c3

+quite the awesome post, thanks for the work and details. If you read the data, the DX is a bit faster as parity drive, and the DM is the faster drive writing. I was indifferent to the DX vs DM, but I would now suggest hunting for DMs (even for parity) in the unRAID application. You're looking for low cost storage with a bit a sensitivity to in home usage (heat/noise), DM wins. That small performance increase of the DX is not going to seen in daily operation, but the heat and possible fan noise will.

Not going say 100%, but that approach is more likely.

There is interest! +1

Vibration has a multitude of negative impact, reliability, performance, power, even noise (a positive feedback loop). In a chassis like backblaze 1.0, traditional RAID would fail out drives for response timeout. The vibration effects heads, forcing them off track, requiring reseek, since the entire stripe is delayed, the drive is dropped. All the reseeking drives up the power and adds vibrations. This area has a lengthy patent list...

So how do you justify present it as a choice?

Let us know how it is working for you. A drive fails, and the data on the drive is unaccessible, until after a rebuild, which requires all the data on all surviving drives to be unavailable. Not a choice I'll be making.

Do not overlook the option of buying a quality router, far less power than most PCs. This is lounge so I'll put this out there, but you can spend plenty of time in the area. I use the ASUS RT-n66u with TomatoUSB from shibby, but it starts out with open source WRT firmware which means AsusWRT-Merlin firmware available with just fixes. This configuration has been field tested supporting more than 50 wireless devices on 802.11n 5Ghz. Nothing on pfSense, but you can do pretty much anything with WRT. This hardware platform is simultaneous dual band radios and plenty of processing/RAM and gigabit ports.

I'm in the US and have no trouble getting 2+ year warranty. The 4TB drives I last purchased have 2 year warranty (Costco $159). All WD Red are 3 year. The credit card companies add one year to warranty. AmEx max is 6 years (5+1), Discover and Citi is 4 years(3+1), and so on.

You wouldn't and every NAS manufacturer tries very hard to strip everything out of whatever OS they are running, but it is nice for home users to be able to fatten it back up for the applications they desire. Hence the plugins, etc being made available.

Disclaimer: I make my living in the ZFS world. Did you miss the part on how both are limited/controlled by Oracle, and only BTRFS is going forward? Yes, there is lots of work being done on ZFS by many factions. Checkout http://en.wikipedia.org/wiki/ZFS#Comparisons for one list of versions. But ZFS continues to be effectively hemmed into boundaries defined by Oracle. While BTRFS is moving to reach the stability and functionality of ZFS. Don't misunderstand the position of NAS OEMs on ZFS, they will switch as soon as the market demands. And several are working to be the first in the switch, to be seen as the market leader. unRAID has feature and function outside the goals of ZFS and BTRFS which are popular reasons why it is the choice of media server builders. I have tried FlexRAID. Yes, that is the product's name, regardless of recent posts by the "creator of FlexRAID". Hit of the buy/download page to see for yourself. Tom, did you hire a great marketing company? First take the FlexRAID name off the product, and then have an unpurchasable product shout your product is second best (but the only purchasable one)!

He/She/It/They can dream. But this is delusional, the stated parameters of whatever "the creator of Flexraid" is hiding has no chance of entering the enterprise world. Too many conflicting functions, some shared by unRAID. For example; -Each drive can be pulled and read in another system by itself. -In case of failure past the tolerance level, surviving drives are fully readable/writable. In order to accomplish this, all of a file has to be written so that it can be removed from the array. This limits the performance/capacity to single drive performance/capacity. Game over man. -With NZFS, you can upgrade your hardware as your favorite OS evolves. http://www.openegg.org/2012/02/17/nzfs-goes-100-kernel-mode/ Windows (and Linux) OS evolution requires kernel level drivers to evolve, or is He/She/It too young to have experienced the life of a kernel driver writer? And I must have missed the discussion of file locks, dedup and compression, not insignificant requirements to enter the enterprise space.

My personal guess, this is the sale price we'll see for a few months. At two years the warranty is better than the one year average for externals, but we void that by opening the hotbox. I keep dreaming on using them inside the shell, but it's a dream

Absolutely possible to make a driver to mediate this, but the hardware will still be exposed until the driver finishes its work, each cycle. Understand the scope of the bug; You take a motherboard/computer out of the box, rack and power it up, no OS. The bug allows the motherboard/computer to be taken offline. Boot an OS, load a driver which then patches the memory resident copy of firmware, and no more threat exposure, until.. Sleep, and you're back to exposed. If the device is not reachable from the internet, very limited risk. knocking machines off the network is not a good propagation method. Disruptive to be sure, but far better than a remote root exploit. This bug does not provide a method for unauthorized access, just disruption. This chip used as your internet ingress, yikes!

I know it has been mentioned before, do not put unRAID on the internet, and here is yet another reason. http://blog.krisk.org/2013/02/packets-of-death.html Basically, the very inexpensive 82574, has yet another bug. This one allows the ethernet port to drop link with the correct packet. I found the defect to be present on a Supermicro motherboard I use, so I post in the motherboard section. Previously discuss risks include the IPMI security risk. Since getting a corrected firmware seems unlikely, I chose to self ping the inoculate packet as part of startup.

Assuming you have a free port and you are using a 3TB parity, then installing a new 4TB as parity and moving the 3TB parity into a data slot, you then gain 3TB not 1TB. True it is at a price premium over just buying anew 3TB, but you are now ready to take full advantage of adding future 4TB into data slots. But in the mean time, you've still at least upped your capacity by 3TB. Well said! The yield and cost of switching drives is complex.

me too. Stop putting words out there that I have not said. I never said there was a shipping 4TB drive using only 4 platters. I never said there was an article comparing 4TB drives. All I said was I read a comparison of the same sized seagate vs hitachi quite a while back that showed the numbers. It was around the 2TB / 3TB era. Not now. I said specifically. Sorry, my head has the thread title stuck in it, 4TB drives. I've been using 4TB drives for almost two years, so I have many old articles half remembered in there.

I'd like to see the article on a shipping 4TB drive using only 4 platters.

The Costco 4TB externals have Seagate 7200rpm drives in them for $160.

5 Platter designs are good for RANDOM I/O. I'd like to see the math behind that statement. This drive has the same average latency, 4.16ms. Less head movement from what I remember. In other hitachi designs that had 5 platters there was a measured improvement of random I/O. It was slower in some sequential I/O but faster in Random I/O. It was a while back. I don't remember the article or where I read it. You're saying the lower density platters have faster access times. So older (5 platter) drives are faster than newer (4 platter) higher density drives. That's not a trend the market would allow. I do believe the hitachi 5 platter benchmarked faster than the seagate drive with 1TB platters, but lots of variables in there. One was a 4TB the other 3TB, and the difference was very small. The 1TB platters will have more, smaller movements, while the 5 platter has fewer, bigger movements AND more mass to start stop.

Oh they run hot for sure, especially in the no fan external cases. I pre clear with a 120 moving air across them.

5 Platter designs are good for RANDOM I/O. I'd like to see the math behind that statement. This drive has the same average latency, 4.16ms.

http://www.isuppli.com/Memory-and-Storage/News/Pages/Hard-Disk-Drive-Market-Revenue-Set-for-Double-Digit-Decline-This-Year.aspx one pager from smart people.

ZFS on Linux! Here is the bottom of the "external" drive with the meant to be removed adapter removed. No warranty voided to pre clear these.

4TB externals at Costco for $160, 3TB still $130, so the scale has tipped.

Hoping to get 10 more for a new raidz2 stripe. At this price they are cheaper than typical 3TBs on sale. Manufacturing lots are very large. It is likely all the drives on sale from all Costcos are one lot. But they have traveled and handled differently.