Mem: 1036588k total, 1008564k used, 28024k free, 68k buffers Swap: 1000432k total, 40k used, 1000392k free, 704392k cached PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND 26666 root 17 0 261m 69m 69m R 2.2 6.9 0:00.07 crm114 26662 root 18 0 213m 38m 37m R 1.2 3.8 0:00.04 crm114 25807 bulwark 14 -2 27624 23m 1728 S 44.9 2.3 0:42.08 spamd 26048 bulwark 15 -2 27136 23m 1724 S 21.5 2.3 0:24.76 spamd 26155 bulwark 14 -2 27096 23m 1724 S 0.0 2.3 0:11.69 spamd 26333 bulwark 14 -2 26444 22m 1716 S 0.0 2.2 0:01.34 spamd 14541 root 13 -2 25312 22m 2232 S 0.6 2.2 4:56.24 spamd 31322 per 15 0 34264 21m 15m R 1.2 2.2 19:29.32 konsole 26625 bulwark 15 -2 25312 21m 992 S 0.0 2.1 0:00.00 spamd 10186 root 15 0 56944 12m 1092 S 4.7 1.2 48:13.30 ibwd 31353 per 16 0 27672 10m 8652 S 0.0 1.0 0:04.06 kded 31351 per 15 0 23768 7540 6420 S 0.0 0.7 0:00.03 klauncher 31346 per 16 0 23916 6624 5440 S 0.0 0.6 0:00.03 kdeinit 31349 per 15 0 23264 6204 5112 S 0.0 0.6 0:00.01 dcopserver I understand virtual memory management perfectly well, although possibly not quite so well on Linux. In the above, 40K of swap-space is used, yet it appears that e.g. the two crm114 trasks, the ibwd and kded tasks have large amounts swapped out. I know top isn't exactly optimal for performance monitoring, but this seems completely off the mark. Any explanations? (this is kernel 2.6, SUSE Linux 10.1RC3). /Per Jessen, Zürich