This is just free advice. Build a disk based index then compare with industry standard database storage engines under real world conditions. Compare your work to others in the field such as Tokutek's fractal tree. If a certain index is faster in one specific way most likely it is worse in others, such as creation time, resource requirements, etc.