> For those who use the term "winraid" to descibe some kind of > inferior software raid, I would suggest that you explain what > you mean by don't use it, in light of this post. the "winraid" indeed is software raid. And yes it's generally inferior to linux native raid for RAID0; this has to do with the on disk format. Depending on the disk geometry (odds of hitting the "wrong" geometry are like 50%) and where you put the partition exactly, you can very well end up with a RAID0 stripe boundary being in the middle of a 4Kb block of the filesystem. This results that when the kernel does 1 4Kb IO, the hardware needs 2 IO's in practice, including 2 head seeks. This is actually measurable performance impact. Again this effect depends largely on how the start of the partition aligns with the stripes, which in turn depends on the geometry. Fwiw in linux this raid is supported by dmraid, but the installer side of things isn't quite there yet, nor is the bootloader side.
Attachment:
signature.asc
Description: This is a digitally signed message part