October 21, 200916 yr For a good month now, I've been trying to specify (and get built) an unRAID box from LimeTech; the biggest 'hurdle' has been the hard drives, particularly as many manufacturers have recently come out with low cost higher capacity, plus 'green' types, during this period. I originally thought, and still feel, that a top-line 2TB as the parity drive is the way to go, irrespective of cost (7200rpm/64MB cache). Low cost data drives 2TB as well, as many are now in the market. Just recently (like, the past week), LimeTech has added the Seagate Barracuda LP ST32000542AS to their small buy list, so that is 'good to go', but in the email back and forth, neither any Seagate (Barracuda) or WD (Black) drives (for the parity) were sprinkled with 'holy water' by LimeTech ('too new' although they've been out for a year or more). The recent release, and cost per GB, of 2TB drives by both Seagate and WD, brought up another question; with a 16 drive system, that means 15x2TB, or 30TB total of data storage, and can unRAID handle that? This question comes up as I believe (or I've been told by Linux 'experts'), that the Linux core can only address 20GB of HD space (I know ext3 can address only 16TB...). I do see that the filesystem used on the 'per disc' part of unRAID is ReiserFS. which does limit the size of each disc to 16TB (a restriction that 'may' present a problem in around what, 10 years?), but haven't tripped across the total system size as of yet. So, two big questions need to be answered, without the emailing back and forth (which has proven to be a bit much): 1. Why no 'approved' HD list on this site? Every competitor has large lists, from every manufacturer, of 'supported' drives, including user lists beyond those of the recent major's (usually older types, precisely the ones unRAID would be targeted at). 2. What is the maximum total data drive space the system will support? (with the new 2TB drives now flooding the market at really good prices....). 15 data drive = 30TB. Any ideas are welcome. I've put this off for too long already, and have had to buy several drives to overpopulate my desktops to hold my 'junk'.
October 22, 200916 yr This question comes up as I believe (or I've been told by Linux 'experts'), that the Linux core can only address 20GB of HD space This is either untrue, or the truth is much more complicated that that simple statement. /dev/fs0 89T 39T 50T 44% /mountpoint Granted the above is not DAS but it is still a single fs being addressed by the kernel in some fashion. The total capacity *is* however DAS via SAN luns at one point in the process - all of which connected to a linux box. (I know ext3 can address only 16TB...). I do see that the filesystem used on the 'per disc' part of unRAID is ReiserFS. which does limit the size of each disc to 16TB (a restriction that 'may' present a problem in around what, 10 years?), but haven't tripped across the total system size as of yet. This might have an implication for user shares - although I'm pretty sure it won't as they're not a 'true' filesystem. So, two big questions need to be answered, without the emailing back and forth (which has proven to be a bit much): I'll do my best given no one else has chimed in! 1. Why no 'approved' HD list on this site? Every competitor has large lists, from every manufacturer, of 'supported' drives, including user lists beyond those of the recent major's (usually older types, precisely the ones unRAID would be targeted at). There are two things at play here and, really, you will have to revert to email to limetech for a definitive answer as that is the only formally supported mechanism. Limetech obviously have to cover themselves when selling you the drives direct as they will be potentially liable for replacements due to failure etc. As such they will be conservative on what models they deem reliable and well enough supported by linux to consider. The rest of us generally just throw any old disk in. So long as theres not a global problem with a specific disk or brand (the recent 1.5 TB seagate firmware issues being a good example) you should be fine. We as forum users don't give you warranty though... There are many users here running 2TB disks. 2. What is the maximum total data drive space the system will support? (with the new 2TB drives now flooding the market at really good prices....). 15 data drive = 30TB. I honestly don't know the answer to this. Based on the above df output, and the fact that the new larger drives on the market and the fact unraid itself has recently had a bump on the number of disks it supports I would suggest you'll be fine. Some people here run large arrays so they might be able to chip in with how close to the above limit they've come. Any ideas are welcome. I've put this off for too long already, and have had to buy several drives to overpopulate my desktops to hold my 'junk'. I understand your concerns but none of the above really holds you back from starting with unraid - one of it's principle points is that you can add as you go with no problem. The several drives you've purchased already could have been or could right now be used to kick start your unraid box. Unless you plan to fully populate all the drive slots with 2TB disks in one go. Which would not be an economically sensible route to go down unless you had an immediate 30TB of data to put on the array in one go (based on the fact that drive prices will continue to go down in the near future). I do understand your reluctance to start down this route without being sure if your end goal can be achieved however. I hope this helps.
October 22, 200916 yr I can't image a hard drive not working unless the manufacturer really messed it up. boof is dead on here. I'm no expert but based on my experience I suspect max size is related to the the maximun a file system can utilize as a single volume. Example.. I recently built a Windows machine using HP SAS card single volume total size ended up to in the range of 6TB. Being the boot partition Windows had an issue with it. Once I broke out the drives and partitioned them individually the issue was gone. The boot partition has some max size limits. By setting up 2 volumes , a boot mirror and a raid 5 for data, I worked around the size limitations. If I had reached the max on the data partition I would have to set 3 partitions, one boot and 2 for data. In Linux with the max as stated, if you added 10x 2TB drives in a HARDWARE raid, I would expect an issue. It can be easily worked around as I stated above. Just a mater of configuring the paritions with the limits in mind. UnRaid uses the drives as indivudals so if we could buy a 17TB drive today I would expect a problem. Granted I could be misreading this entirely but I really don't see an issue. By the time the drive technology surpases present limits we would be on a new file system. Support for Ext4 is in the linux kernel now. Ext4 has an max volume size of 1 Exbibyte. I suspect you'll be fine unless your planning to download the internet
October 22, 200916 yr For a good month now, I've been trying to specify (and get built) an unRAID box from LimeTech; the biggest 'hurdle' has been the hard drives, particularly as many manufacturers have recently come out with low cost higher capacity, plus 'green' types, during this period. I originally thought, and still feel, that a top-line 2TB as the parity drive is the way to go, irrespective of cost (7200rpm/64MB cache). Low cost data drives 2TB as well, as many are now in the market. The best "price-point" for hard disks constantly changes... I purchased a pair of 2TB drives a week ago for $159.99 each. There are NO perfect disk drives. They will ALL eventually fail, it is just a matter of time. It is exactly why we have RAID arrays. We expect the failures, and can deal with them. unRAID make it much easier than some other implementations of RAID. Many require identical drives. Imagine if I tried to find a replacement for my 750Gig IDE drives today... they might be available somewhere, but only at a very high price compared to the 2TB drives. (I just looked.. there are two vendors I could find with a quick search... one price was $199.98, the other $232.00) No way I'll ever replace them with identical sized drives. Today's best size/price-point is going to be different than next year's. The drives I purchased have a three year warranty... some drives have as much as a 5 year warranty. My original two 500Gig drives have over 34,000 hours on them (3.88 years) and are sill going strong. Just recently (like, the past week), LimeTech has added the Seagate Barracuda LP ST32000542AS to their small buy list, so that is 'good to go', but in the email back and forth, neither any Seagate (Barracuda) or WD (Black) drives (for the parity) were sprinkled with 'holy water' by LimeTech ('too new' although they've been out for a year or more). Lime-tech is just providing the drives they can purchase at a reasonable price and is re-selling them a near zero markup. (by the time they burn them in for 48 hours and pre-clear them, they can't be making much profit. The advantage is they might identify drives prone to an early failure... you can do the same burn-in and pre-clear on any drive you purchase. The question is do you spend a few dollars more and let Lime-technology do it for you, or do it yourself. When I purchased my server from Tom, nearly four years ago, I purchased a pair of 500gig drives from him (biggest available at the time in the market) His price at that time was actually better than I could get from newegg, etc. I seem to remember they were over $300 each. Obviously, he purchased in bulk quantities and got a discount. The recent release, and cost per GB, of 2TB drives by both Seagate and WD, brought up another question; with a 16 drive system, that means 15x2TB, or 30TB total of data storage, and can unRAID handle that? This question comes up as I believe (or I've been told by Linux 'experts'), that the Linux core can only address 20GB of HD space (I know ext3 can address only 16TB...). I do see that the filesystem used on the 'per disc' part of unRAID is ReiserFS. which does limit the size of each disc to 16TB (a restriction that 'may' present a problem in around what, 10 years?), but haven't tripped across the total system size as of yet. Actually, the current drive limit in unRAID is 20 drives (19 data drives). Therefore, today you can have as much as 38TB of combined storage. Yes, today ReiserFS has a limit on 32 Bit operating systems of 17.2TB per-file-system. If disk drives sizes double every few years, we'll get 16 TB drives in 6 years or so IF THE MARKET EXISTS for them. At that point, you'll be able to get 300TB in a single unRAID server and spend a few days to perform a parity check even at 100MB/s. So, two big questions need to be answered, without the emailing back and forth (which has proven to be a bit much): 1. Why no 'approved' HD list on this site? Every competitor has large lists, from every manufacturer, of 'supported' drives, including user lists beyond those of the recent major's (usually older types, precisely the ones unRAID would be targeted at). Linux has no dependency on any hard disk, size, or brand. It must just be reliable and of a type supported as an IDE or SATA device. I've got an old 8Gig Quantum drive in my server I use for testing. It works just as well as the newer drives... it just does not hold anywhere near as much data. The only drive that would not be recommended would be those with either a horrible mechanical failure rate or buggy firmware (some Seagate 1TB and 1.5TB drives had firmware that caused time-out errors or reported a drive size of 0 bytes... those drives are NOT recommended, even after updating the firmware, they still caused time-out errors for some users. That is why Lime-tech did not recommend their use in unRAID.) 2. What is the maximum total data drive space the system will support? (with the new 2TB drives now flooding the market at really good prices....). 15 data drive = 30TB. Actually, it is 19 data drives = 38 TB. Any ideas are welcome. I've put this off for too long already, and have had to buy several drives to overpopulate my desktops to hold my 'junk'. The beauty of unRAID is the ability to purchase the drives one at a time when they are on sale. You can expand the array or upgrade at will. Purchase what you need now... the drives will get cheaper over time, unRAID and Linux will evolve over time. (even the 750Gig drives I described above are cheaper, or nearly identical to their original prices when I first purchased them) The drives will get bigger over time. Expect to replace the parity drive at some point when 4Gig drives become available... if you need the space. I've seen corporate databases in fortune-500 companies where the size was under 30TB. 30TB is a huge number of movies... even with full-bit-rate HD content, it probably represents 1000 movies or more. It will be a tough decision as it will be at a premium price initially. Joe L.
October 22, 200916 yr Author First, initial build will be 3 drives (parity + 2data), simply because the 2TB drive costs are VERY soft (they've dropped almost 50% just in the last 3-4 months with new models), and the cost/GB curve v. the 1.5TB drives has just about been reached or exceeded (2TB would have to reach $160), but as I don't have an 'old box of drives' to use, and I'm not about to 'raid' my 2 current RAID5 arrays for drives, it's all 'new' stuff. If anyone has run a cost curve on the total unRAID system, you'd find that unRAID v. widely available 'stand-alone' NAS commercial systems (RAID5/6), that unRAID doesn't begin to achieve true cost benefit until the number of data drives exceed ~10. Of course, this is with all new components, I don't have a box of old drives, motherboards or multi-port sata cards, to bash together something, and don't intend on spending loads of time troubleshooting the problems that older components would introduce. The problems that new ones would be enough to handle... As far as 'build as you go', all modern RAID boxes now support the same flexibility (in some cases even more) that unRAID does, the only limitation is the number of drives supported in a single box/array, and there are easy (and built-in) ways around that, to 'stack' systems until one is blue in the face. That's why the first thing I did was run a cost curve to try and get a grip on what the initial v. out years cost would look like. I'm somewhat amused by daniel.boone's comment about d/l'ing the internet In the last five years my desktop systems have all gone from <500GB to between 4-8TB each. Since the introduction of perpendicular recording (and widespread adoption of SATA) a few years ago, it appears that drive size has reached a plateau at the 2TB level (cost is the target now), and we are about to see a new generation of iinterfaces in the next round (6GB SATA v. USB3). But SATA I/II will be around for a good while yet. The old adage that one can't ever have enough storage space is many times more true today than ever, with streaming 50GB HD movies (and the storage space that eats up) on the table. Unless we're all sitting here 5 years from now scrapping together the pennies to buy a loaf of bread, 100+TB home digital storage systems will be 'the next new thing'. That a fair percentage of that is already here, means it's time to get cracking.
October 22, 200916 yr If anyone has run a cost curve on the total unRAID system, you'd find that unRAID v. widely available 'stand-alone' NAS commercial systems (RAID5/6), that unRAID doesn't begin to achieve true cost benefit until the number of data drives exceed ~10. Of course, this is with all new components, I don't have a box of old drives, motherboards or multi-port sata cards, to bash together something, and don't intend on spending loads of time troubleshooting the problems that older components would introduce. The problems that new ones would be enough to handle... I think this depends on what you define as a cost benefit. My unraid build including case, parts and ~3TB of space was cheaper than an empty 4-5 Bay NAS appliance from soneone like ReadyNas or Qnap. I'm already way ahead on the cost curve. Theres also the point that unraid doesn't compete with those products on feature set given it doesn't do parity striped raid. Thus the nature of data protection is very very different. And for many one of the big advantages of unraid. As far as 'build as you go', all modern RAID boxes now support the same flexibility (in some cases even more) that unRAID does, the only limitation is the number of drives supported in a single box/array, and there are easy (and built-in) ways around that, to 'stack' systems until one is blue in the face. That's why the first thing I did was run a cost curve to try and get a grip on what the initial v. out years cost would look like. I'm not aware of any modern raid box that does unstriped parity? WHS is the closest I can come when coupled with an HP home server hardware build or similar. But even then it's not parity, but it's own file level duplication concept. Drobos and the likes, I believe, will penalise you much more heavily than unraid for using mismatched drives by essentially discarding space as unusable in a 'single small drive amongst many large drive' scenario. Perhaps you've found a better NAS box than I've come across? I'd be very interested in finding out as I'm always evaluating storage solutions. I'm happy to move from unraid if I find a better solution (and I'm coming close to that) I'm somewhat amused by daniel.boone's comment about d/l'ing the internet In the last five years my desktop systems have all gone from <500GB to between 4-8TB each. Since the introduction of perpendicular recording (and widespread adoption of SATA) a few years ago, it appears that drive size has reached a plateau at the 2TB level (cost is the target now), and we are about to see a new generation of iinterfaces in the next round (6GB SATA v. USB3). But SATA I/II will be around for a good while yet. I could be cynical and suggest the 2TB limit is more down to windows / GPT limits. New interfaces won't fix that. The old adage that one can't ever have enough storage space is many times more true today than ever, with streaming 50GB HD movies (and the storage space that eats up) on the table. Unless we're all sitting here 5 years from now scrapping together the pennies to buy a loaf of bread, 100+TB home digital storage systems will be 'the next new thing'. That a fair percentage of that is already here, means it's time to get cracking. Very true and I don't think any of us would be here discussing it if that wasn't the case. But for many of us unraid is the most cost effective way of getting there - new or old hardware. In 5 years time the storage game will have changed many ways beyond what is now available anyway and I'd be inclined to suggest that at that point whatever solution you settle on now might not be the best offered or suit you anymore come that time. Which is why it's also important that any solution you use now allows easy migration away - which unraid does. I'm less sure of other NAS applicances due to either RAID striping or proprietary filesystem tinkering. Some of what you've said (I may have misunderstood) makes me concerned that you may not be aware of unraids full feature set and so the inherent benefits and drawbacks it brings. Have you tried running the free copy to evaluate?
October 22, 200916 yr Author I think this depends on what you define as a cost benefit. My unraid build including case, parts and ~3TB of space was cheaper than an empty 4-5 Bay NAS appliance from soneone like ReadyNas or Qnap. I'm already way ahead on the cost curve. Theres also the point that unraid doesn't compete with those products on feature set given it doesn't do parity striped raid. Thus the nature of data protection is very very different. And for many one of the big advantages of unraid. Only cheaper if, again, you're using 'used' components. A good basic RAID box (5-6 drive) is about $800, and that is around the same cost as a good multi-PCIe (several x4/x8 slots) motherboard, multi-core cpu, ram, power supply, and case cost. Start adding multiport SATA cards in those PCIe slots, and the cost rapidly exceeds it. BTW, that 'target' crossover point (by my calculation, with drive costs) is around $3K. Below that point, the NAS boxes are cheaper per/TB, above it, unRAID continues it's cost curve downward. I've read a lot of the threads here on the unRAID protection scheme (having the single parity drive) and would say that if the system has an Achilles heal, that's it. A 'standard' RAID array using stripping with distributed (or multiple) parity (RAID5 and above), is a better system overall (with modern drives) than the unRAID, but that's perhaps just my opinion (but echoed by many here that would like the option for multiple/distributed parity added to the unRAID system). On the 'plus' side is the ability, matched only by the other Linux variants and the Drobo systems, to utilize a 'mix' of different drives and such. But, drive slots, even as plentiful as in the unRAID, are still a limited resource that needs the maximum utilization. So having a 'matched' drive set is just not that big of a deal when 2TB drives are <$180 ea. (!). Again, it all makes more sense if one is hacking together a collection of used parts, drives, and the like. It's been my experience that at the end of the day, any savings one realizes from such an approach is fairly rapidly lost by time, effort, and money replacing bad or variable components. But hey, that's me. The biggest plus of unRAID, that of being able to throw together a collection of odd drives, though, is rapidly disappearing. On might get an 'inner glow' about getting some use out of some drives one originally spent a dear amount of cash on in years past, but those ports aren't infinite (or of no/low cost) either, and it's a choice as to utilizing them or replacing them with newer/low-cost variants. Slice of my 'cost' calculations: W/2TB drives: NAS 'appliance' - 16TB RAID 5 (10 drives) = $2840 - $177.50/TB UnRAID - 16TB data drives + Parity/Drive (9 drives) - Base unit: $1400, + 9 Drives ($1620) = $3000 - $187.50/TB That's the 'crossover' point, or thereabouts. The 'appliance' cost curve is 'flat' at that point, more storage costs exactly the same, whereas unRAID continues to fall towards around $150/TB at the maximum supported drives (but of course is more expensive at the smaller end of the scale, IF one builds out the unRAID to support more drives/ports at the start, but the differential with more drive slots/sata ports is really minimal). Oh well, the part that worries me the most, again, is potential failure, and the grinding away of rebuilding a drive in the array. That's why the (perhaps) excessive attention to the parity drive in the unRAID, whereas with a RIAD5 array, that grinding and thrashing is distributed among all the drives in the system, and the bottleneck is distributed rather than relying on a single drive/port.
October 22, 200916 yr Only cheaper if, again, you're using 'used' components. A good basic RAID box (5-6 drive) is about $800, and that is around the same cost as a good multi-PCIe (several x4/x8 slots) motherboard, multi-core cpu, ram, power supply, and case cost. Start adding multiport SATA cards in those PCIe slots, and the cost rapidly exceeds it. BTW, that 'target' crossover point (by my calculation, with drive costs) is around $3K. Below that point, the NAS boxes are cheaper per/TB, above it, unRAID continues it's cost curve downward. All my parts were new. And my case gives me expanstion capability to ~15 drives. I would have to get a extra controller along the way but until then my 8 sata port motherboard does the job. Again, by the time I buy an empty and drive slot restricted dedicated NAS box I'm already up and running with a full unraid system and a handful of terabytes. If I'm ahead from the beginning I would have to do something daft to fall behind by that much over the lifetime of the server. You're also talking about high end motherboard, multi core cpu and multiple high end drive controllers. In many ways these are completely useless for unraid as it won't take advantage of them. Again this goes back to my concern if you understand the performance limitations of unraid. As there is no striping maximising the performance of all components with a view to peak drive throughput really doesn't apply as you're going to cap out far, far below a single drives write speed ultimately. Maximising bus throughput will help with full parity checks but these are only a periodic automated task so the benefit of throwing lots of cash / hardware at it doesn't make sense. I've read a lot of the threads here on the unRAID protection scheme (having the single parity drive) and would say that if the system has an Achilles heal, that's it. A 'standard' RAID array using stripping with distributed (or multiple) parity (RAID5 and above), is a better system overall (with modern drives) than the unRAID, but that's perhaps just my opinion (but echoed by many here that would like the option for multiple/distributed parity added to the unRAID system). On the 'plus' side is the ability, matched only by the other Linux variants and the Drobo systems, to utilize a 'mix' of different drives and such. But, drive slots, even as plentiful as in the unRAID, are still a limited resource that needs the maximum utilization. So having a 'matched' drive set is just not that big of a deal when 2TB drives are <$180 ea. (!). It depends how you define 'better'. I'm of the opionion that my data is safer with unraid than a raid-6 array. But the trade off is performance and flexibility by quite a drastic fashion. You're clearly talking about 'big booming' all your storage and full drives slots at once and getting them all matched. Otherwise if you're not, what happens in a years time when a 2.5 or 3Tb disk is more cost effective than a 2TB? do you still crippled yourself by buying the 2TB disk and forgoing the extra space / price sweet spot? This is where mixing and matching drives becomes as much a bonus as starting off with 'whatevers in the cupboard'. Again, it all makes more sense if one is hacking together a collection of used parts, drives, and the like. It's been my experience that at the end of the day, any savings one realizes from such an approach is fairly rapidly lost by time, effort, and money replacing bad or variable components. But hey, that's me. To many here it makes as much sense cobbling together parts or buying new. There are lots of people here running arrays with nothing but 1.5-2TB disks in them. I'm really not sure where your hang up on unraid being mostly useful with used parts is! The biggest plus of unRAID, that of being able to throw together a collection of odd drives, though, is rapidly disappearing. On might get an 'inner glow' about getting some use out of some drives one originally spent a dear amount of cash on in years past, but those ports aren't infinite (or of no/low cost) either, and it's a choice as to utilizing them or replacing them with newer/low-cost variants. As above, not at all. In a years time I can buy bigger, cheaper drives and just slot them in. I can keep doing that until I run out of slots or decide to recycle older smaller capacity drives with cheap big ones. Or be forced to do so as disks fail. Slice of my 'cost' calculations: W/2TB drives: NAS 'appliance' - 16TB RAID 5 (10 drives) = $2840 - $177.50/TB UnRAID - 16TB data drives + Parity/Drive (9 drives) - Base unit: $1400, + 9 Drives ($1620) = $3000 - $187.50/TB What NAS appliance is this? its clearly not a drobo / readynas etc that I've been assuming if it supports 10 drives. My frame of reference is clearly wrong. Oh well, the part that worries me the most, again, is potential failure, and the grinding away of rebuilding a drive in the array. That's why the (perhaps) excessive attention to the parity drive in the unRAID, whereas with a RIAD5 array, that grinding and thrashing is distributed among all the drives in the system, and the bottleneck is distributed rather than relying on a single drive/port. Unlike striped systems if a drive in unraid fails its recovered by parity or, if two drives fail, you lose only the contents of those drives. Not everything due to completely lost stripe. However rebuilding in unraid still (I believe, I could be wrong here) thrashes all the disks. As the parity disk still needs to read all the other disks parity to figure out the missing bit pattern on the replaced drive. Having never done this, I could be wrong,but in my head this is how it would work! I think we clearly disagree on many points so I'll bow out. I will leave you with my concern that you're misunderstanding some key unraid concepts and what they relate to in terms of performance and other issues. If you're planning on throwing a lot of money at high end hardware you *will* be dissapointed when your writes to the parity protected array cap out at 20 megabytes per second. If you re evaluate away from high end hardware to buying budget and frugally you will find your price comparison changes drastically down to where I, and most others here, see it. Good luck!
October 22, 200916 yr Oh well, the part that worries me the most, again, is potential failure, and the grinding away of rebuilding a drive in the array. That's why the (perhaps) excessive attention to the parity drive in the unRAID, whereas with a RIAD5 array, that grinding and thrashing is distributed among all the drives in the system, and the bottleneck is distributed rather than relying on a single drive/port. This should not worry you or anyone here. This "grinding away" of rebuilding a drive in the array is going to happen if you are raid5, raid6 or unRAID. Sometimes I shake my head at the misconception here that your drives are so fragile that they are going to fail because of a sequential pass through the whole drive. In fact this is a very gentle way to access the drive. There are some people who use such old drives they do not trust them. But that is their choice. With unRAID, if you loose parity drive, you rebuild parity from all the data drives. If you loose a data drive, you rebuild that drive with the remaining drives and the parity drive. Same thing if it were raid5. You would still pass through all the disks and all the sectors. I've read a lot of the threads here on the unRAID protection scheme (having the single parity drive) and would say that if the system has an Achilles heal, that's it. A 'standard' RAID array using stripping with distributed (or multiple) parity (RAID5 and above), is a better system overall (with modern drives) than the unRAID, but that's perhaps just my opinion (but echoed by many here that would like the option for multiple/distributed parity added to the unRAID system). The Achilles heel of the single parity disk is write speed. Same issue if it were RAID4. A 'standard' RAID array using stripping with distributed parity is better if you need very fast reads and writes. For my use it's actually not better. I don't want 16 drives spinning to access a movie. I don't want a filesystem that is so large, I cannot copy it to another drive in an easy manner. (or back it up to USB in an easy manner). but echoed by many here that would like the option for multiple/distributed parity added to the unRAID system). Not quite. What is asked for many times is RAID6 type protection so that two failed drives can be tolerated. The other request is multiple arrays. I cannot remember when anyone asked for distributed parity. If that is what is needed regular RAID5/RAID6 would suffice for them. unRAID's benefit is the JBOD approach to storing data in a protected manner. I really love this concept and prefer to have it this way. Having user shares present a combined "view" is an optimum approach.
October 22, 200916 yr I agree with boof, it seems like you are missing the point of unRAID. The way I see it, unRAID is intended to protect your data and allow great flexibility all on a very reasonable budget. Even if you aren't recycling old hardware, you can still save money by buying a low power CPU (a single core AMD Sempron will work fine), budget RAM, and Green or LP hard drives (as opposed to RAID class hard drives). Also, if you factor in 'riding the wave of sweet spots' (meaning today, you would buy a 1.5 TB drive, then in a few months, maybe a 2 TB drive, and so on) you will save lots of cash over time. Also, unRAID's lack of distributed parity is one of its greatest benefits, not its 'achillies heel'. This means that if for some reason your unRAID server went completely haywire, as a last resort you would be able to hook up the individual data drives to a standard Linux computer and run standard Linux data recovery tools on the hard drives. If the same thing happened with a RAID 5 or a RAID 6 server, you would loose all of your data and would have no option of recovering any of it.
October 23, 200916 yr Author Still don't see any answers to my 2 questions; guess I'll check back after the weekend and see if there any.
October 23, 200916 yr boof answered your 1st question. Joe L. answered your 2nd question. Reread the posts above & you will find the answers to your questions.
October 23, 200916 yr Still don't see any answers to my 2 questions; guess I'll check back after the weekend and see if there any. Your questions were answered back in JoeL's post above. unRAID can currently handle 19 data drives and with each of them being 2TB drives that is 38TB of unformatted space. I am sure that as drive sizes increase unRAID will adapt/move to a new FS so that those larger sizes can be supported. As it stands now, you can max out a machine and not have a problem. All the points made in this thread by the other posters are bang on. I choose unRAID because it alows me to recover some data from a 2 drive failure unlike raid 4/5. RAID six was an option when iI as looking but frankly the hardware card to do a RAID6 was to expensive for my taste. You need to recalc your estimates for building the unRAID server and not include drives in the price. the drive cost will be the same across the board so don't even worry about that cost. Take a look at the Pimp Your Rig thread to get an idea of some of the components others are using. There is also the Hardware Compatibility Page with a list of all the motherboards that have been used and proven to work. As you will undoubtedly notice not many of the boards on the list are "Server" class boards. I built my system for just shy of 300 (not including drives; which I bought 2x750GB and had 2x500GB). The board I chose was the Abit AB9 Pro and it can host 10 drives if I chose to use the external eSATA for an internal drive. Granted the AB9 Pro is a little more heafty on built-in ports but still, get a board with 6-8 ports and buy a controller card for the rest. The cost of an unRAID system does not have to be NEARLY as high as you make it out to be.
October 23, 200916 yr However rebuilding in unraid still (I believe, I could be wrong here) thrashes all the disks. I would really like to see people get past this concept of thrashing the disks during parity check/generation. Rebuilding/Recalculating parity does not thrash the disks. It reads each sector from each disk in a sequential manner. It's actually a gentle read of every sector cylinder by cylinder What causes more stress is if you start writing to the array on some of the disks. Now the heads have to jump back and forth to read/recalculate parity/rewrite sectors in manner that is not sequential. Thrashing is when you do so much activity and you make such little progress that it becomes almost useless. Frankly, if you lost a disk in a raid5/raid6 array, the system and disks would be quite busy too. Even more busy because all disks "have" to spin. If I loose a disk in unRAID, there is a potential that some disks do not have to be spinning if I'm reading from a disk that is not virtual.
October 23, 200916 yr However rebuilding in unraid still (I believe, I could be wrong here) thrashes all the disks. I would really like to see people get past this concept of thrashing the disks during parity check/generation. Rebuilding/Recalculating parity does not thrash the disks. It reads each sector from each disk in a sequential manner. It's actually a gentle read of every sector cylinder by cylinder What causes more stress is if you start writing to the array on some of the disks. Now the heads have to jump back and forth to read/recalculate parity/rewrite sectors in manner that is not sequential. Thrashing is when you do so much activity and you make such little progress that it becomes almost useless. Frankly, if you lost a disk in a raid5/raid6 array, the system and disks would be quite busy too. Even more busy because all disks "have" to spin. If I loose a disk in unRAID, there is a potential that some disks do not have to be spinning if I'm reading from a disk that is not virtual. Poor choice of words on my part and I was largely echoing the phrase used by the OP. I agree with you entirely. Apologies.
Archived
This topic is now archived and is closed to further replies.