TruGuru's,
I have two systems running Oracle8 and Oracle8i (under Oracle Financials
11i), on a DS20 (2x6/500, 2G RAM) and ES40 (2x5/600, 2G RAM) respectively.
I think I need more RAM, but with Oracle's reputation for eating any
available memory and my lack of understanding of vmstat and vmubc, I'm not
100% convinced.
The systems usually maintain a "load average" below 1, but frequently, in a
bursty fashion, spike up to 3-5 for short periods of time.  Of course Oracle
is always the dominant process.  vmstat indicates to me that there is very
little (30K pages) memory, but there is almost no paging out to swap which
is under 10% in-use.  vmubc shows that half of the "active" memory is
inactive.
My thoughts are that if I add a bunch of memory either the systems wills
scream as I have never seen or Oracle will just gobble up any additional
memory and the systems will continue to run as they currently are.
Below is a snippet of a "vmstat 5" command on each system:
DS20
procs      memory        pages                            intr       cpu
r   w   u  act free wire fault  cow zero react  pin pout  in  sy  cs us sy
id
3 316  35 204K  30K  20K   64M   5M  32M    2M  18M  34K  56 829 439  6  1
93
3 316  35 204K  30K  20K     2   16   17     0   28    0  13 484 480  1  0
99
3 316  35 204K  30K  20K     5    0    1     0    1    0  19 170 221  0  0
100
3 328  35 204K  30K  20K   258    0   51     0   12    0  15 397 231  0  1
99
3 328  35 204K  30K  20K     0    0    0     0    0    0  11 212 221  0  0
100
3 328  35 204K  30K  20K     0    0    0     0    0    0   9 271 178  1  0
99
3 324  35 204K  30K  20K    18    1    0     0    1    0  18 159 214  0  0
100
3 320  35 204K  30K  20K    48    1   10     0    2    0  26 184 250  0  0
100
3 319  35 204K  30K  20K     0    0    0     0    0    0   9 117 152  0  0
100
3 316  34 204K  30K  20K    68    1    0     0    1    0  15 190 222  0  0
100
3 322  34 207K  27K  20K  3629  173 3002     0  471    0 152  1K 726  7  3
90
3 320  34 207K  27K  20K    23    0    1     0    7    0 186  1K 621  1  1
98
3 319  34 206K  27K  20K    24    0    1     0    0    0  71 598 320  1  0
99
ES40
procs    memory         pages                          intr        cpu
r  w  u  act  free wire fault cow zero react pin pout  in  sy  cs  us  sy
id
3378 26  209K  12K  34K  101    0  101    0    0    0   8  1K 273   1   1
99
3378 26  209K  12K  34K  101    0  101    0    0    0   9 755 322   1   1
99
3378 26  209K  12K  34K  101    0  101    0    0    0   3 524 276   0   1
99
3378 26  209K  12K  34K  101    0  101    0    0    0   9 715 288   0   1
99
3378 26  209K  12K  34K  101    0  101    0    0    0  12  1K 306   1   1
98
3377 26  209K  12K  34K  101    0  101    0    0    0   9 238 248   0   1
99
3377 26  209K  12K  34K  101    0  101    0    0    0   8 260 241   0   1
99
3377 26  209K  12K  34K  101    0  101    0    0    0   9 259 247   0   1
99
3378 26  209K  12K  34K  101    0  101    0    0    0  11  1K 288   1   1
99
3378 26  209K  12K  34K  101    0  101    0    0    0   9 552 282   0   1
99
3378 26  209K  12K  34K  101    0  101    0    0    0  11 679 311   1   1
99
3379 26  209K  12K  34K  101    0  101    0    0    0  10  1K 327   1   1
98
3379 26  209K  12K  34K  101    0  101    0    0    0   9  1K 282   0   1
99
Note that the systems are idle at this time of day.
What are your thoughts?
Douglas L. Thomas
dthomas_at_glgt.com
Received on Thu Nov 01 2001 - 23:25:27 NZDT