I'm looking at building a 20 drive budget build as outlined here:

http://greenleaf-technology.com/category/portotype-builds/

-----------------------------------

20 Drive Budget Build (tower)

Mobo: ECS A885GM-A2 (V1.1)

CPU: AMD Sempron 145

RAM: Crucial 2GB DDR3 1333 (feel free to upgrade, pretty much any DDR3 1066 or 1333 RAM should work)

PSU: CORSAIR Enthusiast Series TX650 V2 650W

Case: Antec 1200

Hot Swap Drive Bays: Norco SS-500 x 4 (Two alternatives are listed below. Both have their pros and cons. You can read my review of all three cages here.)

 

    Icy Dock MB455SPF-B x 4 (replacement option 1)

    Supermicro CSE-M35T-1B x 4 (replacement option 2)

 

Fans:

 

    Coolink SWiF2-801 80mm x 4 *optional (Norco SS-500/Icy Dock MB455SPF-B fan upgrade if quieter performance is desired)

    GELID Solutions FN-PX09-20 92mm x 4 *optional (Supermicro CSE-M35T-1B fan upgrade if quieter performance is desired)

 

SATA Expansion Cards: Supermicro AOC-SASLP-MV8 8-Port SAS/SATA Add-on Card x 2

Cables:

 

    0.5m 30AWG Internal Mini SAS 36pin (SFF-8087) Male w/ Latch to SATA 7pin Female (x4) Forward Breakout Cable x 4

    Locking SATA (Black) x 2 (2 included with motherboard, also available in Blue, Red, UV Blue, and UV Green)

 

Hard Drives: Green drives (5400 rpm or 5900 rpm), such as WD Green, Hitachi CoolSpin, Samsung EcoGreen, and Seagate LP

Cost: $985 (without the optional fans) + shipping + cost of hard drives

Capacity: 38 TB

Expandability: None.

 

---------------------------------------------

 

Now, I have a core 2 duo, 4gb ram, and a mobo with 2 PCI-e slots. I've also got a 750w Rosewell PSU. I'll be putting the components in the Antec 1200, as outlined. I'm also looking at getting 4xIcy Dock MB455SPF-B x 4 bays. I'm planning on starting with 3 drives in the 3tb area. I was doing some reading and it looks like I should grab 2-3tb drives and 1-4tb drive for parity.

 

Because I'm going to be running 3 drives for a while I won't need the SATA expansion cards and I'll be able to run the free unraid OS for a while before upgrading to pro.

 

Before I sink about $1,000 into this build, I wanted to get any advice you could provide. I read that the WD Red drives are more power efficient and better quality so I'll probably look for those.

 

Thanks!

If i were building mine today,I would use the iStar docs http://www.newegg.com/Product/ProductList.aspx?Submit=ENE&DEPA=0&Order=BESTMATCH&Description=BPN-DE350SS&N=-1&isNodeId=1.  A friend at work built his with these 4 drive bays and he loves them.  If you go the cage route, make sure you cycle your drives through and test every slot in the cage.  I had a dead slot in one of mine when I purchased it.

 

I would only use drives targeted for NAS applications.  I would also start with the 6TB Red, assuming your MB supports large drives.  Yes you will pay a premium now for them, but it will also delay purchasing new drives allowing the price to come down.  At a minimum, I would only purchase 4TB drives.  But use any drives you currently have for now.

 

If you have room for the Norco 4220 or 4224 case, I would highly recommend you look at using it. By the time you purchase 4 cages plus the case, you are going to have spent over $500, probably closer to $600

http://www.newegg.com/Product/ProductList.aspx?Submit=ENE&N=100007731%2050010473%20600010491%20600010490&IsNodeId=1&name=Norco%20Technologies%20Inc.

Personally, I would go with 8GB of RAM.  Even if you never install enough plug-ins to use all of it, unRAID will use any unneeded RAM as a cache for writes.  I get write speeds approaching 110MB/sec when Imgburn begins its transfer of BluRay iso's.  It slows down to about 45MB/sec as soon as the cache fills. 

 

If you ever find it necessary to upgrade, you will probably have to pitch the 2GB of memory as you will not be able to find a place to use it.  (I know that I have a box of  too small memory sticks that I can't even give away.)

I would not buy any new drives smaller than 4TB. Starting with 2 (one parity and one data's is fine).

 

NAS designation means little although they can be good drives. I'd look at the BackBlaze and similar studies and buy something reliable. (NAS drives implement something called TLER which is not needed for unRaid but IS needed for most stripping RAID arrays found in many NAS boxes.).

 

I happen to love the Supermicro 5in3s and dislike the Norcos which completely cover the hottest party of the drive. Surprisingly, I also like the Rosewill 4in3s (which only cost $45) and have two of them screwed together mounted on top of an Antec 900 containing 3 SM 5in3s with long cables feeding them. This gives me 23 drives in my backup server.

 

The CPU and memory is really driven by your intended usage.If you have the stuff listed, use it. If you are buying new, I'd have other suggestions.

Agree with nothing smaller than 4TB ... in fact I'd use a 6TB for parity, so you're not restricted on the size you can add later (at least not until drives > 6TB are available).

 

As for drive cages ... I've used the SuperMicro, IcyDock, and IStar 5-in-3's, and while they are all nice, none of them cool as well as these CoolerMaster 4-in-3's:  http://www.newegg.com/Product/Product.aspx?Item=N82E16817993002

 

... But the CoolerMaster's aren't hot-swap -- they just hold the drives nicely.

 

There's a new IcyDock unit ("Black Vortex") that is what I'll buy the next time I need cages:

http://www.newegg.com/Product/Product.aspx?Item=N82E16817994171&cm_re=4_in_3-_-17-994-171-_-Product

 

... it's also available in a non-hot-swap version:  http://www.newegg.com/Product/Product.aspx?Item=N82E16817198059&cm_re=4_in_3-_-17-198-059-_-Product

 

These should both rival the CoolerMaster for keeping the drives cool.

 

The black vortex is interesting because they're so much cheaper. In fact, I think they're even half the price of the 5 bay icy dock cage.

 

I'll look into a 6tb drive for parity.

 

Speaking of which, from what I understand the parity drive has to be the same as or bigger than all other drives, correct?

 

Also, I've currently got an antec 900. I may just keep that, order 3 cages and more drives. I can always upgrade to a bigger case later, eh?

 

The Antec 900 is a very nice case that will certainly provide plenty of storage ... especially if you use 6TB drives !!

 

Yes, the parity must be >= all other drives in size.

 

I need to do more research to fully understand everything. I've got a few questions.

 

1). If any drive fails, it's covered as long as the drive is replaced and rebuilt. Same goes for the parity drive, correct?

 

2). If I want to upgrade the parity drive in the future, is it as simple as removing the drive and putting a larger one in it's place or is there more to it?

 

3). If the mother board fails it should be a simple fix where you replace the board or board and CPU and unraid should fire right up regardless, correct? Is unraid sensitive to amd/intel cpu swaps?

 

4). If I were to get a 4 drive cage now and then want to upgrade to a 5 drive cage later, how would I go about doing that?

 

 

(1)  Yes

 

(2)  Yes  (no, there's not more to it)

 

(3)  Yes  (no, it's not sensitive to CPU changes)

 

(4)  Just change it ... sell or re-use the 4-drive cage.

 

In regards to the cage, you'd have four drives that you'd have to remember what slot they go in and that's the part that seems dicey to me. I can't imagine you could just put them back in the new cage in any order, they'd have go to in a specific order, right?

 

How much of a hassle is it to upgrade the parity drive?

Resent versions of unRaid remember drives by their serial numbers, so you can attach them in any order. Parity upgrades are easy but time consuming depending on the parity size. The array is usable during a parity rebuild albeit at reduced performance.

Resent versions of unRaid remember drives by their serial numbers, so you can attach them in any order. Parity upgrades are easy but time consuming depending on the parity size. The array is usable during a parity rebuild albeit at reduced performance.

 

That... makes sense. Brilliant though. So you could actually play a shell game with the drives and the machine wouldn't skip a beat. Hah.

Yes, with v5 and later you can freely swap the drives around to various SATA ports and controllers.    As long as UnRAID "sees" all the drives it expects, everything will work fine.

 

As I assume you know, this is NOT the case with v4.7 and earlier.

 

One of my suggestions from long ago.

So at this point, how would you actually face data loss? I'd guess that if two drives fail it's game over. Otherwise, it seems like the only other failure point is if there's a disaster to the machine itself. Are there any other ways?

Data loss can occur in a number of ways ...

 

=>  User error ... you could accidentally overwrite or delete something you didn't mean to.

 

=>  As you noted, a disaster in the machine itself that wiped out multiple drives.

 

=>  A dual drive failure ... if two drives fail, it's not "game over" -- but you WILL lose the data from those two drives.    Note that with traditional RAID-5 array, where the data is striped across all of the drives, you'd lose EVERYTHING in that situation ... but with UnRAID you only lose the data on the two failed drives.

 

=>  A "bug" in the OS.    The only one I'm currently aware of that has significant data loss potential is the "user share" bug, which can cause data loss if you copy data from a folder that's part of a user share, but using the disk share reference, to the same user share.    This is easy to avoid; but is catastrophic if you do it.

 

ALL systems can result in data loss -- RAID is NOT a substitute for having your data backed up.

 

So at this point, how would you actually face data loss? I'd guess that if two drives fail it's game over. Otherwise, it seems like the only other failure point is if there's a disaster to the machine itself. Are there any other ways?

 

Not quite.  If you have two drives fail, you only lose the data stored on those two drives.  The data on the other drives can still be read.  (It should also be pointed out that all data on a failed drive will be lost only if a data drive is completely unreadable.  If the drive can be read, you can get back the data from the portion of the drive which has not failed.  Of course, per Murphy's Law, if only one file on a disk is the only one that is irreplaceable, that is the one file which can not be read!)

 

The biggest potential for disasters is operator error.  Never panic and do something simply because you feel that SOMETHING must be done without thinking it through and getting advice before doing it.

So at this point, how would you actually face data loss? I'd guess that if two drives fail it's game over. Otherwise, it seems like the only other failure point is if there's a disaster to the machine itself. Are there any other ways?

Frank1940 beat me, but I will post anyway since I had some additional points.

 

If two drives fail then they can't be rebuilt from parity. However, no other drives are affected when those fail and all of their data is intact since each drive is independent and each file is completely contained on a single drive. Also, you should have a backup of anything you can't afford to lose, perhaps even offsite.

 

The risk of multiple drives failing increases the more drives you have. I question the need for 20 drive arrays in this day of 6TB drives. Perhaps multiple servers would be a better approach.

The risk of multiple drives failing increases the more drives you have.

 

Which is, of course, the reason that RAID-6 has essentially replaced RAID-5 for most data centers.    Not so you can tolerate two drive failures -- but so you can tolerate a 2nd drive failure during a rebuild (which is statistically fairly likely with modern, high-capacity drives ... especially in larger arrays).

 

... this is also why many of us consider dual parity a significant need for future UnRAID enhancement (and it's definitely on Limetech's list ... although I don't know how far off it might be).

 

 

I question the need for 20 drive arrays in this day of 6TB drives. Perhaps multiple servers would be a better approach.

 

No matter how many servers you split your data among, it doesn't eliminate the importance of maintaining BACKUPS of your data.    That is the real solution to data loss.  ANY data you don't want to lose should always be stored on at LEAST two different systems ... regardless of whether or not those systems are fault tolerant.    Fault tolerance is NOT a substitute for backups.

 

Is it worth buying a nice 6tb, 7200 rpm, 128mb cache drive at $600 for the parity drive? I guess what I'm asking is whether or not increasing the speed of the parity drive would be beneficial?

Is it worth buying a nice 6tb, 7200 rpm, 128mb cache drive at $600 for the parity drive? I guess what I'm asking is whether or not increasing the speed of the parity drive would be beneficial?

 

In my opinion, No.

 

A 7200rpm parity drive can be beneficial if you'll have multiple users accessing the machine at once, since parity is involved in EVERY write.  But it has NO impact on reads, and is also irrelevant if there's only one user accessing the system, since the write speed will then be primarily a function of the slowest drive involved in the write.

 

Personally, I'd go with the 6TB WD Red, but the enterprise-class Seagate (which is what I assume you're looking at) IS a very nice performer and easily bests the WD Reds in most performance tests.    What I don't like about it is there's only a 1 year warranty ... which is strange for an enterprise drive.

 

 

Is it worth buying a nice 6tb, 7200 rpm, 128mb cache drive at $600 for the parity drive? I guess what I'm asking is whether or not increasing the speed of the parity drive would be beneficial?

 

Depends on your wallet. For my wallet I'd say "no". The speed improvement would be slight when you are doing multiple simultaneous writes to the array, or writing a a data disk with the same specs (if you had one).