There has been lots of documentation already written about RAID storages, why this one more you can ask. Alright, I want to write it clear, creatively and comprehensive for all whoever wants to understand how RAID works.
In the 1990’s there was a cult TV show “Twin Peaks”. Beautiful girls handsome men, beautiful Laura Palmer, FBI agent Cooper dreamt of by many women. Good and evil in a small town Twin Peaks.
In the second season of the show an evil genius Windom Earle emmerges. A former FBI agent and coleague of agent Cooper.
Windom was a great chess player and a proficient expert in poetry. He writes a poem by Percy Shalley on a sheet of paper, stripes the sheet in three parts and sends the each to the three Queens of Twin Peaks.
Season 2, part 16.
None of the girls could read the poem until they had met in the Roadhouse and put all the three stripes together. Only then they could read it.
So, they read Percy Bysshe Shalley’s poem “Love’s Philosophy”.
|
See the mountains kiss high Heaven And the waves clasp one another; No sister-flower would be forgiven If it disdained its brother; And the sunlight clasps the earth And the moonbeams kiss the sea: What is all this sweet work worth If thou kiss not me? |
If we ilustrate “RAID 0” in a table these three stripes of the poem would look the following way.
| Audrey: ("DISK 0" or "D0") | Donna: ("DISK 1" or "D1") | Shelly: ("DISK 2" or "D2") |
| D0:B00: See the m | D1:B00: ountains kiss | D2:B00: high Heaven |
| D0:B01: And th | D1:B01: e waves clasp | D2:B01: one another; |
| D0:B02: No sister | D1:B02: flower wo | D2:B02: uld be forgiven |
| D0:B03: If it di | D1:B03: sdained its | D2:B03: brother; |
| D0:B04: And th | D1:B04: e sunlight | D2:B04: clasps the earth |
| D0:B05: And | D1:B05: the moonbea | D2:B05: ms kiss the sea: |
| D0:B06: What | D1:B06: is all this sw | D2:B06: eet work worth |
| D0:B07: If | D1:B07: thou kiss no | D2:B07: t me? |
Technically speaking “RAID 0” rrequires at least two disks, in our case we have three, since the girls were three. “RAID 0” is block oriented, i.d. it reads/writes in blocks.
“RAID 0” has an excellent performance as the data is striped in blocks evenly, and thus with the three disks each of the disks has thrice less input/output operations, the speed is thrice higher as well. By organizing a “RAID 0” array of three physical disks you can approach the speeds of an SSD (Solid-State Drive), since the reads are performed simultaneously from all the drives of the arrays. So if one HDD can output 25mb per second, the three disks will output 75 logical mega bytes per second. In fact “RAID 0” is fastest of all arrays.
“RAID 0” has NO redundancy, has NO mirroring, has NO parity. “RAID 0” just simply evenly divides the input/output load between the available disks in the array.
“RAID 0” is highly UNrecommended for usage on critical systems where the storage reliability is of the utmost importance! As soon as any single disk is broken — all the data are lost with no single chance to recover at all. You won’t be able to read from the other healthy disks, for the data would be considered inconsistent without a broken disk of the array.
Just the same way as none of the girls could read the poem without two others, as well as any two girls would be unable to read the poem without any third of them.
The total storage space is the sum of all the disks. If you have three disks 1TB each, so you have 3TB for storage.
In the "RAID 0" the data are written on the disks as lined up stripes, it means that if you use the disks of different sizes in the array, say, 500GB, 1TB, 750GB - then the size of the array will be 1.5TB. That is the smallest disk will be multiplied by three, thus 500GB of the 1TB disk and 250GB of the 750GB disk will be wasted.
“RAID 1” also requires two disks minimum. “RAID 1” is block oriented, i.d. it reads/writes in blocks. The blocks are mirrored. The redundancy is excellent, NO striping, NO parity. “RAID 1” maintais duplicate sets of all data on each disk of the array and represents them as one logical set of data.
The performance is good, but not excellent like for “RAID 0”. In fact “RAID 1” was designed rather for protection from data loss and redundancy than for speed and performance. Actually the disks in the array keep the identical copy of the data. When a write “operation” is performed it is done on the first disk and then the copy is written on the rest of the disks in the array. A “read” operation is satisfied by data delivery from any of the disk OR from all of them simultaneously as in “RAID 0”.
The total storage space is 50% of the capacity of all the drives in the array.
If you have three disks 1TB each, so you have 1.5TB for storage.
In case of an array composed of three discs of different sizes - 500GB, 1TB and 750GB - the usable space will be equal to 500GB. Again as in case with "RAID 0" 500GB and 250GB are wasted.
| Audrey: ("DISK 0" or "D0") | Donna: ("DISK 1" or "D1") | Shelly: ("DISK 2" or "D2") |
| D0:B00: See the m | D1:B00: See the m | D2:B00: See the m |
| D0:B01: ountains kiss | D1:B01: ountains kiss | D2:B01: ountains kiss |
| D0:B02: high Heaven | D1:B02: high Heaven | D2:B02: high Heaven |
| D0:B03: And th | D1:B03: And th | D2:B03: And th |
| D0:B04: e waves clasp | D1:B04: e waves clasp | D2:B04: e waves clasp |
| D0:B05: one another; | D1:B05: one another; | D2:B05: one another; |
| D0:B06: No sister | D1:B06: No sister | D2:B06: No sister |
| D0:B07: flower wo | D1:B07: flower wo | D2:B07: flower wo |
| D0:B08: uld be forgiven | D1:B08: uld be forgiven | D2:B08: uld be forgiven |
| D0:B09: If it di | D1:B09: If it di | D2:B09: If it di |
| D0:B10: sdained its | D1:B10: sdained its | D2:B10: sdained its |
| D0:B11: brother; | D1:B11: brother; | D2:B11: brother; |
| D0:B12: And th | D1:B12: And th | D2:B12: And th |
| D0:B13: e sunlight | D1:B13: e sunlight | D2:B13: e sunlight |
| D0:B14: clasps the earth | D1:B14: clasps the earth | D2:B14: clasps the earth |
| D0:B15: And | D1:B15: And | D2:B15: And |
| D0:B16: the moonbea | D1:B16: the moonbea | D2:B16: the moonbea |
| D0:B17: ms kiss the sea: | D1:B17: ms kiss the sea: | D2:B17: ms kiss the sea: |
| D0:B18: What | D1:B18: What | D2:B18: What |
| D0:B19: is all this sw | D1:B19: is all this sw | D2:B19: is all this sw |
| D0:B20: eet work worth | D1:B20: eet work worth | D2:B20: eet work worth |
| D0:B21: If | D1:B21: If | D2:B21: If |
| D0:B22: thou kiss no | D1:B22: thou kiss no | D2:B22: thou kiss no |
| D0:B23: t me? | D1:B23: t me? | D2:B23: t me? |
"RAID 0+1" means arrays implemented as "RAID 1", whose elements are "RAID 0" arrays. Such implementation has the benefits of "RAID 0" speed and "RAID 1" safety.
It is also much easier to implement than "RAID 3", "RAID 5" or "RAID 6". The main drawback of such a solution is its cost.
| Audrey: (Three Disks) | Donna: (Three Disks) | Shelly: (Three Disks) | ||||||
|---|---|---|---|---|---|---|---|---|
| D0:0 | D1:0 | D2:0 | D0:1 | D1:1 | D2:1 | D0:2 | D1:2 | D2:2 |
| See the m | ountains kiss | high Heaven | See the m | ountains kiss | high Heaven | See the m | ountains kiss | high Heaven |
| And th | e waves clasp | one another; | And th | e waves clasp | one another; | And th | e waves clasp | one another; |
| No sister | flower wo | uld be forgiven | No sister | flower wo | uld be forgiven | No sister | flower wo | uld be forgiven |
| If it di | sdained its | brother; | If it di | sdained its | brother; | If it di | sdained its | brother; |
| And th | e sunlight | clasps the earth | And th | e sunlight | clasps the earth | And th | e sunlight | clasps the earth |
| And | the moonbea | ms kiss the sea: | And | the moonbea | ms kiss the sea: | And | the moonbea | ms kiss the sea: |
| What | is all this sw | eet work worth | What | is all this sw | eet work worth | What | is all this sw | eet work worth |
| If | thou kiss no | t me? | If | thou kiss no | t me? | If | thou kiss no | t me? |
A "RAID 1+0" array is implemented as "RAID 0", whose elements are "RAID 1". It combines the same advantages of "RAID 0" (speed) and "RAID 1" (safety), but in different way. "RAID 1+0" creates a large stripe of small mirrors. After failure of a disk, while it is being replaced, only the small fragment of a whole array is rebuilt.
The drawback is the same as of "RAID 0+1" - high cost of implementation.
| Audrey: (Three Disks) | Donna: (Three Disks) | Shelly: (Three Disks) | ||||||
|---|---|---|---|---|---|---|---|---|
| D0:0 | D1:0 | D2:0 | D0:1 | D1:1 | D2:1 | D0:2 | D1:2 | D2:2 |
| See the m | See the m | See the m | ountains kiss | ountains kiss | ountains kiss | high Heaven | high Heaven | high Heaven |
| And th | And th | And th | e waves clasp | e waves clasp | e waves clasp | one another; | one another; | one another; |
| No sister | No sister | No sister | flower wo | flower wo | flower wo | uld be forgiven | uld be forgiven | uld be forgiven |
| If it di | If it di | If it di | sdained its | sdained its | sdained its | brother; | brother; | brother; |
| And th | And th | And th | e sunlight | e sunlight | e sunlight | clasps the earth | clasps the earth | clasps the earth |
| And | And | And | the moonbea | the moonbea | the moonbea | ms kiss the sea: | ms kiss the sea: | ms kiss the sea: |
| What | What | What | is all this sw | is all this sw | is all this sw | eet work worth | eet work worth | eet work worth |
| If | If | If | thou kiss no | thou kiss no | thou kiss no | t me? | t me? | t me? |
Above I showed you just a poetical representation of “RAID 0”, but before I continue to "RAID 5" I need to go a little deeper into technical details — about blocks, bytes and bits. Poetry will always stay with us, but we have to consider some technicals beyond the poetry, whatever - you cannot avoid it.
The data on the disks are well structured for quick access, the units on the disk where the data are stored are called blocks. Disks write information in blocks. Imagine that a page of your exercise book is a disk block. Each square is a byte. A byte is just a symbol, for example letter "S" is a byte represented in machine binary codes as "01010011", where those "zeros" and "ones" are bits.
So the letter "S" is a byte, represented with eight bits in binary code.
An invisible symbol "[SPACE]" in binary representation looks like: "00100000". Another "invisible" symbol "[new line]" (it looks like "\n") is "00001010", and so on...
Below is an exercise book page - just a clean disk block with no data.
When you create a new file and save it on the disk the system creates a symbolic link (file name) and connects it with the address of the block on the disk where the data of the file are actually stored.
| Audrey: ("DISK 0" or "D0") | Donna: ("DISK 1" or "D1") | Shelly: ("DISK 2" or "D2") |
| D0:B00: See the m | D1:B00: ountains kiss | D2:B00: high Heaven |
| D0:B01: And th | D1:B01: e waves clasp | D2:B01: one another; |
| D0:B02: No sister | D1:B02: flower wo | D2:B02: uld be forgiven |
| D0:B03: If it di | D1:B03: sdained its | D2:B03: brother; |
| D0:B04: And th | D1:B04: e sunlight | D2:B04: clasps the earth |
| D0:B05: And | D1:B05: the moonbea | D2:B05: ms kiss the sea: |
| D0:B06: What | D1:B06: is all this sw | D2:B06: eet work worth |
| D0:B07: If | D1:B07: thou kiss no | D2:B07: t me? |
| Audrey: ("DISK 0" or "D0") | Donna: ("DISK 1" or "D1") | Shelly: ("DISK 2" or "D2") |
| D0:B00: 01010011 01100101 01100101 00100000 01110100 01101000 01100101 00100000 01101101 | D1:B00: 01101111 01110101 01101110 01110100 01100001 01101001 01101110 01110011 00100000 01101011 01101001 01110011 01110011 00100000 | D2:B00: 01101000 01101001 01100111 01101000 00100000 01001000 01100101 01100001 01110110 01100101 01101110 |
| D0:B01: 01000001 01101110 01100100 00100000 01110100 01101000 | D1:B01: 01100101 00100000 01110111 01100001 01110110 01100101 01110011 00100000 01100011 01101100 01100001 01110011 01110000 00100000 | D2:B01: 01101111 01101110 01100101 00100000 01100001 01101110 01101111 01110100 01101000 01100101 01110010 00111011 |
| D0:B02: 01001110 01101111 00100000 01110011 01101001 01110011 01110100 01100101 01110010 00100000 | D1:B02: 01100110 01101100 01101111 01110111 01100101 01110010 00100000 01110111 01101111 | D2:B02: 01110101 01101100 01100100 00100000 01100010 01100101 00100000 01100110 01101111 01110010 01100111 01101001 01110110 01100101 01101110 |
| D0:B03: 01001001 01100110 00100000 01101001 01110100 00100000 01100100 01101001 | D1:B03: 01110011 01100100 01100001 01101001 01101110 01100101 01100100 00100000 01101001 01110100 01110011 00100000 | D2:B03: 01100010 01110010 01101111 01110100 01101000 01100101 01110010 00111011 |
| D0:B04: 01000001 01101110 01100100 00100000 01110100 01101000 | D1:B04: 01100101 00100000 01110011 01110101 01101110 01101100 01101001 01100111 01101000 01110100 00100000 | D2:B04: 01100011 01101100 01100001 01110011 01110000 01110011 00100000 01110100 01101000 01100101 00100000 01100101 01100001 01110010 01110100 01101000 |
| D0:B05: 01000001 01101110 01100100 00100000 | D1:B05: 01110100 01101000 01100101 00100000 01101101 01101111 01101111 01101110 01100010 01100101 01100001 | D2:B05: 01101101 01110011 00100000 01101011 01101001 01110011 01110011 00100000 01110100 01101000 01100101 00100000 01110011 01100101 01100001 00111010 |
| D0:B06: 01010111 01101000 01100001 01110100 00100000 | D1:B06: 01101001 01110011 00100000 01100001 01101100 01101100 00100000 01110100 01101000 01101001 01110011 00100000 01110011 01110111 | D2:B06: 01100101 01100101 01110100 00100000 01110111 01101111 01110010 01101011 00100000 01110111 01101111 01110010 01110100 01101000 |
| D0:B07: 01001001 01100110 00100000 | D1:B07: 01110100 01101000 01101111 01110101 00100000 01101011 01101001 01110011 01110011 00100000 01101110 01101111 | D2:B07: 01110100 00100000 01101101 01100101 00111111 |
The defaut minimal block size is 512 bytes, the size is defined during setup, and you can specify 512 bytes or 1024 or 4096 bytes. It depends which files you are going to keep on those disks. If you store big video files, so then you’d better specify a larger size of the block. For small files — a smaller size of the block would be more optimal.
Compare this with a long text, which takes every square on the page, you do not leave blank spaces on a page and paper is not wasted.
Or a short note on a page taking a couple of lines, and the rest of the page is wasted as a result.
When the operating system writes data on disk it divides the data in blocks, if your data are 1024 bytes long, then exactly two block will be filled. If your data are, say, 1300 bytes, then two 512 byte blocks will be filled exactly to their size, and the third block will contain the rest 276 bytes. 236 bytes of the block size are actually wasted.
You can finish a chapter by writing only a couple or so lines on a fresh page, but anyway the next chapter of your text you will begin at the next clean page. Though the previous page is almost clean the system considers it “written”, any new information will not be written on the clean space of that page, the rest of the block is just “lined up” by zeros, like a big “Z” below your handwritten text. Yes, it’s a sort of a loss of the disk space.
Let’s assume that the columns of the poems are blocks.
Just imagine that you can write 512 symbols on a page (letters, figures, spaces, commas, periods, etc.),
if you did not fill up the page and have to start a new chapter of your great book, you start it just on a new page, an leave the previous page clean with a trailing text from the previous chapter at the top. In the “RAID 5” blocks that unused space is filled in by zeros to gurantee the parity calculations.
Let’s look at the poem stripes below and correct it according to the blocks idea described above.
| Audrey: | Donna: | Shelly: |
| See the m0000000 | ountains kiss 00 | high Heaven00000 |
| And th0000000000 | e waves clasp 00 | one another;0000 |
| No sister 000000 | flower wo0000000 | uld be forgiven0 |
| If it di00000000 | sdained its 0000 | brother;00000000 |
| And th0000000000 | e sunlight 00000 | clasps the earth |
| And 000000000000 | the moonbea00000 | ms kiss the sea: |
| What 00000000000 | is all this sw00 | eet work worth00 |
| If 0000000000000 | thou kiss no0000 | t me?00000000000 |
| Audrey: | Donna: | Shelly: |
| See the m0000000 | ountains kiss 00 | PARITY_CHECK0000 |
| high Heaven00000 | PARITY_CHECK0000 | And th0000000000 |
| PARITY_CHECK0000 | e waves clasp 00 | one another;0000 |
| No sister 000000 | flower wo0000000 | PARITY_CHECK0000 |
| uld be forgiven0 | PARITY_CHECK0000 | If it di00000000 |
| PARITY_CHECK0000 | sdained its 0000 | brother;00000000 |
| And th0000000000 | e sunlight 00000 | PARITY_CHECK0000 |
| clasps the earth | PARITY_CHECK0000 | And 000000000000 |
| PARITY_CHECK0000 | the moonbea00000 | ms kiss the sea: |
| What 00000000000 | is all this sw00 | PARITY_CHECK0000 |
| eet work worth00 | PARITY_CHECK0000 | If 0000000000000 |
| PARITY_CHECK0000 | thou kiss no0000 | t me?00000000000 |