14.122 Nested RAID Levels 1+0; 10; 0+1; 01
Вставка
- Опубліковано 15 вер 2024
- My book "Patterns in Data Management" is now available both as an ebook or a print book (with color graphics!). See: amzn.to/1Ts3rwx This book is not a standard textbook on database techniques. This book was written extending and complementing preexisting educational videos (including this one).
What is RAID 10? Why would we call this nested RAID? And what is RAID 01 then? Is it possible, in principal, to combine any kinds of RAID-levels?
Video for my inverted classroom "Database Systems".
The complete list of videos and additional material is available at datenbankenlern...
Computer Science, Saarland University:
Bachelor (in German): www.cs.uni-saar....
Master (in English): www.cs.uni-saar....
Ph.D./Grad School: gradschool.cs.u...
Notice: B1,B2~B8 are all partition of file B.
RAID 0+1:
First, think about 1(mirror), we have B and Bm. (Bm means the mirror of B)
Second, striping: B1~B8, Bm1~Bm8
RAID 1+0:
First, think about 0(striping), we have B1~B8
Second, mirror: B1~B8, Bm1~Bm8
Generally, RAID 1+0 performs better.
eg. B1 and Bm2 are broken, RAID 1+0 can work because Bm1 and B2 are still good to use.
But RAID 0+1 can not, because B1~B8 is broken due to B1, and Bm1~Bm8 is broken too due to Bm2
all got from this video. Thanks for uploading.
Thanks helped a lot in a short time. Thumbs up.
Thanks, this is very helpful!
come here after bytebytego explaining discord sets raid01 to its new db
but which is best to use RAID 10 or 01 ?
+Akash Jadhav the german wikipedia claims RAID 10 is more failsafe than RAID 01, and RAID 10 can recover faster from failures. sadly they dont justify their claims. and i dont know either. maybe @Jens Dittrich can tell us more?
@@immanuelhaffner6915 RAID 10 is indeed more failsafe, the reason is that RAID 10 is essentially a stripe of mirrors, if it loses a drive the the stripe and both mirrors survive albeit one mirror is degraded, at this point a drive failure on the degraded mirror will kill the array however a drive failure on the good mirror is survivable leaving a good disk in each mirror and the stripe intact, this gives RAID 10 a 2:3 chance of surviving a second drive failure (66% chance).
RAID 01 by comparison is a mirror of stripes, if it loses a drive then the mirror survives but one of the stripes is completely dead. This leaves just two drives online (in the good stripe) and if either of those drives fail the stripe will die and the mirror with it, giving RAID 01 a 0:2 chance of surviving a second drive failure (0% chance).
@Lord_Edge thanks for the answer (on a 3 years old comment :D)
I agree with your argument about RAID10. In RAID01, however, one disk in the broken RAID0 is still intact. Although of no use, it is still operable. So I would say, that the already damaged RAID01 can still survive one more disk failure, namely that intact disk in the broken RAID0. The chance that the next disk that fails is this particular disk is then 1/3, or 33%.
Still, your conclusion still stands that RAID10 is more failsafe than RAID01. Thanks :)
I feel like I should throw my take in here. In RAID 01 you need to read from both drives in the good striped array to rebuild the data in the bad RAID 0 array. In RAID 10 you just need to re-mirror the bad RAID 1 array of a single stripe. Having another failure in 01's good array is complete failure, while having another failure in 10's bad array is complete failure. 01's good array has two drives, 10's bad array has one drive left. If another drive were to fail while rebuilding, 01 has a 2/3 chance of complete failure, 10 has a 1/3 chance.
@@llynellyn but lets say data in 01 is spread as
a a b b
If disk 1 fails, then we can still retrieve data from disk2, and we still have a chance to reconstruct data if disk 3 / 4 fail since now we will be left with
x a x b
Or
x a b x
So we can still reconstruct data, how does it have a 0 chance of surviving a second disk failure?
My teacher couldnt do better ...thanks