Date: Fri, 2 Dec 2005 13:37:19 +0100 (CET) From: Wojciech Puchar <wojtek@tensor.3miasto.net> To: freebsd-questions@freebsd.org Subject: a bit of FFS questions Message-ID: <20051202132526.I52820@chylonia.3miasto.net>
next in thread | raw e-mail | index | archive | help
can someone help me. 1) newfs -a <more than 16> successfully creates the filesystem but dumpfs shows: . . . maxbsize 16384 maxbpg 2048 maxcontig 16 contigsumsize 16 what's the difference between maxcontig and contigsumsize? will FreeBSD kernel actually make user of maxcontig >16 (with MAXPHYS>128k it make sense) 2) -b <more than 16384> - does it make real sense? i tested with large files and it CPU load is small anyway, while transfer speed is the same. the answer may depend on answer for question 1 however. 3) -i <large value> - it would save space to use -i 16384, 32768 or more, and reduce fsck times. but if inode counts is closely tailored to amount of files, will it slow down filesystem and how much? i think if something like half or less inodes are allocated, OS have higher degree of freedom of where to allocate inode. but i imagine too that having even more free inodes doesn't make things. what is the ratio of used to free inodes, below which there will be just wasted space? 4) -o space/time - in my tests i can't see much difference in CPU load, maybe on 386 it was a difference but not on Athlon64 or P4. isn't -o space better to be always used? 5) -m <low percents> - does -m change anything in FFS algorithms, or it's just a quota? i would prefer to set it low just in case of urgent need of space, while trying to maintain 8-10 free space at normal operation. am i right?
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20051202132526.I52820>