I'm afraid the question is utterly stupid, but why does one use block devices as RAIDframe components and not raw devices? I would think data is then buffered twice, once for the sd/wd/ld device and once again for the raid device.