How much power does unRAID save compared to traditional RAID systems?  I have an existing general purpose Linux server with 6 drives in a RAID10.  I'm looking at building a new unRAID system.  Presuming a similar number of drives, how much power does unRAID save compared to just going with Ubuntu Server Edition on RAID10?

It's all depends on what you do wiht it. It is not only about power saving either.

See the main advantage of unraid is that even if you loose 2 drives only data on those drives are lost.

 

Sent from my SGH-T889 using Tapatalk

 

 

I'd really rather not lose anything, but I do understand how unRAID works in that way.  (One of the reasons I'm on RAID10 right now is that you can lose one drive from each pair and not lose data.)

 

Anyone else?

Again, it depends on what you use it for (aka how you use it). 

 

With your current RAID 10 all 6 drives have to be spinning to perform reads or writes.  With unRAID only the drive being read from must be spun up.  For writes the drive being written to and the parity drive must be spun up. 

 

So at a minimum (and assuming the files being read/written were on a single disk in a 6 disk array), you would save the power of 5 disks for reads and 4 drives for writes.  The average difference between spun down state and operating state for a current-gen 7200 rpm spinner is ~7W.  So ~35W savings for reads and ~28W savings for writes.

As noted above, it very much depends on the specific drives you're using and the way in which the array is being used.

 

In rough terms, a typical "green" drive will draw about 1 watt when spun down; and about 6 watts in operation -- so you save about 5 watts when it's spun down.

 

A 7200 rpm unit that difference is (as dirtysanchez noted) about 7 watts.

 

Without trying to model the "average hrs/disk spinning" for the array, let's just assume you average at least 20 hrs/day of "spun down" time for your disks/day.    That's not at all unusual for UnRAID ... in fact it's probably even higher, since many disks probably never get spun up on most days.    But at 20hrs/day and 5w/disk savings, that 100wh/disk/day = 36.5kwh/yr/disk.  At a typical $0.12/kwh, that's a savings of $4.38/disk/year.    A 7200 rpm unit would be about 40% higher than that, or roughly $6/yr/disk in savings.    If you're using NAS-rated drives like WD Reds the numbers are a bit lower ... the savings are about 20% lower than the Green drives -- or about $3.50/disk/year.

 

So for a 6-disk system using WD Reds, you'd likely save $21/year using UnRAID vs. a traditional RAID that's spinning 24/7.

 

... an afterthought ==>  remember that power savings isn't really the primary goal of using UnRAID.    The ability to mix/match drive sizes;  the potentially longer live of the drives since they're idle much of the time (it's of course arguable whether that's actually better or worse than leaving them spinning 24/7);  the lower heat generation in your system;  the less catastrophic impact of multiple drive failures (which SHOULD be irrelevant if you're properly backed up ... but many folks aren't);  etc.  are all additional factors to consider.

 

Hmmm.  Thanks for the thoughts.  Here's where I am.  Generally, I'd tend towards a generic Linux server due to the number of ancillary services I plan on putting on the system (Squeezeserver, MythTV, possibly Asterisk, etc.) and the importance of the data residing on it (family photos).  However, our power costs average $0.31/kWh here, so power savings can be substantial.  Based on garycase's calculations, I'd probably expect savings of $11/disk/year, or about $500 over an 8-year system lifetime.  Everything important will be backed up by Crashplan, so the RAID is there to protect against the inevitable drive failure. 

 

For me, the wins of unRAID are cost savings, both in power and drives from the N-1 RAID layout compared to N/2 of RAID10 (more capacity), and the easy mix/matching of drive sizes.  The downsides are the need to use plugins to adapt arbitrary programs to the unRAID disk layout, the slight drop in reliability compared to RAID10, and more complicated mix/matching of drive sizes.

 

Am I missing anything?  You all have been quite helpful so-far.

... the wins of unRAID are cost savings, both in power and drives from the N-1 RAID layout compared to N/2 of RAID10 (more capacity), and the easy mix/matching of drive sizes.

 

True relative to RAID10; but not with respect to RAID5, where the data capacity is the same (N-1).    Certainly true that the easy mix/match of sizes is a nice feature.

 

 

... The downsides are the need to use plugins to adapt arbitrary programs to the unRAID disk layout, the slight drop in reliability compared to RAID10, and more complicated mix/matching of drive sizes.

 

Any complexity in the plugins is due to the underlying Linux distro -- not to the UnRAID disk layout.  From the perspective of the programs, UnRAID is simply presenting a large data store ... whether it's RAID10, RAID5, or the UnRAID layout is irrelevant.

 

I don't think the mix/matching of drive sizes presents any complications -- the only specific requirement is that the parity drive must be >= the largest data drive.

 

If you want the highest reliability RAID, use RAID6 ... which will tolerate 2 drive failures without loss.  That's also true for RAID10, but ONLY if the 2 failures are in specific sets of drives -- with RAID6 it can be any 2 drives.    Note that one of the features planned for a future release of UnRAID is dual parity support -- which will give it the same dual-drive failure protection as RAID6.    There's no clear timeline for adding this support, however, so I wouldn't factor that into any buying decision.

 

Any complexity in the plugins is due to the underlying Linux distro -- not to the UnRAID disk layout.  From the perspective of the programs, UnRAID is simply presenting a large data store ... whether it's RAID10, RAID5, or the UnRAID layout is irrelevant.

That may be true, but it's irrelevant to me.  unRAID has selected that distro, and requires the plugin system.  I have to deal with that if I want to use unRAID in the same way as the majority of users.

 

I don't think the mix/matching of drive sizes presents any complications -- the only specific requirement is that the parity drive must be >= the largest data drive.

Good catch.  That was me mixing my options up.  What I meant to say was that mixed drive sizes is a minor disadvantage of a traditional Linux system, where it can be done, but it's complicated.

 

If you want the highest reliability RAID, use RAID6 ... which will tolerate 2 drive failures without loss.  That's also true for RAID10, but ONLY if the 2 failures are in specific sets of drives -- with RAID6 it can be any 2 drives.

Actually, RAID10 is better than that.  When one drive fails in a RAID10, you only have data loss if the other drive in that mirrored pair fails.  That's 1 out of the N-1 remaining drives.  Certainly, RAID6 is better on that factor, but it's not as much as it might seem at first glance.  RAID5 and RAID6 also have the disadvantage that the recovery process requires a full disk read on all disks to reconstruct the missing data.  With RAID10,  you only have to do a full disk read on the mirror of the failed drive(s).  I do worry about the possibility of cascading failure during a RAID5/6 reconstruction operation. 

 

OTOH, I could just increase the complexity an order of magnitude, and run UnRAID on top of ESXi and have the best of both worlds.