DU 4.0 memory usage

From: Lou Muccioli <lou_at_mailhub.ansys.com>
Date: Thu, 12 Mar 1998 19:59:51 -0500

I have an AlphaServer 4100 running Digital UNIX 4.0A with 1 GB of RAM
and 2.8 GB of swap. The vm subsystem settings I have in
/etc/sysconfigtab are:

vm:
        vm-maxvas = 8589934592
        vm-vpagemax = 524288
        vm-mapentries = 1000000
        ubc-maxpercent = 50

What I don't seem to understand is I expect more RAM to be available than
what "top" or any other memory monitor says I actually have available.
For example:

top output displays:

load averages: 2.00, 2.06, 2.06 19:22:47
96 processes: 3 running, 12 sleeping, 80 idle, 1 zombie
CPU states: 50.0% user, 0.0% nice, 0.1% system, 49.8% idle
Memory: Real: 539M/1002M act/tot Virtual: 37M/2804M use/tot Free: 315M

  PID USERNAME PRI NICE SIZE RES STATE TIME CPU COMMAND
12219 jmv515 42 0 335M 79M run 30:41 99.90% ansys55t
17633 jmv515 44 0 335M 45M run 13:42 99.90% ansys.e55
 1156 lou 44 0 3032K 344K run 0:00 0.10% top
  854 root 39 -5 2568K 598K sleep 1:28 0.00% ansys_lm
  269 root 44 0 1680K 131K sleep 0:38 0.00% portmap
17273 root 44 0 1776K 90K sleep 0:09 0.00% mountd
  517 root 44 0 2120K 163K sleep 0:05 0.00% cvlm
  519 root 44 0 2112K 163K sleep 0:05 0.00% cadds
25882 lou 42 0 4488K 425K sleep 0:05 0.00% xterm
12917 jmv515 44 0 335M 385K sleep 0:01 0.00% ansys.e55
  394 root 44 0 2904K 122K sleep 0:01 0.00% os_mibs
   96 root 44 0 1648K 90K sleep 0:01 0.00% syslogd
  160 root 44 0 1648K 81K sleep 0:01 0.00% routed
  273 root 44 0 1600K 0K sleep 0:00 0.00% nfsd
  275 root 44 0 1592K 0K sleep 0:00 0.00% nfsiod

"top" says 539M used but only 315M free of 1002M. (This doesn't add up)

Is the UBC using up the rest? Now, here's the real interesting part.
I just ran a small C program that malloc'ed 1024 MB of memory and exits
normally. I ran top immediately after running the C program and top shows:

95 processes: 3 running, 17 sleeping, 74 idle, 1 zombie

Memory: Real: 106M/1002M act/tot Virtual: 102M/2804M use/tot Free: 787M


  PID USERNAME PRI NICE SIZE RES STATE TIME CPU COMMAND
17633 jmv515 42 0 335M 25M run 24:37 100.10% ansys.e55
12219 jmv515 42 0 335M 36M run 41:42 99.80% ansys55t
31432 lou 42 0 3032K 278K run 0:00 0.00% top
    1 root 42 0 440K 40K sleep 2:48 0.00% init
  854 root 39 -5 2568K 368K sleep 1:28 0.00% ansys_lm
  269 root 44 0 1680K 131K sleep 0:38 0.00% portmap
17273 root 44 0 1776K 90K sleep 0:09 0.00% mountd
  509 root 42 0 2056K 163K sleep 0:08 0.00% lmgrd
  507 root 42 0 2032K 163K sleep 0:08 0.00% lmgrd
  511 root 42 0 2048K 155K sleep 0:08 0.00% lmgrd
25322 lou 42 0 4488K 360K sleep 0:07 0.00% xterm
  517 root 42 0 2120K 163K sleep 0:05 0.00% cvlm
  519 root 42 0 2112K 163K sleep 0:05 0.00% cadds
30487 lou 44 0 2464K 253K sleep 0:02 0.00% csh
12917 jmv515 44 0 335M 385K sleep 0:01 0.00% ansys.e55

All of a suddent I have 787M available and the same ANSYS processes are
still running? What's up with that?

Also, the "ps" for the kernel displays:

USER PID NI PRI VSZ RSS S %CPU %MEM STARTED TIME COMMAND
root 0 -12 32 647M 38M R < 0.0 3.7 Mar 04 01:08:08 [kernel
idle]

Any clarifications of this whole memory issue would be greatly appreciated.


Thanks,

Lou

-----------------------------------------------------------------
Lou Muccioli
Systems Engineering Group
ANSYS, Inc. (http://www.ansys.com)
M/S: 2N1
Office: 2N092
Southpointe
275 Technology Drive
Canonsburg, PA 15317
Phone: 724-514-2888
FAX: 724-514-5096
E-Mail: lou.muccioli_at_ansys.com
-----------------------------------------------------------------
Received on Fri Mar 13 1998 - 01:59:50 NZDT

This archive was generated by hypermail 2.4.0 : Wed Nov 08 2023 - 11:53:37 NZDT