Negative caching

[Hal Burgiss]

On Fri, Jun 19, 2009 at 12:20:00AM -0600, Karl F. Larsen wrote:
> 	Slow can be a shortage of RAM. And I think you have lots but it might
> not all may be working. Use a Terminal with top to measure your RAM and
> swap usage. Let us know what it shows when you you slow down.

I guess you missed the original post:

atop data:

PRC | sys   0.76s | user   1.70s | #proc    200 | #zombie    2 | #exit      ?
CPU | sys      7% | user     17% | irq       1% | idle      0% | wait     75%
CPL | avg1   2.90 | avg5    2.54 | avg15   2.23 | csw     9527 | intr    5228
MEM | tot  992.4M | free   91.3M | cache 316.9M | buff    3.2M | slab   44.1M
SWP | tot    2.9G | free    1.4G |              | vmcom   3.2G | vmlim   3.3G
PAG | scan  10364 | stall      0 |              | swin    4293 | swout    585 (red)
DSK |         sda | busy    100% | read    1625 | write    608 | avio    4 ms (red)

That was after a long idle period when everything was really slow, response to
mouse clicks, keyboard events, etc. Click on a background window, wait 5-6
seconds for it to start to come to the foreground. That kind of thing (and no
compiz!).

As you can see, there is plenty of free swap space. You can also see the
PAGING and DISK utilization stats are off the charts. There was no user
initiated disk intensive activity at the time. None. No downloads going on, no
copying, no unpacking, rsyncing, etc. The disk usage you see there was in
response to simple mouse clicks (like to raise a window) and/or keyboard
like typing in a text terminal.

If its not kernel swap, its something horribly askew in 9.04 (at least on my
system). The same system has run 7.04, 7.10, and 8.04 just fine. 8.10 was a
trainwreck (because intel chipset debacle). 9.04 is better, but it is a system
that requires rebooting occasionally to be functional (at least the desktop).
This is like one step forward, and 3 steps backwards. The notion of rebooting
to maintain system integrity just rubs me the wrong way. This is a microsoft
legacy.

To compare, current atop:

PRC | sys   0.66s | user   1.68s | #proc    180 | #zombie    0 | #exit      ? |
CPU | sys      7% | user     17% | irq       0% | idle     76% | wait      1% |
CPL | avg1   0.90 | avg5    0.45 | avg15   0.43 | csw    11319 | intr    2174 |
MEM | tot  992.4M | free   23.4M | cache 363.7M | buff    2.0M | slab   44.3M |
SWP | tot    2.9G | free    2.1G |              | vmcom   2.9G | vmlim   3.3G |
PAG | scan      0 | stall      0 |              | swin      11 | swout      0 |
DSK |         sda | busy      1% | read      12 | write      4 | avio    5 ms |

halb at honey:~$ uptime
 08:45:56 up 1 day, 14:29,  2 users,  load average: 0.70, 0.45, 0.42

Life is good. For now. 

-- 
Hal