ehrlich@psuvax1.cs.psu.edu (Dan Ehrlich) (07/06/90)
I just noticed that when running newfs the file system size that is reported is about 105% of what it should be. For example # newfs /dev/rid000h /dev/rid000h: 159900 sectors in 130 cylinders of 15 tracks, 82 sectors 81.9MB in 9 cyl groups (16 c/g, 10.08MB/g, 4608 i/g) Using my trusty TI-55-II calculator to compute (159900*512)/1048576 yields a result of 78.076172MB. Even worse 9g * 10.08MB/g = 90.72MB, not 81.9MB. This is off by about 12%. Can anyone explain to me why these values would be so far from what it would appear they should be? Thanks. Dan Ehrlich <ehrlich@cs.psu.edu>/Voice: +1 814 863 1142/FAX: +1 814 865 3176
jms@tardis.tymnet.com (Joe Smith) (07/11/90)
In article <9702@brazos.Rice.edu> ehrlich@psuvax1.cs.psu.edu (Dan Ehrlich) writes: >/dev/rid000h: 159900 sectors in 130 cylinders of 15 tracks, 82 sectors > 81.9MB in 9 cyl groups (16 c/g, 10.08MB/g, 4608 i/g) >Using my trusty TI-55-II calculator to compute (159900*512)/1048576 yields >a result of 78.076172MB. Even worse 9g * 10.08MB/g = 90.72MB, not 81.9MB. 159900 * 512 = 81,868,800 = 81.9*10^6 bytes 16 cyls * 15 tracks * 82 sectors * 512 bytes = 10,076,160 = 10.08*10^6 bytes 130 cyls / (16 cyls/group) = 8.13 groups, must round up to 9 cyl groups Given that 1M = 1,000,000 those numbers are accurate. A lot of vendors are using the convention that "1 megabyte" of memory is exactly 1,048,576 bytes, but "1 megabyte" of disk is 1,000,000 bytes. (It's more impressive on the sales literature to say that you sell an 81 megabyte drive instead of a 78 megabyte drive.) Joe Smith (408)922-6220 | SMTP: jms@tardis.tymnet.com or jms@gemini.tymnet.com BT Tymnet Tech Services | UUCP: ...!{ames,pyramid}!oliveb!tymix!tardis!jms PO Box 49019, MS-C41 | BIX: smithjoe | 12 PDP-10s still running! "POPJ P," San Jose, CA 95161-9019 | humorous dislaimer: "My Amiga speaks for me."
nobody@kodak.com (07/20/90)
In article <9843@brazos.Rice.edu> jms@tardis.tymnet.com (Joe Smith) writes: >130 cyls / (16 cyls/group) = 8.13 groups, must round up to 9 cyl groups One handy thing that I learned by taking a one morning tutorial from Dr. Kirk McKusick on the fast file system was what cylinder groups were and why they existed. It was a few years ago, but I did remember that a cylinder group is an entity that is a reference for certain things such as the number of inodes/group. So there is no such thing as a fractional cylinder group, you may have 8 cylinder groups that are of the size 16 cylinders and 1 cylinder group which has only 2 cylinders in it, but you do really have 9 cylinder group. Newfs makes sure of that. One of the other things that is done by newfs in each cylinder group is to allocate the inodes. Things may have changed with mkfs (please correct me if I am wrong), but mkfs didn't used to do fractional inode allocations. I.E. It will allocate 4608 inodes in your 2 cylinder - cylinder group, just the same as it will in a 16 cylinder cylinder group. (4608 i/g was contained in the original posting). If you ever have a disk on a new system that you can play with formatting, try making a cylinder group with just 2 cylinders in it. You would be amazed at how much of the file system is occupied by inodes. <S> - S curve - go slow. Personally, I collect up what I would have put in small cylinder groups, and often shift it into one partition to make at least a 8 cylinder - cylinder group. Otherwise I add it to the swap. I have tried to vary the number of cylinders per group (I did it a few years ago on an Ultrix system), and although the option is there in newfs and mkfs, it had absolutely *no* effect .. they all came out 16 cylinders per group. I have been out of it (system administration) for a while you see. So please test out some of the things that I have said, or check with someone with source. I still layout partitions always on even cylinder boundaries with the mininmum size of a cylinder group being 8 cylinders. Most of the time I layout all partitions so that all cylinder groups have all 16 cylinders in them. Any "left over" cylinders, I usually add to the swap (which doesn't have a file system on it) -- if the disk has a partition used for swap. Hope this helps. James H. Moore User Services Coordinator Eastman Kodak Company Rochester, NY 14653-5403 Email: jmoore@ssd.kodak.com