>>Well.. I never aware that RAID 5 is stripe at the byte level. But I seriously doubt it can spread evenly between HD (like RAID 0) because this is where you gain a true performance. No redundancy, no fault tolerant, no mirroring, just performance. And that's also why we cannot called RAID 0 as a "real" RAID
>
>If you look at the graphic, at the link I gave (http://www.acnc.com/04_01_05.html), it seems that in the example (5 hard disks), for the first 4 bits (?), the data is spread evenly among 4 disks, whereas the 5th. hard disk provides redundancy.
>
>Later, for other data blocks, the disks are changed (that is, another disk takes over the "redundancy" role).
>
>In the example, one would suppose that data can be read 4 times as fast as when you don't use RAID - and you have redundancy.
>
>> ...The controller will do it's job, it will switch automatically if one of the disk is damage. But if you don't replace it with another one than you won't get a good fault tolerance. But I maybe wrong about this.
>
>Yes, that is my understanding, too. If one of the disks gets damaged, it has to be replaced ASAP.

It seems like if we use at least 5HD with RAID 5, the data can be spread evenly between HD. Well, this is actually what I thought about RAID 5+0. But if we can also do it with RAID 5, then why RAID 5+0 exist ? I think I have to find out more about it, and get a clearer information.

Regards