On Sun, 2006-12-03 at 12:13 +0800, Hadders wrote: > David G. Miller wrote: > > Ubence Quevedo <r0d3nt@xxxxxxxxxxx> wrote: > > > >> Hi Everyone, > RAID 0 - striping, two disks, add them together to get the container > size. With just two disks, no failover, any disk fails and your > container is toast. You can implement 0+1, which is striping plus > mirroring, but then you need four disks. > RAID 1 - mirroring, two disks, one is usually master, other is slave and > gets a copy of everything that does on the master, any disk fails you > keep on ticking and rebuild once you get a replacement. > RAID 5 - minimum of three disks, two data, one for parity (end up with > size roughly equal to two disks), however can span as many disks as need > and make very large containers. One disk can fail and your container > still works, but in degraded mode, get the replacement and rebuild on > the fly > RAID 6 - less used, but like 5, but handles more than a single disk failure. Thanks for this information. I will have to look closer at RAID 6 for my new file server. I was looking at Hardware solutions and the only reason that I can find is the controllers for the drives. The biggest reason that I have found to be against any hardware solution is dependency on that solution if there is a problem. If you have a motherboard or RAID card fail, you have to get the same controller to continue to access your drives. With software RAID, just load the software on any computer and your drives are available if you can access the drives. Finding a replacement controller could be a costly headache.