Talk:Non-standard RAID levels/Archive 1

Z level

 * RAID 0: Striped Set (2 disk minimum), without parity. Provides improved performance and additional storage but no fault tolerance from disk errors or disk failure. Any disk failure destroys the array, and is twice as likely to fail.
 * RAID 1: Mirrored Set (2 disks minimum), without parity. Provides fault tolerance from disk errors and single disk failure. Increased read performance occurs when using a multi-threaded operating system that supports split seeks, very small performance reduction when writing. Array continues to operate with one failed drive.
 * RAID-ZZZ: Sleeping Set (1 disk minimum), without parity. Provides improved power savings by powering down disks, but no read performance increase or fault tolerance from disk errors or failure.

SCNR. By me, j.engelh 02:57, 26 March 2007 (UTC)

Paltry Diagrams
The diagrams have redundant (useless) text on them. The diagram "Diagram of a Matrix RAID setup", for example,: 1.) "Diagram of a" is redundant and useless text; the entire label under the diagram is useless text because the diagram already has the words "Matrix-RAID" (which should not be hyphenated) on it. 2.) The word "partition" on the diagram should be "volume" (a RAID volume can be partitioned just like a single drive). —The preceding unsigned comment was added by TonyWest (talk • contribs) 20:21, 7 May 2007 (UTC).

RAID-X
What about XIV storage RAID-X? —Preceding unsigned comment added by 216.17.194.146 (talk) 19:41, 4 January 2008 (UTC)

Linux md section vs IBM ServeRaid
It seems the Linux md section contains all the IBM ServeRaid info. —Preceding unsigned comment added by 89.132.218.46 (talk) 20:58, 22 July 2008 (UTC)

DROBO
The Drobo's storage functions similarly to WHS... not sure where there's a good technical description...
 * If by WHS you mean Windows Home Server then no, not really. They both pool randomly sized disks, but that's about it. WHS can replicate data to create redundancy, but only by creating copies (RAID 1-ish, needs twice the storage). Drobo mixes RAID levels as good as it can and only needs ~1/4 of the available storage for backup. WHS can use any disk attached to the system, even via USB or Firewire, Drobo is limited to 4 internal drives. Sneeka2 (talk) 14:13, 21 March 2009 (UTC)

X-raid
What about Infrnat's X-raid? --81.129.155.240 16:55, 15 June 2007 (UTC)

http://www.readynas.com/forum/faq.php#What_is_the_advantage_of_using_X-RAID_over_RAID_5%3F —Preceding unsigned comment added by 96.42.234.28 (talk) 02:22, 17 May 2009 (UTC)

BeyondRaid ans SHA1
"BeyondRaid offers a RAID 6–like feature and can perform hash-based compression using 160-bit SHA1 hashes to maximize storage efficiency." WTF? Hashing isn't compression! Something is wrong, or needs better explanation in article.

BTW. Linux MD, also supports creating AFAIK all raid levels on drives of different sizes, and will do not waste space. It isn't recomended probably (as different areas of array will have different performance and reliability properties), but it works. —Preceding unsigned comment added by 91.213.255.7 (talk) 13:09, 20 April 2011 (UTC)

Matrix Raid
It is my understanding that Intel Matrix Raid supports more than 2 disks, and more than just RAID 1 and RAID 0 volumes.

This section seems to imply that Matrix Raid = 2 disks with a RAID 1 and a RAID 0 volume - this is misleading.


 * Microsoft Windows and Novell Netware OS software RAID follows a similar scheme whereby disks can be divided into partitions which are then combined into a RAID volume. For instance Novell Netware allows you to split three disks into two partitions each, then mirror the partitions against each other, then span a volume across the three mirrors, creating a RAID similar to 1E but without the block striping. --81.138.95.60 (talk) 08:33, 29 July 2011 (UTC)

Overdefined polynomial
Anyone know what "overdefined polynomial which characterizes the data" means? Timbatron 07:20, 21 February 2007 (UTC)
 * Try looking at Secret sharing —The preceding unsigned comment was added by 217.147.80.137 (talk) 20:34, 10 March 2007 (UTC).
 * There isn't any reference to the term "overdefined polynomial" in that article. I have a vague understanding of what it means, but there are nearly no references to the term "overdefined polynomial" on google except to this article (and copies of it). Someone more knowledgable about this topic should change this to something in more common usage I think... Timbatron 16:40, 22 March 2007 (UTC)

'overdefined polynomial' is referring to a system of linear equations where you have more linear equations than variables. Basically, you have more information than you need to solve the problem. In the context of this article, it is saying that the additional parity is being used as parity on different data than the first parity, rather than being redundant like the two parities in RAID6. My suggested wording for the sentence is "Unlike RAID 6, the second set is not used for redundant error correction." 74.93.207.201 14:04, 12 July 2007 (UTC)

Might I suggest, "Unlike in RAID 6, the second set is not used for redundant error correction. Instead it serves as an alternate set of error correction, calculated from a different set of data." Or, perhaps, "... from a different group of data blocks." Unfortunately, terms like 'overdefined polynomial', while a mathematically valid description, will result in the majority of people, who don't happen to be up on linear algebra, missing the important concepts in this section. These concepts can be explained just as well without the use of terms that most people might not understand. Ge0nk (talk) 23:34, 8 August 2011 (UTC)

RAID 10 ???
Here RAID 10 is several times mentioned as a standard RAID level. However neither RAID articel nor RAID standard levels article mention it.

For what i can tell, it usually is interpreted like mirror of stripes (RAID 1+0), however why cannot it be interpreted in more exotice ways ? Striping of mirros for example. Or two RAID 5 in stripes (RAID5 + RAID 5) Or RAID 5 made of stripes. Or whatever.

Either RAID 10 really had been standardised - in which case it should be mentioned in RAID and RAID-standard-levels articles. Or it had not been, then RAId 10 is to be determined what is it a beast in this article near each reference to it. —Preceding unsigned comment added by 212.44.136.228 (talk) 11:03, 25 April 2008 (UTC)

RAID 10 is mentioned in the Nested Raid Levels article, which is linked from the RAID article, iirc. --Bobbozzo (talk) 06:22, 5 July 2008 (UTC)

RAID 10 IS RAID 1 + RAID 0, i.e., a mirrored set of striped sets. This is completely standardized in the RAID vernacular, although it is occasionally written as RAID 1+0. Similarly, you can have RAID 0+1, which is a striped set of RAID mirrors (and, therefore, FAR less common than RAID 10, which is a relatively common way of protecting large striped sets.) Ge0nk (talk) 02:45, 19 August 2011 (UTC)

Parity bits vs checksum bytes?
Appropriate links or a discussion of the difference between the types of disk array parity blocks (some of which actually do contain parity bits, like RAID2 for example, but also some don't) would be extremely helpful here. — Preceding unsigned comment added by 205.153.180.229 (talk) 15:11, 1 November 2011 (UTC)

Windows 8 Storage Spaces
With the Consumer Preview of Win8 released I would suggest there should be a section on the 'Storage Spaces' feature and it's relation to WHS' Drive Extender. Doyna Yar (talk) 20:25, 14 March 2012 (UTC)

Propose moving RAID 6/Double parity to /wiki/Standard_RAID_levels
RAID 6 has become very common. Propose moving it to Standard_RAID_levels. Thoughts? Ktoth04 (talk) 21:27, 8 September 2012 (UTC)

RAID-Z automatically vs atomically
"it writes to a new location and then _automatically_ overwrites the pointer to the old data." Did it mean atomically? Jackieku (talk) 14:23, 1 January 2013 (UTC)


 * Reworded the sentence and this is now gone. 81.187.162.109 (talk) 23:30, 11 April 2013 (UTC)

UnRAID
This section doesn't provide any useful information on this RAID system. It may be that this is just proprietary software than a non-standard RAID level/methodology.


 * looks like it's covered now. --Bobbozzo (talk) 01:07, 3 July 2008 (UTC)


 * unRAID bascally seems like a slight variation of RAID 3 / 4? (except that it has separate volumes) 197.96.84.106 (talk) 07:05, 23 April 2013 (UTC)

No mention of X-RAID
It is a newer, proprietary (and patent pending) level introduced by Infrant Technologies, who were recently merged with Netgear. Since it is a patent pending technology, exact details aren't known, but it allows for dynamic volume size allocation without loss of data. i.e. You have a RAID set containing 3 500GB drives (SATA in their case). X-RAID would configure this as a ~1TB volume with parity protection against a single drive failure. You can simply add a 4th 500GB drive and it will add it to the existing volume, increasing it to a ~1.5TB volume with your data, and parity information, intact. I believe the process is handled by an embedded Linux OS using LVM. Because this is a "software" solution, it may not warrant a mention in this article. —The preceding unsigned comment was added by 216.76.89.144 (talk) 19:28, 14 May 2007 (UTC).


 * Why not? linux MD raid is software/FakeRAID too. 24.55.213.129 (talk) 04:09, 4 May 2013 (UTC)

Replaced citation 3298 with USENIX paper
I this cite because it didn't directly support the claim it was given to back up ("sometimes known as row diagonal parity"). In addition it was a dead link. Saved here (and reformatted as a web site from archive) in case it may be useful for other claims.



Instead I have given a cite found in that paper, which already backs it up by its title.



--SpecMade (talk) 21:11, 22 November 2013 (UTC)


 * Thumbs up! -- Dsimic (talk) 22:37, 22 November 2013 (UTC)

RAID-Z vs RAID 5
RAID-Z is often compared with RAID 5 (as it is in the article), however there are some serious shortcomings (especially when backing a database). I was considering adding:


 * RAID-Z does not increase random read IOPS performance in the same way that RAID 5 can - it performs near the same rate as just a single disk in the array.

Note: (Due to random write "aggregation″, random write performance is similar in performance to RAID 5 (or even better?))

Other things I have noted from Adam Leventhal's comparison - youtube slides:
 * RAID-Z resilvering is faster than RAID 5 for mostly empty disks, but slower for a full rebuild (due to walking the file structure - RAID 5 is done using continuous access for the whole disk, RAID-Z by random access of the used area of disk).
 * RAID-Z space accounting is complicated - For Z disks, array capacity for RAID 5 is fixed at (N-1). Capacity for RAID-Z array is between (N-1) when doing large writes to (N/2) for a disk full of 1-block writes.  For a disk full of small writes, RAID-1 could give the same capacity (and better random read IOPS)

E.g. Assuming 100GB disks


 * Hello there! These additions are looking good to me, please feel free to add them into the article. &mdash; Dsimic (talk | contribs) 05:53, 31 March 2014 (UTC)


 * Thanks Dsimic. Added now. Ajoiner (talk) 07:06, 2 April 2014 (UTC)


 * You're welcome. Though, we'll need to provide better references. &mdash; Dsimic (talk | contribs) 07:12, 2 April 2014 (UTC)

Inaccurate reference about RAID-DP
The RAID-DP section references "Hitz, Dave; Bhargava, Akshay (February 2006). "A Storage Networking Appliance". NetApp.com. Network Appliance. Retrieved 2014-06-07." as validation for the claim that RAID-DP overhead is only 2% greater than typical RAID-4 systems. I can't find any support for the claim in that document. It doesn't say 2% anywhere. A valid, relevant reference should be supplied or the claim removed as unsupported. — Preceding unsigned comment added by 50.169.160.33 (talk) 23:42, 3 July 2014 (UTC)

RAID-Z
I had made a large edit to the information in the RAID-Z section less than 6 hours ago. This information, and the remainder of the section, was subsequently deleted in its entirety. I assume this was due to uncited material. I am attempting to update the section and was simply cleaning up the format to start. can User:Voidxor please explain why such a large amount of info was removed without consulting the talk page prior? certainly RAID-Z is a important topic on the subject of non-standard RAID levels, no? is it not acceptable to take the current, unsourced information and edit it as required with proper citations? deleting the entire section is most unhelpful in this regard. please allow a little more time for discussion before deleting the contributions of others, it is only my intention to improve the quality of the content in this article.. thanks -Jchap1590 (talk) 07:27, 26 October 2014 (UTC)

I have the following references to offer, although I'm not sure of their validity: https://calomel.org/zfs_raid_speed_capacity.html, http://pages.cs.wisc.edu/~remzi/Classes/736/Fall2007/Projects/BrianKynan/paper.pdf, https://blogs.oracle.com/ahl/entry/what_is_raid_z Jchap1590 (talk) — Preceding undated comment added 09:15, 26 October 2014 (UTC)

EDIT: here are some more good ones right from Oracle: https://blogs.oracle.com/ahl/entry/double_parity_raid_z, https://blogs.oracle.com/ahl/entry/triple_parity_raid_z, https://blogs.oracle.com/bonwick/entry/raid_z, https://blogs.oracle.com/relling/entry/zfs_raid_recommendations_space_performance, https://blogs.oracle.com/relling/entry/raid_recommendations_space_vs_mttdl Jchap1590 (talk) 09:50, 26 October 2014 (UTC)


 * I removed the RAID-Z section (including your addition) because it was uncited. Please be sure to cite your sources when adding content in the future. Without references, nobody can fact check your work. Worse yet, you could have plagiarized (which opens Wikipedia up to legal action)! Furthermore, that is exactly what gives Wikipedia a bad reputation with a lot of people for being dubious. It is the author's duty to add supporting references. Please don't expect other editors to come along behind you and scramble to fact check, plagiarism check, and find sources for you. I know it's hard work, but that's why it's unfair for you to expect me or any other editors to do it for you.


 * As far as you demanding that I explain my actions, I did so in my edit summary. The edit summaries on the history page are the first place you should look for an explanation of why any change was made.


 * As far as warning, please read the maintenance tags. The RAID-Z section had been tagged Uncited section since April. That's effectively six months of fair warning that the content was subject to removal if references weren't added. Sadly, people keep on writing content from their own personal knowledge as if this is their private blog, rather than a fact-checked encyclopedia.


 * I hope I answered your question. You may certainly reintroduce any content that I removed so long and you add suitable references as well. One of the references that you offer is a link to a blog, and blogs are almost never suitable references. I would not use that one. Let me know if you need any additional help, though. – voidxor (talk &#124; contrib) 23:59, 26 October 2014 (UTC)


 * Just as Voidxor explained it above, it's all about providing references. Though, with RAID-Z there seems to be very few good references, for some reason; for example, a related part of the Oracle's publicly accessible documentation is quite brief and refers to a couple of blog posts and a wiki page for further information.  I've looked around for some time, and pretty much all that pops up for RAID-Z is a mix of blog posts, together with a few wiki pages.  The only reference that lives up to Wikipedia's stadards is the one Jchap1590 already provided above –  ZFS and RAID-Z: The Über-FS?. &mdash; Dsimic (talk | contribs) 01:58, 27 October 2014 (UTC)


 * Also, there's this document, which also provides some RAID-Z information. &mdash; Dsimic (talk | contribs) 02:22, 27 October 2014 (UTC)

It was never my intention to leave uncited content in the article. I'm not asking anyone to verify my edits to the article. As I said, I was simply cleaning up the format to have a basis to work with, find reliable sources for what was already presented, and finally omit any unreliable information. The section was removed before I could finish any substantial amount of research. RAID-Z seems like pertinent topic on the subject of non-standard RAID, so I feel it should be included in the article; however brief. Jchap1590 (talk) 04:02, 27 October 2014 (UTC)


 * RAID-Z probably does belong here; nobody's arguing that. Dsimic and I are just trying to convey that references are needed. I've heard the, "Oh, I was going to add references later..." excuse before. Why not try building your data and references in a sandbox, and then move the fruits of your labor into this article when you're done? That way, your work won't be interrupted by other editors. – voidxor (talk &#124; contrib) 04:35, 27 October 2014 (UTC)

I came to the Wikipedia page to learn about RAID-Z and, when I saw there were almost no references in the section, I decided to start researching and improve the article. I fully understand the purpose and importance of references. Thanks for the tip about using sandbox. Jchap1590 (talk) 18:39, 28 October 2014 (UTC)


 * Yeah, just go ahead and we'll help with everything that's needed so we end up with a good and well-referenced description of RAID-Z in this article. &mdash; Dsimic (talk | contribs) 19:19, 29 October 2014 (UTC)

I'm redirecting RAID-Z3 and creating the RAID-Z2 redirect to here. The ZFS article that RAID-Z3 previously redirected to has some information on RAID-Z already, particularly in this section. 93 00:37, 23 January 2015 (UTC)

This is complete BS. If Voidxor were serious about improving wikipedia, he/she/it would research and add citations rather than mass deleting content. The old days, the "build phase" of wikipedia were so much better before the asshats banded together to ruin it (and yes, that can be cited, google the myriad problems with the wikipedia community). RAID-Z exists, the page redirects here, I know little about it, and I should get to read something about it considering that a few well meaning people have put work into educating us. Leave the uncited tag up and the reader can decide how much they trust the info. If you don't trust the info, change the specific things you don't trust to more trustworthy material that you have uncovered. Add your work to wikipedia, do not delete other people's work. 68.173.49.156 (talk) 17:33, 29 January 2015 (UTC)