We just piped the backups through pigz for compression; rapidgzip also exists for parallelized decompression (or any other compression algorithm you’d like to use, of course).