February 7, 201511 yr ... So, parity will be a tad bigger than 8TB data drive, so there is a hope that this might actually work. Clearly that will work fine -- good thing the size works out (I just confirmed that this is universally true -- at least for the drives I checked (1TB, 1.5TB, 2TB, 3TB, 4TB, and 6TB in addition to this 8TB) => i.e. 2 x any drive has more bytes than the single drive of the same size. I suspect that's not at all coincidental Hi, I just did a few tests on my test server with two disk Raid 0 as parity drive and have no success. i tried 2 configurations: - 2x 1TB in Raid 0 and one 2TB drive - 2TB disk was larger in size than Raid 0 volume - 2x 2TB in Raid 0 and one 4TB drive - 4TB disk was larger in size than Raid 0 volume i'm using Dell Perc H710P as Raid controller. what i'm doing wrong?
February 7, 201511 yr Personally, I think contemplating using a Raid 0 for parity is the wrong way to do things. No matter what quality of drives you are using, your odds of a parity failure are going to skyrocket. Sure, you can use the absolute highest quality enterprise drives you can find to try and mitigate the situation, but the failure rate is still going to be higher. Not until you use a RAID 1+0 to mirror each drive in the set are you going to get the failure rate back down. But then you're using 4 drives to make up the parity drive. To each his own I suppose, but I don't think the risk / performance benefit is worth it on a production server
February 7, 201511 yr Therefore guessing. Only if you don't believe in the specifications that the manufacturer has released.
February 7, 201511 yr I am about to try this with 2 3T disks to create a 6T parity to use with new 6T HGST. My math shows the RAID0 will be larger, but until I do it I won't know for sure. These same 2 3T have been used for parity for quite a while and I've just slowly increased the size, and used the leftovers as a RAID1 mirrored pair that I used outside the array. If the RAID0 is just a tiny bit too small, you can create a small HPA in the data disks to make them slightly smaller. I used this technique a while back when unRaid was limited to 2.2T drives. I bought some 3T drives in anticipation of big disk support, and created an 800MB HPA to drop the capacity to under the magic 2.2T size until then. You'd probably need a tiny HPA to drop the size of your 8T drive under the 8T RAID0 size. A slight PITA but maybe better than the alternative! BTW my 3T parity drives are Hitachi - the most reliable drives on the planet. I would not have done it with Seagate or WD based on personal and documented failure rates of those brands. Squid is right that your failure rate is impacted, but choosing a drive with a much much lower failure rate evens the score. I personally would not want an archive disk as parity as the write performance hit would affect writes to even my high performance drives. But I understand the tradeoffs.
February 7, 201511 yr Still guessing. I have hitachi drives due to their reliability as well and feel uncomfortable using seagates.
February 7, 201511 yr If you are interested in creating HPAs look at THIS THREAD for a discussion of how to use the hdparm tool to create the HPA. Note the I'd recommend creating therm on a motherboard port. After they are created you can move the drives to a different controller.
February 7, 201511 yr Personally, I think contemplating using a Raid 0 for parity is the wrong way to do things. No matter what quality of drives you are using, your odds of a parity failure are going to skyrocket. Sure, you can use the absolute highest quality enterprise drives you can find to try and mitigate the situation, but the failure rate is still going to be higher. Not until you use a RAID 1+0 to mirror each drive in the set are you going to get the failure rate back down. But then you're using 4 drives to make up the parity drive. To each his own I suppose, but I don't think the risk / performance benefit is worth it on a production server I had a 20 drive server. The drives that failed were the data drives, not the RAID0 parity drive. In fact, the RAID0 parity set never failed. Maybe I was just lucky in the 5 years of it's use. What I liked about the Areca controller is the built in alarm and bios automatic spin down. While I knew there was the loud alarm, I also never heard it from a failure. While it seems obvious that a two drive set is likely to fail more often, there has to be consideration of data drive reads vs data drive/parity writes.
February 7, 201511 yr If you are interested in creating HPAs look at THIS THREAD for a discussion of how to use the hdparm tool to create the HPA. Note the I'd recommend creating therm on a motherboard port. After they are created you can move the drives to a different controller. Thanks for an idea about HPA, will take a look how to do that and yes, the difference in size was really small in both tested variants.. i was wondering about lack of an option not to use a whole drive - we all know how drive performance is affected when we read/write to inner sectors..
February 7, 201511 yr I had a 20 drive server. The drives that failed were the data drives, not the RAID0 parity drive. In fact, the RAID0 parity set never failed. Maybe I was just lucky in the 5 years of it's use. What I liked about the Areca controller is the built in alarm and bios automatic spin down. While I knew there was the loud alarm, I also never heard it from a failure. While it seems obvious that a two drive set is likely to fail more often, there has to be consideration of data drive reads vs data drive/parity writes. Its just statistics. If you assume that all the drives are equivalent in terms of reliability then with a RAID-0 parity set it is exactly twice as likely to fail as the other drives.
February 7, 201511 yr I had a 20 drive server. The drives that failed were the data drives, not the RAID0 parity drive. In fact, the RAID0 parity set never failed. Maybe I was just lucky in the 5 years of it's use. What I liked about the Areca controller is the built in alarm and bios automatic spin down. While I knew there was the loud alarm, I also never heard it from a failure. While it seems obvious that a two drive set is likely to fail more often, there has to be consideration of data drive reads vs data drive/parity writes. Its just statistics. If you assume that all the drives are equivalent in terms of reliability then with a RAID-0 parity set it is exactly twice as likely to fail as the other drives. If it's used on a smaller percentage of operations, i.e. only writes or reads during a parity check, those statistics change. On top of that, in this given scenario where there is a known performance penalty of the drive's architecture, having a spare 4tb spindle isn't all that expensive. Given the choice of RAID0 using older burned in and tested drives vs the unknown of the new drive's architecture, I think it's worth the risk. Given the choice of a single 8TB drive of high reliability without the performance penalty, I would go that route.
February 7, 201511 yr If it's used on a smaller percentage of operations, i.e. only writes or reads during a parity check, those statistics change. No it doesn't. Under that argument, you would have no problem setting up an array using high quality data drives and a parity drive that you found in a dumpster. But it's really a moot point. There is no real-world statistics of the failure rate of 8TB drives, nor will there be for quite a while. Performance is another consideration, and I'm not talking about performance vs reliability. I'm merely speaking to reliability
February 7, 201511 yr If it's used on a smaller percentage of operations, i.e. only writes or reads during a parity check, those statistics change. No it doesn't. Under that argument, you would have no problem setting up an array using high quality data drives and a parity drive that you found in a dumpster. You're putting words and assumptions into the counterpoint. I wouldn't use low quality drives no matter what. Nobody would. There are a few of us who have done the RAID0 parity with quality drives on a quality controller. I would invite them to report failure rates vs the data drives. With a good maintenance review and backup plan, I think the points loose weight. Yes, RAID0 has a greater chance of failure, but it also depends on how many writes vs reads. i.e. The RAID0 array can likely be in a spin down state for quite a while.
February 7, 201511 yr You're putting words and assumptions into the counterpoint. I wouldn't use low quality drives no matter what. Nobody would. Np. I'm going to drop this before it becomes too argumentative and OT. But, your argument that the statistics change based upon usage patterns is also misleading. The parity disk is no more and no less important than any other drive in the array for recovery purposes. My dumpster argument was merely to emphasize that point. My only assumption for all of this is that the reliability for all drives is equivalent. And that right now is an unknown.
February 7, 201511 yr There are a few of us who have done the RAID0 parity with quality drives on a quality controller. I would invite them to report failure rates vs the data drives. I have never had a parity drive fail. My data drives have also been pretty reliable, but I have had 3-4 failures with emphasis on Seagate and WD (in that order). I have used the RAID0 parity on the Areca ARC-1200 controller for at least 2-3 years (following Weebo's example) with the same pair of 7200RPM 3T Hitachi's, and they have been 100% perfect. I think people tend to think of the parity drive as being the most important drive in the array. I don't agree. The parity drive only has to work when another drive fails, and at that time EVERY other drive in the array has to be working perfectly also. Although all drives are important, I feel that the parity drive is the least.
February 7, 201511 yr There are a few of us who have done the RAID0 parity with quality drives on a quality controller. I would invite them to report failure rates vs the data drives. I have never had a parity drive fail. My data drives have also been pretty reliable, but I have had 3-4 failures with emphasis on Seagate and WD (in that order). I have used the RAID0 parity on the Areca ARC-1200 controller for at least 2-3 years (following Weebo's example) with the same pair of 7200RPM 3T Hitachi's, and they have been 100% perfect. I think people tend to think of the parity drive as being the most important drive in the array. I don't agree. The parity drive only has to work when another drive fails, and at that time EVERY other drive in the array has to be working perfectly also. Although all drives are important, I feel that the parity drive is the least. This has been my experience as well. Data drive fails, I'm very concerned. Parity drive fails, I make sure my backups are up to date and replace the parity drive.
February 7, 201511 yr My only assumption for all of this is that the reliability for all drives is equivalent. And that right now is an unknown. Plus the assumption that the usage patterns are the same. That is surely not going to be the case in a mostly read environment. my RAID0 set was in spin down state most of the time. Eventually I had to write a job that would keep it spinning from 6am to 11pm. Even then it didn't fail after 5 years of use. I had the same FUD to a fault. I kept a couple spare spindles, burned in, tested and in carriers expecting a failure. Frankly I was surprised, the drives outlasted the server itself.
February 7, 201511 yr ... The parity drive only has to work when another drive fails, and at that time EVERY other drive in the array has to be working perfectly also. Although all drives are important, I feel that the parity drive is the least. I agree that ALL drives are important, but it's certainly NOT true that "... The parity drive only has to work when another drive fails ...". The parity drive is involved in EVERY write to the array; and NO reads from the array, unless there's a failed drive and the drive is being emulated, in which case EVERY other drive is being read as well to simulate the failed drive. From a fault tolerance perspective, every drive is equally important -- if any one of them isn't working correctly when a drive fails, then the failed drive can't be correctly rebuilt or emulated. I'd hardly say that "... the parity drive is the least ..." important => but it may be true that it's the least used for arrays that are primarily write once; read many (as is true for a lot of media servers).
February 7, 201511 yr OK... few more bits and pieces. The 8TB Seagate seems to have capacity of: 8,001,563,222,016 bytes. My 4TB HGSTs, pair of which I plan to RAID-0 to obtain 8TB parity drive have 4,000,787,030,016 bytes each, two will be: 8,001,574,060,032 So, parity will be a tad bigger than 8TB data drive, so there is a hope that this might actually work. Now, I'll have to wait patiently until the drives arrive and pass preclear.... a week, not less. Oh my. SWEET!!! I can't wait to see the outcome of this. What controller are you using to do the RAID0? The plan is to use the other half of Addonics AD4SA6GPX2. It has 4 SATA connectors, two are doing RAID-0 cache drive, the other two are available.
February 8, 201511 yr ... The parity drive only has to work when another drive fails, and at that time EVERY other drive in the array has to be working perfectly also. Although all drives are important, I feel that the parity drive is the least. I agree that ALL drives are important, but it's certainly NOT true that "... The parity drive only has to work when another drive fails ...". The parity drive is involved in EVERY write to the array; and NO reads from the array, unless there's a failed drive and the drive is being emulated, in which case EVERY other drive is being read as well to simulate the failed drive. From a fault tolerance perspective, every drive is equally important -- if any one of them isn't working correctly when a drive fails, then the failed drive can't be correctly rebuilt or emulated. I'd hardly say that "... the parity drive is the least ..." important => but it may be true that it's the least used for arrays that are primarily write once; read many (as is true for a lot of media servers). I agree that all drives are important, but the integrity of the parity disk is only depended upon should a data disk fail. The integrity of data disks is constantly required. And if I had a double drive failure, I'd sure rather one was the parity disk!
February 8, 201511 yr ... The parity drive only has to work when another drive fails, and at that time EVERY other drive in the array has to be working perfectly also. Although all drives are important, I feel that the parity drive is the least. I agree that ALL drives are important, but it's certainly NOT true that "... The parity drive only has to work when another drive fails ...". The parity drive is involved in EVERY write to the array; and NO reads from the array, unless there's a failed drive and the drive is being emulated, in which case EVERY other drive is being read as well to simulate the failed drive. From a fault tolerance perspective, every drive is equally important -- if any one of them isn't working correctly when a drive fails, then the failed drive can't be correctly rebuilt or emulated. I'd hardly say that "... the parity drive is the least ..." important => but it may be true that it's the least used for arrays that are primarily write once; read many (as is true for a lot of media servers). I agree that all drives are important, but the integrity of the parity disk is only depended upon should a data disk fail. The integrity of data disks is constantly required. And if I had a double drive failure, I'd sure rather one was the parity disk! I agree if two drives are going to fail, it'd be nice if one was the parity disk (since you'd only then lose one disk's worth of data). But as for the integrity of the parity disk being less important than the integrity of a data disk -- I don't agree. If a data disks fails, you can still read it; reconstruct it; etc. because of the integrity of the parity disk. If the parity disk fails, you can reconstruct it because of the integrity of the data disks. The simple fact is the integrity of ALL disks is important, because if ANY disk fails (regardless of whether it's a data disk or the parity disk), ALL of the OTHER disks are necessary to reconstruct that failed disk.
February 8, 201511 yr The simple fact is the integrity of ALL disks is important, because if ANY disk fails (regardless of whether it's a data disk or the parity disk), ALL of the OTHER disks are necessary to reconstruct that failed disk. Yet if a RAID0 parity set has failed, no data has been lost. Only the ability to reconstruct another failed disk conveniently without resorting to backups. This makes it 'slightly' less important, at least in my personal sense of urgency. Frankly, I'm so much more concerned when a data drive gets a pending sector then my parity drive failing. A pending sector can cause a drive rebuild to fail no matter what drives are in play. After that, there is the issue of 'what file is suspect'.
February 8, 201511 yr If you are interested in creating HPAs look at THIS THREAD for a discussion of how to use the hdparm tool to create the HPA. Note the I'd recommend creating therm on a motherboard port. After they are created you can move the drives to a different controller. Thanks for an idea about HPA, will take a look how to do that and yes, the difference in size was really small in both tested variants.. i was wondering about lack of an option not to use a whole drive - we all know how drive performance is affected when we read/write to inner sectors.. Hi all, i will just confirm, HPA was setup on my test server and now test array is up and running with 2x 1TB Raid 0 as parity and one 2TB disk. in this config i lose about 1.1GB of space - that's acceptable and raid 0 will spin down too cos my Dell H710P supports it..
February 9, 201511 yr ... The parity drive only has to work when another drive fails, and at that time EVERY other drive in the array has to be working perfectly also. Although all drives are important, I feel that the parity drive is the least. I agree that ALL drives are important, but it's certainly NOT true that "... The parity drive only has to work when another drive fails ...". The parity drive is involved in EVERY write to the array; and NO reads from the array, unless there's a failed drive and the drive is being emulated, in which case EVERY other drive is being read as well to simulate the failed drive. From a fault tolerance perspective, every drive is equally important -- if any one of them isn't working correctly when a drive fails, then the failed drive can't be correctly rebuilt or emulated. I'd hardly say that "... the parity drive is the least ..." important => but it may be true that it's the least used for arrays that are primarily write once; read many (as is true for a lot of media servers). My real life experience is that there are tools running indexing those media folders, subtitles getting added, covers added.. My data drives are far from "read only"..
February 9, 201511 yr ... My real life experience is that there are tools running indexing those media folders, subtitles getting added, covers added.. My data drives are far from "read only".. Yet another reason why the parity drive isn't the "least important".
Archived
This topic is now archived and is closed to further replies.