I use all three and for large scales they’re considerably faster than the alternative; which at some scales (peta/exabyte) is “none.”