fleureck@imec.be (Marc Fleureck) (06/04/91)
Hello, Recently I copied a directory on a SR10.2 DNXXX system to an SR10.3 HP400 system. The directory on the SR10.2 was about 60 M in size, but when I did a "du" on the HP400 it gave a size of about 104 Mbytes. If I check with "df" on the HP400 after the copy, I can see that about 70M has been added to the disk in stead of the expected 60M. I suppose HP changed the block sizes of the file system ? So the final and strange question is : how much disk space has been allocated ? 104M, 70M or 60M. _______________________________________________ | | | _ /\ | | Marc Fleureck || __||__ | | Sys Admin Unix || | _||_ | | | IMEC vzw _||_||_||_|| | | Belgium |_||_||_||_|| | | ||_||_||_||_ | | |____||_||_| | | || || | | | | E-Mail : fleureck@imec.be IMEC vzw | |_______________________________________________|
thompson@PAN.SSEC.HONEYWELL.COM (John Thompson) (06/05/91)
> Recently I copied a directory on a SR10.2 DNXXX system to > an SR10.3 HP400 system. The directory on the SR10.2 was > about 60 M in size, but when I did a "du" on the HP400 it > gave a size of about 104 Mbytes. > > If I check with "df" on the HP400 after the copy, I can see > that about 70M has been added to the disk in stead of the expected > 60M. > > I suppose HP changed the block sizes of the file system ? > So the final and strange question is : how much disk space has been > allocated ? 104M, 70M or 60M. None of the above. The real answer is "42." :-) (See Hitchhikers Guide To The Galaxy if you're confused) Seriously, there were some changes, and there are other reasons your disk space could show up weird. 1) At 10.3, the disk block size went from 1K to 4K (except on the DN10K, which always used a 4K block size). This means that a 1 byte file will take up 4K bytes on a 10.3 formatted disk, and 1K byte on a 10.[012] formatted disk. If you have a lot of tiny files, this can give you a 4X increase in disk usage. We noticed about a 10% increase in our average, which is what HP/Apollo said to expect, too. (I can't remember where I read about this, but I remember reading about it....) 2) All of the disk-usage routines give "bad" results when you have hard-links. The routines don't bother checking whether an object has been counted once, so they count its space for every entry in the listing. We are using up 740MB of our 440MB disk, and it still has 45MB free! 3) The Unix 'ls' command gives the total space of a sparse object, not just the allocated part. This may cause some "Huh?" as well. (In fairness, Aegis lists the total space too, but they also list the blocks used, which is a more reliable indicator of actual usage. -- jt -- John Thompson Honeywell, SSEC Plymouth, MN 55441 thompson@pan.ssec.honeywell.com When in danger, when in doubt -- run in circles, scream and shout.
krowitz@RICHTER.MIT.EDU (David Krowitz) (06/05/91)
SR10.2 systems allocate disk space in 1Kb blocks for the Motorola 68020/68030 machines. The newer 68040 based HP400 series machines have a slightly different memory management unit that wants to use 4Kb blocks. If you directory has lots of small files in the 1 to 2 Kb range, then the HP400 series machine will wind up allocating a lot more disk space than the DN3xxx/4xxx machines. -- David Krowitz krowitz@richter.mit.edu (18.83.0.109) krowitz%richter.mit.edu@eddie.mit.edu krowitz%richter.mit.edu@mitvma.bitnet (in order of decreasing preference)
rees@pisa.citi.umich.edu (Jim Rees) (06/06/91)
In article <9106051318.AA01565@pan.ssec.honeywell.com>, thompson@PAN.SSEC.HONEYWELL.COM (John Thompson) writes:
3) The Unix 'ls' command gives the total space of a sparse object, not just
the allocated part.
I don't think that's true. The length of the file (as reported by 'ls -l')
is just the seek key of eof (to use Aegis-speak). But the size (as reported
by 'ls -s') is the true (allocated) size (modulo 1k/4k bugs in stat()).
The 'ls' man page is wrong about this, by the way, or at least misleading.
It uses the term "size" to mean both length and size. This concept is
confusing enough without the documentation contributing to the chaos.wjw@ebh.eb.ele.tue.nl (Willem Jan Withagen) (06/06/91)
In article <9106051345.AA08802@richter.mit.edu>, krowitz@RICHTER.MIT.EDU (David Krowitz) writes:
=> SR10.2 systems allocate disk space in 1Kb blocks for the
=> Motorola 68020/68030 machines. The newer 68040 based
=> HP400 series machines have a slightly different memory
=> management unit that wants to use 4Kb blocks. If you
=> directory has lots of small files in the 1 to 2 Kb range,
=> then the HP400 series machine will wind up allocating
=> a lot more disk space than the DN3xxx/4xxx machines.
Well I'd think a usenet-news system would fall in this category. :{
Haven't checked it, but my guess is that about 90% is less than 1K.
This would shure be a killer on an HP400 (or is it all 10.3)
I just checked:
/bin/pagesize on DN4500 gives 1024
/bin/pagesize on HP400 gives 4096
So I won't get into trouble yet when upgrading my news-gate DN4500 from
10.2 to 10.3.
But when I go and put it on a HP400, can I create a seperate disk-partition
with 1k blocksize just for the news volume??
I know that this should be a trivial question, since it has to be done
on any Unix-box, except of course Apollo's.
Willem Jan
--
Eindhoven University of Technology DomainName: wjw@eb.ele.tue.nl
Digital Systems Group, Room EH 10.10
P.O. 513 Tel: +31-40-473401
5600 MB Eindhoven The Netherlandschen@digital.sps.mot.com (Jinfu Chen) (06/11/91)
In article <1202@eba.eb.ele.tue.nl> wjw@eb.ele.tue.nl writes: > /bin/pagesize on DN4500 gives 1024 > /bin/pagesize on HP400 gives 4096 > >So I won't get into trouble yet when upgrading my news-gate DN4500 from >10.2 to 10.3. Not unless your DN4500 will be upgraded to DN5500 (040 based). DN5500 will use the same 4096K page size and your harddisks on DNX500 will be reformatted as well.
wjw@ebh.eb.ele.tue.nl (Willem Jan Withagen) (06/11/91)
In article <676362150.43@egsgate.FidoNet.Org>, John.Thompson@f98.n250.z1.FidoNet.Org (John Thompson) writes: => => > Recently I copied a directory on a SR10.2 DNXXX system to => > an SR10.3 HP400 system. The directory on the SR10.2 was => > about 60 M in size, but when I did a "du" on the HP400 it => > gave a size of about 104 Mbytes. [And more] This article has passed us before!! but then with the right origin. It seem like somebody has setup an echo somewhere in the fidonet. I guess that the person doing this has not completly got it right. So could this be corrected. Ciao Willem Jan The path is: eba!svin02!hp4nl!mcsun!unido!math.fu-berlin.de!ira.uka.de!\ sol.ctr.columbia.edu!zaphod.mps.ohio-state.edu!rpi!\ news-server.csri.toronto.edu!torsqnt!lethe!telly!moore!eastern!egsgate!Uucp -- Eindhoven University of Technology DomainName: wjw@eb.ele.tue.nl Digital Systems Group, Room EH 10.10 P.O. 513 Tel: +31-40-473401 5600 MB Eindhoven The Netherlands
hj412fr@unidui.uni-duisburg.de (Frik) (06/11/91)
In <52175350.3593b@digital.sps.mot.com> chen@digital.sps.mot.com (Jinfu Chen) writes: >In article <1202@eba.eb.ele.tue.nl> wjw@eb.ele.tue.nl writes: >> /bin/pagesize on DN4500 gives 1024 >> /bin/pagesize on HP400 gives 4096 >> >>So I won't get into trouble yet when upgrading my news-gate DN4500 from >>10.2 to 10.3. >Not unless your DN4500 will be upgraded to DN5500 (040 based). DN5500 will >use the same 4096K page size and your harddisks on DNX500 will be reformatted >as well. I am confused: We also received an 68040 upgrade but have not installed it yet as we do not have the new compilers yet. NOW, will we have to REFORMAT our disk to run the /sau14 operating system??? -- Martin Anantharaman FB7, FG7 (Mechanik) Work: +49 (203) 379-3336 Universitaet -GH- Duisburg Home: +49 (203) 37 65 89 Lotharstr. 1 FAX: +49 (203) 379-3052 4100 Duisburg 1 E-Mail: martin@mechanik.uni-duisburg.de West Germany hj412fr@duc220.uni-duisburg.de hj412fr@unidui.uni-duisburg.de