Talk:Nested RAID levels/Archive 1

Edited the section on RAID 30
I removed the comment from the RAID 30 section which claimed the highest level of redundancy and performance for RAID 30, since it directly contradicts the statement in the preceding paragraph that, after one drive failure, all the other drives in the failed drive's cluster become a single point of failure. RAID 60, on the other hand, can afford to have half the drives in the entire array fail before data loss occurs. Gepard 12:30, 1 February 2007 (UTC)

RAID 51
There is a typo (or copy paste error) in RAID 10 and RAID 0+1 sections: "hybrid approaches such as RAID 0+1+5 (mirroring above single parity) or RAID 0+1+6 (mirroring above dual parity)." Mirroring above single parity is logically RAID 51 (this is consistence with main RAID article). I will correct this and add an external link to http://www.linux-tutorial.info about RAID. Kopovoi 15:31, 20 February 2007 (UTC)

Inaccuracy in RAID 60 section
The section contains the following sentence:

more than half of the disks must fail in the above example in order to lose data

If three disks from the same RAID 6 set fail, then all data is lost. Three is less than half of the disks. —Preceding unsigned comment added by 194.151.55.34 (talk) 15:45, 25 January 2008 (UTC)

RAID 00
Is RAID 0+0 possible? what would be the drawbacks and advantages? for example:

Hard1    Hard2     Hard3     Hard4 A1       A2        A3        A4 B1        B2        B3        B4 C1        C2        C3        C4 D1        D3        D4        D4

--83.183.127.178 (talk) 14:40, 14 June 2008 (UTC)


 * It's the same as plain RAID 0 across more disks, though working out what the strip size will be can be 'interesting' 86.13.76.46 (talk) 09:38, 24 June 2008 (UTC)

A note regarding edit by 87.194.157.28
For reference, that IP is mine (i.e. static), and I made that edit before I had an account. Feel free to delete this note.

--Elrebrin (talk) 14:26, 31 December 2008 (UTC)

Restrictive definition of RAID 0+1 and 1+0
At the lowest nested level, the sub-arrays are not limited to pairs of disks. It is perfectly valid to have for example, a RAID 1+0 array using mirrored sets of 3 or more disks. Can this be somehow incorporated into the article? --Spook (my talk 06:20, 5 February 2009 (UTC)

Faulty example
The article says: "Below is an example where three collections of 240 GB RAID 5s are striped together to make 720 GB of total storage space" with 9 disks of 240 GB each you will get 2160 GB. This needs to be addressed. —Preceding unsigned comment added by 87.79.87.190 (talk) 21:51, 3 March 2009 (UTC)

Citations / References
I removed one external link from References that was obviously for commercial purposes, but then I moved another link that was in References to the External Links section. I did this because it had no corresponding citations. Since this left the References section empty, I removed the References section, period. In reality, someone needs to go through and add the citations AND the corresponding references (as close to simultaneously as possible). —Preceding unsigned comment added by 66.195.226.134 (talk) 04:26, 19 March 2009 (UTC)

Images and Diagrams
Please do not remove and images or diagrams. They're all there for a reason. The diagrams were added by various authors to prove a point. Eventually all the diagrams should be converted to images. Diagrams are easier to edit by most so that's what they start off as until someone with artistic talent updates them to an image.Avernar 15:53, 17 February 2007 (UTC)
 * Not a problem. Sorry about that. Koweja 17:47, 17 February 2007 (UTC)
 * How about this? I can make changes, but do you think it is a good start?
 * [[Image:Raid_0+1.svg|thumb|none]]
 * Paulish (talk) 22:19, 18 April 2009 (UTC)

Explanation for Added WikiProject Computing Template Parameter Values
I added the WikiProject Computing Template since this article is listed under the "Open Requests" section in this WikiProject.

Class: I rated this article as a stub-class article because it has absolutely no sources. According to the WikiProject Computing/Assessment page, if an article has no listed sources, it is a stub, regardless how informative, well-written, or long it is. This is simply because such articles are not verifiable, which Wikipedia holds in very high regard.

Importance: In the WikiProject Computing Assessment page (http://en.wikipedia.org/wiki/Wikipedia:WikiProject_Computing/Assessment), an article is of low importance if the article elaborates on "Optional add-ons that are not fairly important". Nested RAID levels are not, in any way, integral to computers; RAID levels are optional additions to computing systems, making Nested RAID levels of even lower importance. Hence, I gave this article a low importance rating.

Portal: I added this article to the computing portal because it is fitting.

Needs Infobox: I noted that this article lacks an infobox simply because it technically lacks an infobox.

Hardware and Hardware Importance: This article was added to the Computer Hardware task force and was marked as Unknown-importance in this area before I added the WikiProject Computing template. Nested RAID levels are computer hardware, but since they are not integral pieces of hardware (computer will run without them), I believe they are of low importance in this category as well.

WikiProject Computer Science and WikiProject Computer Science Importance: I added this article to WikiProject Computer Science because I thought it was obviously befitting. I marked this article of low importance in this category as well becaused Nested RAID levels are not integral to the study of computer functioning. Some Old Man (talk) 21:49, 8 May 2009 (UTC)

Deletion of
Since Wikipedia users have done an admirable job in adding diagrams to this article, I deleted the request for diagrams.

--Some Old Man (talk) 04:20, 8 June 2009 (UTC)

Description of this Article's WikiProject Computing Template Update
Since Wikipedia users have done an admirable job improving the quality of this article (especially in raising the number of sources from 0 to 2), I have updated the article's quality rating from stub-class to start-class. The article's quality still seems to be limited by its number of references, since two references are probably not enough references for an article of this size and technical nature to be completely verifiable (in accordance with http://en.wikipedia.org/wiki/Wikipedia:WikiProject_Computing/Assessment#Quality_Statistics). However, the article is undergoing obvious improvement.

--Some Old Man (talk) 12:27, 30 June 2009 (UTC)

RAID 60 (RAID 6+0) - bad example
I would like to point out that the visual example given for this section is not able to demonstrate the value of RAID 60 over RAID 10. The reason for this is that the demonstrated RAID 60 has 50% redundancy (2x 4-drive RAID 6 striped together, with 2 disks worth of parity per stripe). The same 8 disks could be used in a RAID 10 with the same level of redundancy, and possibly identical fault-tolerance (complex, and arguable. Faster rebuilds probably give 10 the edge), but with the definite downside of the RAID6 of slower rebuilds, more processing necessary, more expensive controllers needed, and some other issues that I don't have the time to write at the moment, but that should be enough.

I suggest that an example be used with at least 5 drives per stripe in order to give it the edge in storage efficiency over RAID 10, otherwise the example shows no clear advantage of RAID 60 over RAID 10.

203.45.41.216 (talk) 01:37, 2 September 2009 (UTC)

Graphics Standard
I was going to do the graphics that was undone (Raid 30) and i find the graphics had no standard, so i want to know what standard use to do it, and i may redo other graphics as well for the sake of standardization. (Ishnigarrab (talk) 02:39, 30 November 2009 (UTC))

RAID 10 definition
It appears that someone has got it wrong. This article, http://msdn2.microsoft.com/en-us/library/ms190764.aspx and others have RAID 10 the other way around. Which is right? 212.212.161.101 10:38, 30 April 2007 (UTC)mct


 * I would not put the question so sharp.

It is indeed the question of definition. As it is stated in []: "The difference between 0+1 and 1+0 might seem subtle, and sometimes companies may use the terms interchangeably." Or "Microsoft SQL Server 2000 High Availability": "Striped mirrors and mirrored stripes often confuse people, and manufacturers might call them RAID 10, RAID 0+1, RAID 1+1, RAID 01, and other terms. Work with your manufacturer to understand which one they actually implement for their disk subsystems." We can mention the situation in the article.

One can define mirror made of stripes to be RAID 0+1, meaning that first you apply stripe (raid 0) and then you apply mirror (raid 1). One can also call it RAID 10, translating directly "mirror of stripes" = "1 of 0" = "10".

Still, I would choose the definition from the site, where both RAID 10 and RAID 01 are mentioned, to be sure that the author is aware of the distinction. I would also prefer the link from hardware manufacturer whose product really support the both RAIDs. Kopovoi 12:55, 30 April 2007 (UTC)


 * There's further confusion in that it should be very simple to switch an array between 1+0 and 0+1 as no data needs to be moved to do the switch (the ondisk layout is the same). This generally should mean that only poor hardware supports both raid combinations but good hardware will be able to do anything the poor hardware can with their stripe and mirror layout. 86.13.76.46 (talk) 09:43, 24 June 2008 (UTC)


 * True. There is absolutely no reason nor excuse to mirror striped two disks in any way without ensuring that every stripe exists on both disks.  While one might stripe two disks to double the speed of access, the risk is actually more than times as great as any failure of either disk causes the entire array to be useless.  Poor implementations that don't make good use of predictability of offsets (unless they are totally formulaic, it's admittedly hard to maintain across many writes, may require occasional rewriting of the whole disk) might sacrifice a few rarely-used stripes and put more copies of commonly-used ones at "near" access points following some kind of Markov chain algorithm, but if they don't make up for that somehow with additional parity or more frequent backup of the rarely-used stripes, it's unforgiveable.  RAID by definition assumes that the risk to each bit of data on the same disk is the same and this is a key design criteria for serious integrators.


 * The real problem in this area is a lack of mapping and formatting standards so that RAIDs that fail can be recovered on entirely different RAIDs.  The use of NAS is going to come to a crashing halt very soon when people realize that even a RAID1 mirrored NAS with no striping leaves them "spare" disks that they cannot simply put into another enclosure and access another way.  This is not what they think they are buying.  The article on NAS should really make this clearer.


 * These technical articles are really not meeting the needs of 99.99% of users who want to understand what RAID and nested RAID levels are for purposes of selecting a piece of hardware, and understanding what risks with their data are involved. —Preceding unsigned comment added by 142.177.6.204 (talk) 16:16, 21 September 2010 (UTC)

Irrelevant Marketing
The page says: "In high end configurations, enterprise storage experts expected PCIe and SAS storage to dominate and eventually replace interfaces designed for spinning metal"

The only source for this statement is a published interview with a marketing person. Of course the marketing person wants people to use their PCIe-based products. But one vendor's marketing person has become "enterprise storage experts" (more than one!).

I'd suggest removing it. —Preceding unsigned comment added by 82.68.80.137 (talk) 10:44, 19 January 2011 (UTC)

Confusing array size notations
RAID 0+1 and RAID 10 use the same capacity calculation, but the article uses two completely different notations for each. Could someone make the article more consistent? Also adding notes to the effect that the usable capacity is the same as in other RAID level might help make it clear so that people don't have to compare math formulas to figure it out. Both formulas can be written as: $$Size = \dfrac{(n \cdot S_{\mathrm{min}})}{2}$$

From the article:

RAID 0+1

 * The size of a RAID 0+1 array can be calculated as follows where n is the number of drives (must be even) and c is the capacity of the smallest drive in the array: $$\begin{align}Size & = \left(n \times c\right) \div 2\end{align}$$

RAID 10

 * The usable capacity of a RAID 10 array is $$(N/2) \cdot S_{\mathrm{min}}$$, where $$N$$ is the total number of drives in the array and $$S_{\mathrm{min}}$$ is the capacity of the smallest drive in the array.


 * I would argue that the available capacity is just: $$(N/2)$$ because 0 is just striping not striping with parity. So, 6 disks 10GB each. 3 mirrored pairs, 10GB available each = 30 GB total.  Striped = 30 GB total.


 * $$(N/2) \cdot S_{\mathrm{min}}$$ would be for RAID 30 or RAID 50

Neither $$(N/2) \cdot S_{\mathrm{min}}$$ nor $$(N/2)$$ is correct; the capacity is Σ $$V_{\mathrm{i,min}}$$ where $$V_{\mathrm{i,min}}$$ is the capacity of the smallest disc in the ith mirror set. Michealt (talk) 10:16, 7 March 2011 (UTC)

Linux RAID 10 driver
I've updated the Linux RAID 10 driver paragraph after reading the driver author's web page and examining the actual driver code. I've removed the line about RAID 10 on two disks and the block diagram as they are not standard RAID 10. Avernar 08:07, 2 January 2007 (UTC)

I've expanded the Linux RAID 10 section in the Proprietary RAID levels article to include the far layout so the block diagram is now there (but with three drives). Avernar 09:48, 2 January 2007 (UTC)

It looks as if someone has put the 2-disc stuff back. I've added a comment that it is non-standard. In fact the Linus thing is so far from the standard definition of RAID 10 that it probably should be taken out of the RAID 10 section and have a separate section for itself.

I believe the whole thing about the FAR configuration doubling serial read performance is just nonsense, based on a seven year old guess by Neil Brown and never as far as I know backed up by actual measurement (although the statement has certainly spread around a lot). Can anyone cite a trustable benchmark on this? Michealt (talk) 11:15, 7 March 2011 (UTC)

Ludicrous RAID 10 FUD/original research
Regarding the following sentence:

Given these increasing risks with RAID 10, many business and mission critical enterprise environments are beginning to evaluate more fault tolerant RAID setups that add underlying disk parity.[citation needed] Among the most promising are hybrid approaches such as RAID 50 (stripe above single parity) or RAID 60 (stripe above dual parity).

This is utter nonsense. For example, in a six disk RAID 10 if one drive fails, it continues to operate with a single point of failure, but that single point is only the other drive in the same mirror. This means if another were to fail, the odds of that crucial drive failing are only 1/5, or 20%.

Conversely, six drives in a 50 setup (2x 3 disk raid 5 striped together) If one drive fails it continues to operate, still good, but there are TWO points of failure in this setup. If EITHER of the remaining two drives in the same RAID 5 fails, then the RAID is dead. This is 2x1/5, or 40% chance that the next drive that fails will be in the degraded RAID.

Clearly then the likelihood of a RAID 50 losing all its data after a second disk failure is double that of a RAID 10 in a six disk setup. I understand that the odds favoring a RAID 10 increases as number of disks increase. However, RAID 50 looks worse as number of disks go up.

The likelihood of a second drive failure killing the whole RAID tends toward 50% in a 2x striped RAID 5 setup, and 33% in a 3x striped RAID 5 setup (as in the examples shown). However the odds of losing the single crucial mirrored disk in a RAID 10 begin at 33% in a 4 disk setup, and tend toward 0% as the number of disks in the RAID go up.

RAID 60 on the other hand will always operate after any two disk failures - however you would never have a RAID 60 with only 6 drives, since you would be using 66% of the capacity of the RAID as parity! In that case, with only 6 drives (or 4) you would likely run a raid 10, particularly since it also reduces processor and rebuild overhead (raid 10 only needing resilvering)

RAID 10 on the other hand can sustain a failure of 50% of the drives in the RAID, so long as one of each of the mirrors survive.

I understand that the examples in the main document use more striping (3 stripes opposed to only 2), but that clearly does not apply to setups with fewer than 9 disks for RAID 50, or 12 disks in RAID 60. It is not uncommon to see RAID setups for enterprise with fewer than 9 disks. The comment quoted above may be for the most part true true, but only in cases where there are more than the numbers of disks cited above.

It also ignores the emerging issue where drives of 1T or over in large distributed parity arrays are increasingly likely to see subsequent disk failures DURING a RAID rebuild, which completely negates the advantage of distributed parity. - http://blogs.zdnet.com/storage/?p=162

I would like to highlight that the addition of a hot spare to a RAID 5 does not in any way negate the emerging problem discussed in the link above. One way to temporarily negate that issue, however, would be to greatly increase the MTBF of the disks used in a RAID 5. It should be clear that greater quality control needed for RAID 5 disks will also lead to high cost - further negating any advantage RAID 5 may have over RAID 10. RAID 5 does however maintain a lead over RAID 10 in physical storage space needed to house the drives, however this is moot if using drives of a high capacity that lead to guaranteed RAID failure in a rebuild.

I would also like to point out that I can see that the issue from the link above does also apply to RAID 10, but due to the speed of resilvering, and the fact that all RAID 10 configurations can continue with up to 50% disk failures, disk storage would still need to increase by orders of magnitude before it becomes a concern of the same degree that it is for RAID 5 with 1T+ disks.


 * Agreed. The fact that "all RAID 10 configurations can continue with up to 50% disk failures" means that a two-disk configuration can continue with one disk for as long as it takes to replace the other disk.  It's the best of both worlds, at least as far as bootable RAID is concerned.  The potential drawbacks of this (due to the need to write multiple blocks before doing another transaction that may read and act on that data) are outlined now in the RAID10 section.  Along with the extremely strategic applications in which RAID10 lets you protect your investment in slower (SATA-II) disks or employ multiple smaller disks as a bootable RAID.  Given these advantages and Linux support for them, and increasingly Windows or Mac donfigurations using them, expect RAID10 to be 99% of all RAID configurations.

My final word on this is that although a RAID 10 only uses 50% of the available space, storage is very cheap these days, and the added cost in buying another $500 or so of storage to set up a RAID 10 is easily rationalised by the added data security over using RAID 5 or 6 with high capacity disks. Octothorn (talk) 01:55, 28 August 2009 (UTC)


 * Exactly. With 1TB of storage selling for about $50, buying another $50 to get a double-speed 1TB bootable RAID which is reliable enough to hold some of your data until network access allows it to be backed up, is an investment that any sane IT manager should make.  Already, in 2010, allowing users to write data to a non-mirrored directly attached data drive that is not instantly backed up to network storage, is an extremely questionable practice and probably unprofessional.


 * Agreed. For that extra $50 you'd get double speed read access without necessarily taking much hit on write access, and data mirroring. If either drive fails, you still have all the stripes on the other and can immediately back them up to the network and write a second drive.  The first the end user has to know about the problem is when someone from IT appears with an already-striped second drive to replace the one that failed, which they can do over a coffee break.  Compare that to data recovery, lost work time, interruption of workflow, or putting up with slow (in the minutes) boot times and it's pretty obvious that a responsible IT manager should already be doing this and that you get more from that than from $50 more of RAM or monitor or video card in an already well-designed workstation that has good interfaces (teamed gigabit ethernet, any modern GPU, 2GB RAM or more).


 * Agreed, the paragraph in the article is nothing but FUD. If one wants increased reliability, one can use RAID 10 with tree-way mirroring. .  This paragraph should be deleted.  Michealt (talk) 10:34, 7 March 2011 (UTC)

I've been so bold as to delete that passage.Wbloos (talk) 19:32, 4 April 2011 (UTC)

raid 0+1 incorrect statement
It says in the section raid 0+1: "It is not as robust as RAID 10 and cannot tolerate two simultaneous disk failures, unless the second failed disk is from the same stripe as the first. " The 'unless' part cannot be true. The raid0 array in which 1 disk fails is out of order. There is no such thing as another disk failing in an array that is not working at all. How about: "It is not as robust as RAID 10 and cannot tolerate two simultaneous disk failures. When 1 disk fails, the raid0 array that it is in will fail also. The raid 10 array will continue to work on the remaining raid0 array. If a disk from that array fails before the first failing disk has been replaced, the data will be lost." — Preceding unsigned comment added by Wbloos (talk • contribs) 18:44, 4 April 2011 (UTC) I've changed it on the page. I hope it's alright. Wbloos (talk) 19:34, 4 April 2011 (UTC)

Review of article classifications
Is it time to revisit some of the classifications for this article? For example, it seems to be substantially more complete than a start-class article. It also seems to merit a higher importance-- I think cases can be made for mid or high, but low certainly seems... well, low. --Spotstubes (talk) 22:27, 12 October 2011 (UTC)

Recommendation regarding capacity of example disks
I don't know about you, but in practice 120gb drives are rare these days. I think, in the interest of keeping this article relevant (and easier to understand the math) the metric be changed to 500gb —Preceding unsigned comment added by 67.84.194.22 (talk) 15:33, 11 April 2009 (UTC) - Garrett W. (Talk / Contribs) 13:08, 1 December 2009 (UTC)
 * Rare? I don't agree. Dlabtot (talk) 15:43, 11 April 2009 (UTC)
 * I agree with this for the sake of simplicity. Thinking about 500gb disks is easier than with 120gb, and using 1Tb is even better. Ishnigarrab (talk) 07:25, 29 November 2009 (UTC))
 * I vote to use 250GB instead of 120 or 500. It's easy to think about, and it also reduces the likelihood that the reader will need to comprehend a jump from GB to TB.
 * Using something similar to Moore's law as a guide ... assuming that capacity doubles every 18 months, the equivalent of the 250 above would now be 1,000. I would strongly urge any person or team of people making changes to the arithmetic here to change the disk sizes to 1 TB. That would be simpler, more realistic and easier to look up and cost actual disks. Beyond that, I suspect that most writers here tend to be 'abstract', while most readers are likely to be 'concrete' thinkers. The simplification would likely not interfere with writing, but would surely make the subject matter accessible to concrete thinkers who must of necessity visualize an actual example of the calculation in order to 'grok' it.  DeepNorth (talk) 00:15, 2 January 2013 (UTC)

would someone address a point of confusion: is stripe size absolute or aggregate? (total, or per disk?) When the RAID BIOS asks what stripe size to use (often gives a list of choices (16k, 64k, 256k), is the stripe size based per disk or for the total array?  example: If a 64kb stripe is chosen with a raid0 of four drives, does this imply A. that there will be one 16kb segment on each of the four drive that totals 64k, meaning the smallest data item available to the OS is 64kb? or B. does it mean that there is a 64k segment on each of the four drives, so that the smallest data item available to the OS is (4 x 64k) 256kb? Accordingly, when formating a file system, should one match the RAID stripe size to the format sector size?    example:  with a raid stripe of 256k, #mkfs.ext4 -s 256k 173.20.39.44 (talk) 03:16, 5 August 2010 (UTC)

Reason for removal?
In a previous edit, I added this to the article:

While different naming conventions may be used (such as 1+0 vs. 10), in both, the names follow a "bottom-to-top" convention, "bottom" and "top" as portrayed in the array diagrams shown herein. The "bottom-most" array type is listed first, while the RAID type which joins the bottom-level arrays, and typically provides redundancy, is listed second (or even third, in the case of RAID 100). For example, an array of type 50 consists of at least 2 RAID 5 arrays joined by RAID 0. Similarly, a RAID 0+1 array consists of RAID 0 bottom-level arrays joined by RAID 1.

Why was this removed? It vanished, but I don't see any explanation in the talk page. It was removed in revision as of 21:48, 31 January 2013. If some modification is needed I'd be happy to do it, but I think a lack of naming convention explanation is a notable deficiency in the article. Spotstubes (talk) 22:38, 10 April 2013 (UTC)


 * Your addition was removed because you did not cite your sources. Also, please provide an edit summary for each of your edits. You can read the edit summaries of other editors (when they provide one, which they really should) on the History tab to determine why something was changed or removed. Lastly, please add new threads to the bottom of talk pages—not the top. Thanks and welcome to Wikipedia! – voidxor (talk &#124; contrib) 07:13, 12 April 2013 (UTC)


 * In that case, I think I'm out of luck, as it is a fact that follows from observation of the published standards of many technical bodies and industry members. I can either cite all of them in a massive list or exclude it, and given the overactive editing of this page, I think any attempt to add so many citations would be immediately whacked. Spotstubes (talk) 20:58, 18 April 2013 (UTC)


 * If you can find several sources to support a fact, that's great--but not required. You need at least one. I recommend two just to support the credibility of the first reference, but you definitely don't need an exhaustive list of references. As far as editing goes, nobody should object to the addition of citations, assuming they're credible and relevant. – voidxor (talk &#124; contrib) 04:41, 19 April 2013 (UTC)


 * It's not so much finding a couple of sources that say, "names follow a 'bottom-to-top' convention...". It is, rather, pointing to sources that have RAID type descriptions and saying to the reader, "Note for yourself that all of these sources follow the 'bottom to top' convention".  That will not survive.  Under the rules, maybe it shouldn't, but I don't care quite enough try and have my effort wasted.  Maybe someone with the authority to slap down overzealous editors can do it, but I'm not holding my breath. Spotstubes (talk) 21:21, 23 April 2013 (UTC)

I'm trying to figure out what are the differences and cannot
Section: RAID 100 (RAID 1+0+0)

What is the difference between, for example the RAID array in the image (RAID 100.png), i.e. a "RAID"-0 of a size of 2 over a "RAID"-0 of a size of 2 over [something], to a RAID array that is composed of a nested RAID of "RAID"-0 of a size of 4 over [the same something] ?

Or more generally, what are the advantages and disadvantages of multilevel striping in contrast to striping in a single level over a large number of physical drives/other blocks? 77.126.46.227 (talk) 16:38, 8 November 2013 (UTC)

"Near versus far, advantages for bootable RAID" doesn't discuss Near, Far, or bootability
As stated. There is a header which references "Near versus far", but neither of those words appears anywhere in the article except in that header and the TOC. Nor does that section appear to discuss "bootable RAID". 108.3.140.128 (talk) 13:12, 21 January 2014 (UTC)


 * That heading seemed pointless so I just removed it. Thanks for reporting it! – voidxor (talk &#124; contrib) 06:37, 22 January 2014 (UTC)

Superscript notation for number of disks
In many groups, eg, a 12 disk RAID5+0, it's necessary to know how many disks in each RAID5 block. Some people use the superscript or caret notation. Eg, 3 groups of (4 disks in a RAID5) arranged in RAID0 is RAID54+0 RAID5^4+0. See http://www.smbitjournal.com/2012/12/standard-network-raid-notation-standard-sam-raid-notation/. Is this technique widely used? is it worth including in this article? Andy Henson 109.231.215.34 (talk) 17:56, 13 May 2014 (UTC)


 * Hm, I haven't seen that notation, neither used in practice nor in written materials. &mdash; Dsimic (talk | contribs) 21:40, 15 May 2014 (UTC)


 * I haven't seen that notation either. – voidxor (talk &#124; contrib) 06:19, 17 May 2014 (UTC)

RAID 10 and spans & pictures
The term "span" is only used in a few places and not properly defined anywhere. From what I understood a "span" in RAID 10 is one of the underlying RAID 1 arrays. I believe the write performance for RAID 10 in the table is wrong, it should be simply the number of spans. — Preceding unsigned comment added by 142.104.191.242 (talk) 00:23, 18 March 2017 (UTC)


 * raid pictures are wrong 03 shows 30. It is first the raid 3 then raid0; raid 10 shows 01, these are stupid mistakes that can be fixed when someone has time to do it

External links modified (February 2018)
Hello fellow Wikipedians,

I have just modified one external link on Nested RAID levels. Please take a moment to review my edit. If you have any questions, or need the bot to ignore the links, or the page altogether, please visit this simple FaQ for additional information. I made the following changes:
 * Added archive https://web.archive.org/web/20110605162202/http://www-03.ibm.com/systems/resources/Demartek_IBM_LSI_RAID_Controller_Performance_Evaluation_2009-10.pdf to http://www-03.ibm.com/systems/resources/Demartek_IBM_LSI_RAID_Controller_Performance_Evaluation_2009-10.pdf

When you have finished reviewing my changes, you may follow the instructions on the template below to fix any issues with the URLs.

Cheers.— InternetArchiveBot  (Report bug) 07:20, 16 February 2018 (UTC)

RAID 01 Space Efficiency
In the section on RAID 0+1 the discussion explicitly says, correctly, that space availability for RAID 0+1 is (N/2), which is 50%, but the comparison table below says that RAID 0+1 space efficiency is 1. It should be 1/2, right? Thanks... 173.79.165.244 (talk) 15:34, 16 February 2018 (UTC)mjd