archive HDD v2 ST8000AS0002, 8TB - 8TB 3.5" HDD (SATA-600) @ 150/150MB/s : € 259,-

 

Both rated at 180TB a year, 1M hours MTBF. So no real difference there on duty cycle.

 

Intended use is write once, read many for very large media files - average size 20+GB

 

There will be MANY reads from these drives.

 

Which would you go for and why ?

 

Thanks for your opinion.

[garycase]

Probably the Archive drives, although with the nearly identical pricing it's a touch choice.

 

For your application, the shingled writes don't matter -- with primarily large files, and a write-once application, the performance will be virtually identical to a PMR drive, so that's not a factor.

 

My choice is based primarily on the lower-RPM and slightly lower power requirements for the Archive units => which translates to less heat generation.  I'm still old-school in that area ... i.e. heat is "bad" and less heat is "good" 

 

Further, I presume since you're asking the question here that these drives are going to be in an UnRAID server, so the higher RPM of the IronWolf will have minimal performance impact, since your bottleneck will most likely be your network speed (I presume you have a Gb network infrastructure -- clearly if you're running 10Gb then that's a different story).

 

[Jim Beam]

Hey Garycase,

 

Again thank you. Maybe you can see where i am going with all this. A layered approach. First how do i run the VM's, then the HDD sub system has to be figured out. Lots more questions coming as I put the pieces in place.

 

You are right about power usage. I want to put 36 drives in a server in a data center in Europe where electricity is charged out as consumed and its expensive (at least compared to where i am physically located).

 

So the power draw per drive is important to me - one of the reasons i want to use UNRAID in the server - drives can be spun down and idle when not in use, but still get the advantage of parity protection and be able to ask "smart" hands to easily pull a drive if it dies and easily replace without taking down a whole array.

 

If the drive sits essentially idle in UNRAID , does it use any significant power at all ?

 

Heat is also an issue. The less the better it has to be said. But in the DC its pretty cool all the time.

[garycase]

... If the drive sits essentially idle in UNRAID , does it use any significant power at all ?

 

Assuming that by " essentially idle"  you mean spun down, then there's not any significant difference.

 

When being accessed, the IronWolf uses about 20% more power than an Archive drive (9.0w vs. 7.5w).  When spinning but no read/write in process, the difference is more like 40% (7.2w vs. 5.0w).  But when in standby (spun down), they both draw less than a watt (0.6w for the IronWolf, unspecified for the Archive -- just says <1w in the spec sheet).  In the spun down state, they're probably virtually identical, since the only thing drawing power is the PCB ... thus the difference in RPM isn't a factor.

 

 

[Jim Beam]

What are "Load/Unload" cycles ?

 

Ironwolf drive has 1m Hours MTBF, 600,00 load/unload cycles,  9.0 W avg/7.2W idle / Max Temp 70C / 265.00 Euro

 

Archive drive has 800,00 MTBF, 300,00 load/unload cycles, 7.5W avg/5.0W idle / Max Temp 60C /259.00 Euro

 

 

Its a wash.

 

 

Do i go for 800,000 hours MTBF ? ....or is 600,00 load/unload cycles more important ?

 

2.5W per drive power usage is about 50W over 22 drives (20 data, 2 parity, 2xSSD's for cache) in front of server. 12 HDD's in the back of the server, but not sure on the back 12 yet.

 

50W per hour x 24 hours a day x 7 days a week x 365 days a year = ? KWh ".17c (euro) per KWh.

 

I guess it all depends on final config and usage and if the HDD's sit in idle mode or actually power off when not used under UNRAID.

 

Not sure about the last part because i have never used UNRAID.

 

 

 

[Jim Beam]

 

 

Assuming that by " essentially idle"  you mean spun down, then there's not any significant difference.

 

But when in standby (spun down), they both draw less than a watt (0.6w for the IronWolf, unspecified for the Archive -- just says <1w in the spec sheet).

 

Ok thanks for that.

 

with UNRAID, when a drive is not reading or writing it spins down - right? So essentially no power used or at least so miniscule that its insignificant.

 

I just want to make sure i have all my facts right.

 

Still got to crunch numbers on power usage cost but with the lower power requirements of the 10TB IronWolf and lower wieght (they will be shipped to me one day and air freight is expensive) it might not be that big a difference in cost over the life of the drives between 8TB and 10TB Iron wolf. I am going to pay for it all eventually, so have to be aware of these things.

 

10TB Ironwolf is 6.8W Avg/4.42 idle/0.8 standby/650Grams

[garycase]

A load/unload cycle is when the heads load -- basically this will be once/spindown cycle.    So in a worst case scenario, suppose you have the drives set to spin down after 30 minutes, and they are accessed every 30 minutes plus 1 second ... so every time they spun down, they immediately spun back up.  That would be a load cycle every 30 minutes ... or 17,520 cycles/year.  And of course in actual use, you'll not get anywhere near that number.  In other words, you don't have to worry about that spec 

 

r.e. power usage => I presume you know how to calculate that, but just for completeness, your 50w/hr = ~ 438 KWH/yr

[50w x 24 hrs/day x 365 days = 438,000 watt-hours = 438 kilowatt-hours]  ... or about 75 Euro per year at the rate you noted.

 

But if your drives are spun down (in standby mode) most of the time, the differential will be FAR less than that, since when spun down the IronWolf and Archive drives uses essentially the same (very low) amount of energy.

 

The 10TB IronWolf's are an interesting alternative -- what's your cost for those?

 

 

[Jim Beam]

The 10TB IronWolf's are an interesting alternative -- what's your cost for those?

 

Good question - out of stock with no price given as result. In the US they are quite expensive so come nowhere near being suitable.

 

10 TB vs 8 TB over 22 HDD's does give 44 TB's more capacity. But at huge cost given what i see in the US.

 

Thank GOD i am not an accountant - i hate counting pennies. But has to be looked at - at least approximately. I'm done for the day - tired of reading spec sheets, looking up websites and crunching numbers on the calculator :-)

 

Thanks for your input Garycase

[JonathanM]

A load/unload cycle is when the heads load -- basically this will be once/spindown cycle.

Definitely happens at spindown / spinup, but also can occur while spun up. After a certain period of no requested I/O, a drive will unload the heads. It will stay spinning indefinitely, until explicitly told to spin down. The timeout period before a drive unload happens is programmable, and controlled by the drive's firmware. Spin down is never initiated by current drives, it must be requested by the OS.

 

Certain models of drives have had unrealistically short timeout periods for unload, causing HUGE LC counts in very few running hours. There are posts around here outlining the problem and the solution for those older drives. I haven't seen any similar complaints on current drives.

[garycase]

I'm well aware of the idle unload issue -- but I don't think it's a problem with any modern drive => especially the very high capacity units, which are very much designed for high capacity data center type applications.    I think the days of WDIDLE are gone 

[John_M]

I'm well aware of the idle unload issue -- but I don't think it's a problem with any modern drive => especially the very high capacity units, which are very much designed for high capacity data center type applications.    I think the days of WDIDLE are gone 

 

WD Greens came out of the factory set to 8 seconds right up until the time they were discontinued. The new WD Blues (ie. the high capacity, low RPM Green replacements/re-brands) seem also be be set to 8 seconds - certainly, the two 4 TB ones I bought in September were.

 

[garycase]

Even with 8 seconds, there aren't likely to be all that many load cycles in a typical UnRAID use case.

 

Remember, if the heads unload after 8 seconds, they won't load again until there's another access to the disk -- and if there hasn't been an access in 8 seconds, it's likely there won't be another one anytime soon in a typical media scenario.  e.g. if you're playing a media file, it'll never go 8 seconds without an access ... and when it does, the disk is likely going to sit idle until the spin down timer and then just spin down => so 8 seconds doesn't add all that much.

 

Just for grins, I had a look at my media server, which still has a bunch of old 2TB WD Greens.  I NEVER changed the idle timeout on these, and the worst case I found was one with 12,741 load cycles  -- and that's after 54,426 hours [6 yrs, 2 months, 14 days, and 18 hours).    Granted, my Reds are much lower -- the largest count was 626 load cycles (after 1 yr and 4 months) ... but none of those numbers are anything to worry about.