On Tue, 19 Oct 2004 12:39:46 -0500 Hoyt Bailey <[EMAIL PROTECTED]> wrote:
> I will be interested in the replys you get on this since I have one > part with 21 % non-contiguous files. My understanding was that it 21% is extremely high, but not out of band with some earlier installs I did a long time ago on a seriously (by today's standard) limited drive. There used to be a 'frag' utility that would tell you what percentage of fragmentation overall you had on a particular partition. I saw fragmentation percentages of 40%-50%. Now I took that with a grain of salt, and I might just do that with your 21% percentage. At issue, I gather, with 'frag', was that it counted tertiary indexed files (any sufficiently large file would suffice) and miscounted them as fragmented when they weren't, or weren't as badly fragmented. On an ext2 filesystem, the biggest file you can have without going to a secondary index of pointers to blocks is about 13-14K per the kernel source. This is because there are that many pointers to disk blocks in the inode for a file. If your file is larger than that, you need a block to store pointers to blocks. And if it really is big, more than maybe 256K+14K (a 1k block could store at most 256*4 (sizeof an integer) pointers to disk blocks for a file) you need to go to tertiary indexed blocks, where you have pointers to pointers to blocks. Ordinarly, ext2 filesystems resist fragmentation, and even if a file is X% fragmented, it really doesn't have the same negative meaning as a fragmented file in FAT(32). Besides that, much of the overhead in a DOS system is involved in finding the first block of a file in a filesystem (much more time-consuming than on ext2fs) not in the head moving from point A to point B on the disk to get to other sections of the file. Obviously, ext2 does some things right, to minimize disk access, for the most part, logically related items are grouped together; for instance, various files in the same directory are grouped closer together than those in some other place on the filesystem. In FAT(32) there is little if any guarantee of this. Over time, there is the likelihood that a file FOO in UTILS may be right next to a file BAR in DOCS, and the next logical file in UTILS may be at the other end of the disk. And so ext2 doesn't guarantee that the filesystem will be utterly devoid of fragmentation; for that, you probably want to switch to an even better filesystem, such as reiserfs. -- ------------------------------------------------------------------------ David E. Fox Thanks for letting me [EMAIL PROTECTED] change magnetic patterns [EMAIL PROTECTED] on your hard disk. -----------------------------------------------------------------------
____________________________________________________ Want to buy your Pack or Services from MandrakeSoft? Go to http://www.mandrakestore.com Join the Club : http://www.mandrakeclub.com ____________________________________________________
