garycase

... and I now carry a 128GB USB flash drive in my pocket (on my key ring) => that's 4,923 times as much storage as that first hard drive had; and it cost me about 1/100th as much

Yes, the question is do you get the same 4 corrections the next time you run a correcting parity check.

My first hard drive was a 14" Seagate 26MB unit for my CompuPro CP/M system -- after a 10% discount it cost me $4500. And yes, I too thought it was an amazing amount of storage that would take a long time to fill up !!

That's not an unrealistic approach. In fact, I still have one server I leave on v5.0.6 because it's got a fairly low-end CPU (Atom D525) and the CPU demands of v6 are appreciably higher than with previous releases. That particular server is rock-solid and will likely serve me well for years to come ... and will almost certainly never be upgraded. My other servers are all on v6, as they have significantly more "horsepower". I agree that v6 has some very nice feature improvements; but quite candidly v5.0.6 with UnMenu does everything I need for a NAS, and I don't miss any of the newer features with it. But since you're building a new server, you'll undoubtedly have plenty of "horsepower" for v6, so that's what I'd recommend -- primarily for the dual parity support; but also because it does have an improved GUI.

That would mostly depend on the number of disks you intend to use. What about 10 disks, or 20? It's a good feature no matter how many disks you have; but certainly with 10 or more it's a very good idea. The primary benefit is that if you have a disk fail, and are doing a rebuild of that failed disk, the rebuild will still be successful if a 2nd disk fails while it's in progress. The fact the system is dual fault tolerant does NOT mean you should wait for 2 failures before replacing the disks !! Remember that RAID -- regardless of how fault-tolerant it might be -- is NOT a substitute for backups. You should still have backups of all of your data (or at least any data you don't want to lose). Some folks don't bother with backups -- a risk I certainly don't recommend -- but if you don't have backups I'd think you absolutely would want dual parity ... no matter how many data disks you might have.

... also, for Justin: It's been over two years since this thread was active. Did you finally figure out a "fix" that worked?

When you run a correcting parity check, do you get the same 4 errors "corrected" every time ??

I'll pass on discussing YOUR purchases with YOUR wife I actually agree with your comment that you should "... buy things based on need vs want ..." ==> within reason. But if it's clear that your FUTURE needs are clearly going to require significant storage increases, then it makes sense to buy a parity drive that will support those; and not put yourself in a situation where you have to upgrade your parity drive multiple times. On the other hand, as I also noted, it makes no sense to buy, for example, a 6TB parity drive unless you are only going to buy 6TB data drives in the future. Buying a 6TB parity drive; and then adding 3TB or 4TB data drives to your array, simply doesn't make sense. Fortunately my spouse doesn't give a rip what I spend on my hobby

Note that the BackBlaze statistics are for the desktop DM series drives -- traditional PMR drives, but NOT the NAS units discussed in this thread. So their statistics don't apply to these drives. I DO think Seagate's NAS drives are very reliable, but you can't use the BackBlaze stats to back that up.

Agree r.e. a bit of conservatism on drive sizes. I see a lot of folks buying 6TB (or even 8TB) parity drives with an array full of 1, 1.5, and 2TB drives. This is, IMHO, a waste. If the array is providing the storage you need; and you want the ability to expand a bit, moving to perhaps a 4TB parity drive would seem to make more sense. ... you'll ultimately be able to better than double the storage capacity in the system with 4TB drives, and won't have one (or two, if you choose to use dual parity -- a good idea, but not something everyone will do), 6TB drives invested in providing fault tolerance for drives 1/3rd that size (or smaller). On the other hand, if you've suddenly changed your usage scenario from an occasional collector of music CD's to an avid ripper of BluRay disks, then you may need to start moving rapidly to the much larger drives -- and if that's your plan, then of course you should get a parity drive that will support the size drives you plan to buy. In other words -- you need to THINK about how your server needs to grow; and be sure any increase in the size of your parity drive reflects what you are realistically going to buy in terms of drive sizes. r.e. the "effective cost per TB" of a new drive, where you compute the cost purely based on how many TB you're adding. This is, of course, a valid argument -- BUT there are two other things to consider: (a) If you primarily just replace drive when they fail; then it's irrelevant -- you're adding the total capacity of a new drive when you do the replacement, since you need to buy a new drive anyway; and (b) using a larger parity drive (or drives) also means you're investing a higher cost in providing fault tolerance, since you have to buy one (or two) of that size drive to provide that feature. e.g. using Franks's earlier example of moving to a 3TB parity and upgrading 2TB drives. (a) If you simply bought the 3TB parity; then replaced 4 2TB drives with 3TB drives; it would indeed be rather expensive compared to using 6TB drives. You could, as Frank noted, use a 6TB parity and replace one of the 2TB drives with a 6TB drive and achieve the same 4TB of increased space in your array. It's not as bad as it sounds, however, when you look at the actual costs. Using today's Amazon prices, 5 3TB WD Reds would cost $545 ($109 ea); whereas a pair of 6TB Reds would cost $462 ($231 ea). So yes, if that was what you were doing, it would cost $83 more to do that upgrade. However, if you were also moving to dual parity, the economics of that would be different -- you'd need 6 3TB drives ($654) or 3 6TB drives ($693) to do the same thing. (b) If you were just upgrading parity to prepare yourself for future capacity growth, then you'd either be spending $109 for a 3TB drive (or $218 if you want dual parity) or $231 for a 6TB drive (or $462 if dual parity) -- and gaining NO space at this point. Then, when you needed additional capacity, or if a drive failed, you'd be buying additional drives. Note, however, that if a drive fails, the replacement drive you're buying is effectively providing ALL of its capacity to the array, since the failed drive is no longer usable. Don't misread my point => I DO think it makes sense to buy as large a parity drive as you think you might reasonably need. But be sure you THINK about what is really reasonable for your particular situation. The question I'd ask yourself is: If I buy an X TB parity drive, am I going to buy X TB data drives from now on? If the answer is no, then X is probably too large a value

How do we get dual parity now? Edit: I just saw it's a beta feature in unRAID Server Release 6.2.0-beta18. Does it seem reliable yet? As trurl noted, 6.2 hasn't been Beta for several months. The current stable release is v6.2.4 => and Yes, dual parity is very reliable, and clearly provides MUCH better fault tolerance than a single parity system.

... forgot to add: The second parity disk, if you upgrade to v6.2 and decide to use dual parity, uses a more complex parity calculation but the basic rule is still the same: i.e. it has to be as large as the largest data disk. So if you DO decide to go with a dual parity setup, simply use two parity disks of the same size.

The first parity disk uses a simple longitudinal even parity calculation => just XOR's all the bits in any given bit position across all drives and forces the end result to be a zero (thus "even" parity). The Wiki has a more detailed discussion of this, but if you understand basic parity calculations, it's very simple. So the SIZE of the parity disk is completely irrelevant in terms of how much "protection" you get => the only constraint is it must be at least as large as any of the data disks (else there wouldn't be a parity bit available to protect some of the data). So if all your disks are 1.5TB, then a 1.5TB parity disk provides exactly the same protection a 6TB parity disk would ... the extra 4.5TB of space on the parity disk would effectively be unused [it would actually still be "checked" and set to the appropriate parity to protect itself]. Your parity checks would, however take a good bit longer, since a parity check will check the complete size of the parity disk. What a larger parity disk does is provides you with the ability to use data disks of that size. So if you had a 6TB parity disk and one of your 1.5TB disks failed (or you just wanted to upgrade it to a larger size), you could replace it with a 6TB data drive. In other words, it doesn't make sense to use a larger parity disk than the largest data drive you plan to use => but be careful when deciding what that might be, as data needs have a way of growing much larger than you might have expected

Your ambitions are too low ==> ask for an 8TB drive

Yes, you can create partitions and assign logical drive letters to them. Just go to Disk Management -- right-click on "This PC", select Manage, then click on Disk Management. If you are restructuring a disk, you'll likely want to first delete all of the partitions on the disk (THIS WILL DESTROY ALL DATA ON THE DISK); then you just right-click in the unallocated space; right-click; and create a new volume. You can name it, assign a drive letter, etc.

This isn't the right thread for a detailed discussion r.e. file system conversion -- just search for "XFS conversion" if you want to read more details. But I'll toss in a couple basic thoughts ... => XFS is the recommended file system for new drives you add to your array, but there's no real reason to convert current Reiser volumes, especially if they're essentially static (i.e. filled with media and rarely changed or written to). => Reiser does indeed have better recovery tools in the event a Reiser volume should get corrupted -- there are plenty of examples of just how good they are on this forum. => If you have a very active volume -- lots of writes, deletes, updates, etc. -- it's probably a good idea to convert it to XFS, as XFS doesn't have the issue Reiser does with slow write performance as a disk gets close to full. If you DO decide to convert your data, be VERY careful. Read the threads about this CAREFULLY and be CERTAIN you understand what the "user share copy bug" is and that you absolutely avoid it -- it will result in complete data loss if not.

A bit of "out of the box" thinking r.e. how to eliminate the performance hit while still leaving the SSDs in the array ... [Note: I have NOT tried this, but it certainly SHOULD work as long as you get the 2 spanned drives for parity in the right order, so the "SSD part" is matching your SSDs] (1) Create a parity "drive" using a 525GB SSD and your current 6TB parity drive spanned ("BIG" setting) using this: https://www.amazon.com/StarTech-com-Internal-Connector-Controller-S322SAT3R/dp/B00M77UNDI This would give you a 6.525TB parity "drive", where the first 525GB was physically on an SSD. (2) Now the first 525GB of writes to any disk in the array will be using the SSD part of your parity "drive" => so if you have 525GB SSDs in the array, there won't be use of the 6TB spinner for the associated parity writes.

As Johnnie noted, using SSDs for storage devices is okay; BUT if you're using them for applications that will be doing frequent writes, the writes will be limited by the speed of the parity drive (in addition to requiring 4 I/O's per write) => so using them as unassigned devices is a better idea. You can back them up to the array periodically; but the normal operation won't have the limitations of array write speeds.

The special during v6 development wasn't really a "sale" on the key, so much as an inducement to get more Beta testers using the product. You had to already have an UnRAID license; and even then it was a limited time offer to induce more folks to set up a 2nd server to help test v6, which was in Beta for a VERY long time. Things have changed; v6 is very stable and is developing nicely with far more frequent releases than previously; the key replacement policy has been both formalized and GREATLY simplified (easy to do online in real time); and I wouldn't anticipate any further price reductions in the key pricing, as it's really quite reasonable for what you get with the latest version.

That WAS a very good deal on the 8-core 2750 setup => but it's already sold. I've had no problem maxing transfer speeds on a D525 setup; but I DID leave it on v5, which is likely why. I know these little Atoms can struggle with the CPU demands of v6.

For a mITX build I don't think the lack of 10Gb or M.2 support is really an issue. It WOULD be nice to have 8 ports instead of 6, but the reality is the PCIe slot can easily take care of that => with either an M.2 card with a spare SATA port; or simply a 2 or 4 port card with a standard SSD. Newegg's status is still "out of stock" ... and they don't have an "auto-notify" option now => this generally means they're not likely to restock; but not always. I'll keep an eye of it -- not sure I really need another server at this point, but that IS a very compelling deal.

Definitely a VERY nice improvement. If I had seen that deal I'd have jumped on it (it's now sold out). => r.e. "... (this one even has active cooling!) ..." ==> You say that like it's a good thing. Most of us would agree that passive cooling is MUCH nicer [totally quiet; never fails (fans DO fail); uses NO power] IF (and here's the catch) a passive cooler can provide adequate cooling for your CPU. In this case, that is apparently not possible, so there's a fan. I'm sure it's still very quiet, but the fact there's a fan is NOT an improvement -- just a necessity. => A 10X gain in PassMark is certainly nice. In addition, the Xeon supports the AVX 2 extensions that make the 2nd parity calculation much more efficient (the Atom does not). So overall it's a MAJOR boost in performance. => In addition to the CPU performance boost, you also just switched to registered (buffered) ECC memory -- another very nice improvement. => I also have a D525-based server (Supermicro board), and it works just fine as a basic NAS. I DID avoid doing the v6 upgrade, based on the increased CPU demands that were obvious from my experimentation on a couple other systems [i have 4 UnRAID setups -- 3 I use, and one I "experiment" with]. So my D525 system is still on v5 => but since it's used purely as a NAS that's fine, and it serves files very well -- I can easily exceed 100Mb/s on reads from the server; and writes to it (no cache) are in the 40-45Mb/s range. In fact, if I had bought one of these Xeon boards, I'd likely have not replaced my D525 unit -- I'd have just built another mini-ITX box with the Xeon. => landS's comments r.e. tweaking to get a D525 to run well with v6 are interesting. It seems you can even run dual parity v6 on the D525 without too much performance degradation. But my D525 will happily remain on v5 [it's completely backed up to my dual-parity server anyway ]

Agree -- if you're using it purely as a NAS, virtually ANY modern CPU will do fine. I have 3 servers (plus a 4th for "experimentation" that doesn't contain any data I care about) => and my favorite of the bunch (not the highest capacity) is my trusty old SuperMicro X7SPE-HF-D525 based unit ... which uses an Atom D525 processor. Completely quiet, VERY low power (idles at 18w with 6 drives), and as a NAS it's as fast at file serving at any of my other much-more-powerful servers. I would not, however, upgrade this one to dual parity, as the impact of that 2nd parity calculation is indeed a significant bump in CPU demands. But I'd say any CPU that scores 2000 or better on PassMark will be just fine for NAS duties. Even low-end current generation Pentium and Celeron CPU's score well over that. Note, however, that it's not unusual for folks to later decide to "do a bit more" with their UnRAID systems -- so you might want to look at CPU's in the 4000+ PassMark range just to provide a bit of "headroom", in case you later decide to stream a transcoded Plex stream or two

The RM550x is fine for this build.

Linux Reader is actually very simple -- you can easily access the files from any Windows utility that uses an Explorer interface.