Newsmeldung von slashCAM: Why 2009 RAID 5 are no longer secure enough of thomas - 22 oct 2008 14:07:00
This article presents a very simple but very interesting example of invoice to: 2009, 2 TB hard every day and be in RAID 5 arrays will be installed to ensure maximum storage capacity available. With an average default probability of a single hard drive of 3% (per year) is the failure of a plate from a larger array is likely - especially the longer the holding time and the more disks in the array are used: in 7 hard disks, the failure probability of 20% already per year. Actually not a problem for a RAID 5 system: from the data of the remaining panels, the data of the failed disk are reconstructed. But there is a second relevant defect size: the "unrecoverable read error rate", ie the probability of not re-producible reading errors, which SATA disks at 10 ^ 14 is located, so in such an error every 12 terabytes read. Falls in a 7x2 TB RAID 5, a plate and needs to be restored, it is very likely that while reading the 12 TB of data of the other plates such an error occurs. The result: the reconstruction of the disk fails and the data of all partitions are lost! Of course, this risk already exists today, but the larger the plates are, the more likely it is total failure. Sure, data is only through a total backup of the data - RAID 5 alone in any case is no longer safe enough for such data sets and RAID 6 (tolerance for 2 errors), only an interim solution.
This is an auto-generated entry
Antwort von camworks:
raid6 why should only be an interim solution? I find this solution quite well, have even a hardware raid6 (Areca 1220) and it gave me 2 times already rescued the rear.
still have backups, because if the computer power supply and a tension abraucht spitze sends into the plates, is synonymous with raid6 from the oven. I ask myself, what with those of multiple terabyte sizes still comfortably backuppen wants. adequate solutions cost a fortune. is time that the industry as what specific offers.
Antwort von r.p.television:
As long as one due to high data rates such as recording of umkomprimierten HD-SDI input is not dependent on a RAID, I would anyway to steer clear of it. Had it been significant data loss. I have now individual plates and once a week doing a backup.
The performance advantage of a RAID system for my data rates irrelevant.
Antwort von camworks:
and if you make your backup on Friday and thursday s.nächsten hops the plate goes, you have just the information a week disposed of. very carefully thought out approach! ;-)
Antwort von wontuwontu:
Basically, the author has quite sure, data can be with a RAID system can not be reached. However, investing the majority of Raid so users probably only be in a Raid (5) system, because there is simply no other similarly effective solution. I myself use a 6TB-Raid5, which we mainly used for the recording of uncompressed 10bit material need, since the data is primarily behind the speed. If, however, times 2TB Footage thing, it is synonymous throughout the not so easy "short time" to secure. A solution to the times I wanted to test was two Raid5 in a RAID1 collaborative activities. That should be the safety yes again increase slightly. Otherwise, hold a good UPS ballast and the most important data on an external disk secure.
Antwort von r.p.television:
and if you make your backup on Friday and thursday s.nächsten hops the plate goes, you have just the information a week disposed of. very carefully thought out approach! ;-) Labor-intensive files such as project data, copy etc. Of course I always on a second folder on an external disk. If my data is lost, then only gebatchte video files that are always produced. And of course I backup once a week.
Antwort von camworks:
So you helping hand after RAID5. much work, everything by hand copying. but where should it be beneficial? your backup disk can just break down as RAID5.
Antwort von r.p.television:
So you helping hand after RAID5. much work, everything by hand copying. but where should it be beneficial? your backup disk can just break down as RAID5. At the risk of it is the impression I rhymes like my answers for a Contra together:
I have course two backup disks, ie the same content on three discs, but not in a raid linked to but once a week and on completion of projects secured. Project files are secured permanent background.
After all, I drive with this method is safer than with the stupid Raid 5th Two years ago, I still synonymous to the alleged benefits of the Raid-grouping and have fooled twice thanks to significant data loss suffer. Once through spikes and two hard disks abgeraucht once the Raid controller, which then unfortunately was out of stock. The data on the 6 plates were then garbage.
And another advantage: The two back-plates are only for the data set. That means less power consumption, less wear, less noise, etc.
So I would never be a RAID-Verbund not entrust recoverable data. I would consider it only for recording a use uncompressed video.
Antwort von camworks:
After all, I drive with this method is safer than with the stupid Raid 5th Two years ago, I still synonymous to the alleged benefits of the Raid-grouping and have fooled twice thanks to significant data loss suffer. Once through spikes and two hard disks abgeraucht once the Raid controller, which then unfortunately was out of stock. The data on the 6 plates were then garbage.
can happen, so I've written synonymous that when tension is lost everything sharp.
but if "only" 2 plates to abandon the spirit, my tuts raid6 yet.
and I need have no fear that my controller "out of stock", because I have a brand controller. tute since even the successor, since the same procedure is applied. I promise and so on with Areca experienced, even more than 2 controllerless generations can raid the old associations still be read.
ultimately, the safety raid6 with quite high, but I will soon have a very big "us" as a backup growth.
Antwort von PowerMac:
So you helping hand after RAID5. much work, everything by hand copying. but where should it be beneficial? your backup disk can just break down as RAID5. Manual backups are anyway a relic of the 90s. Today the intelligent Programs, such as Time Machine.
Antwort von r.p.television:
and I need have no fear that my controller "out of stock", because I have a brand controller. tute since even the successor, since the same procedure is applied. I promise and so on with Areca experienced, even more than 2 controllerless generations can raid the old associations still be read.
That I thought was synonymous. I had a brand synonymous controller. Unfortunately, his successor was not as tolerant as you. I think there is no guarantee and is subject to more happiness.
Manual backups are anyway a relic of the 90s. Today the intelligent Programs, such as Time Machine. I've already considered synonymous. However trau automatisms I do not really understand. I'd somehow afraid that this kind of way, and something is exchanged, etc. It may be hintergrundlos. On the other are the steps already so familiar and not a big expense. The backup is done overnight or in breaks from work and not hinder my workflow.
Antwort von mmohl:
So you helping hand after RAID5. much work, everything by hand copying. but where should it be beneficial? your backup disk can just break down as RAID5. It is not a RAID5 but RAID1 after. The news in the problem occurs because not exactly. When you lose a disk needs to reconstruct only one other error-free plate can be read and not all the other plates of the x raids.
Antwort von LarsProgressiv:
Hi Everybody,
Why is the RAID 5 only in 2009 no longer safe enough?
The article described the danger is surely only in the number of composites used for RAID hard disks. I can just as well with 20 plates á 10GB RAID5 / 6 and thus build the risk into the Height drive.
That with the Hardwarekontrollern and whose default risk is very well with software RAID bypass. The controller runs s._jedem_, s.dem run the plates.
In the current c't 22/2008 are very good article.
Regards
Lars
Antwort von camworks:
So you helping hand after RAID5. much work, everything by hand copying. but where should it be beneficial? your backup disk can just break down as RAID5.
It is not a RAID5 but RAID1 after. The news in the problem occurs because not exactly. When you lose a disk needs to reconstruct only one other error-free plate can be read and not all the other plates of the x raids. true mistake of me.
Antwort von tom:
Hi Everybody,
Why is the RAID 5 only in 2009 no longer safe enough?
The article described the danger is surely only in the number of composites used for RAID hard disks. I can just as well with 20 plates á 10GB RAID5 / 6 and thus build the risk into the Height drive.
Yes, of course. The article says only the 2009 risk due to the larger disks will be even bigger than it is today - a RAID 5 with too many plates (20) is always a bad idea, because only the failure of a disk can be tolerated, the likelihood of multiple failure on many boards but rising - especially as the plates the same brand purchased together were sometimes show a tendency synonymous to similar auszufallen time and larger amounts of data at the stress of the RAID rebuild such a failure may provoke even synonymous.
For RAID 6: because the article says that the still relatively safe, but with even greater amounts of data / records in the future in these default probabilities are synonymous is too uncertain.
Antwort von Meggs:
The article is very pessimistic, because:
1. He assumes that the plates are full, ie 14 terrabyte described and must be read.
2. 1 byte if it is unreadable, how does the Raid?
If indeed the entire disk and thus the entire composite discarded? Or does this one byte ultimately to one or a few files that can not be reconstructed?
As far as I know, announces the failure of a plate, and the raid logs and requires a new disk before the old is completely out. In the case could not yet completely destroyed old record synonymous still be used for reconstruction.
That being said: Raids never replace a backup, just because of technical failure of a plate is secured, not accidentally overwritten by the user or by an error in software. Just as mechanical destruction of the compound by fire, lightning, short circuit, etc..
Antwort von PowerMac:
I maintain quite saucy, a non-byte read is irrelevant. The plate itself, the means of transmission, the operating system files - anywhere but redundancies are built.
Antwort von Meggs:
In any case, would have described the horror of this scenario is not readable byte data, which for the manufacture of the failed disk is important. That is only a fraction of the available data of the case. Otherwise, this byte is synonymous a case for the recovery mechanisms of the raids.
