Re: Slow System

["James E. Tarvid" <tarvid ls net>]

Display from "top" on a fairly well used Linux machine. If "idle" were to
drop below 70% I would upgrade the CPU. Maybe I should install more memory;
memory is so cheap that even 4MB of swap usage is unnecessarily high.
 
 7:25am  up 7 days, 19:34,  6 users,  load average: 0.13, 0.07, 0.01
70 processes: 68 sleeping, 2 running, 0 zombie, 0 stopped
CPU states:  0.9% user,  1.7% system,  2.3% nice, 97.4% idle
Mem:   30868K av,  25648K used,   5220K free,  18040K shrd,   3576K buff
Swap:  32220K av,   4324K used,  27896K free                  6596K cached

  PID USER     PRI  NI  SIZE  RSS SHARE STAT  LIB %CPU %MEM   TIME COMMAND
16011 tarvid     3   0   544  544   404 R       0  2.3  1.7   0:02 top
28967 root       0   0  3176 3128   464 S       0  0.3 10.1   1:28 named
    1 root       0   0   196  168   124 S       0  0.0  0.5  10:38 init
    2 root       0   0     0    0     0 SW      0  0.0  0.0   0:01 kflushd
    3 root     -12 -12     0    0     0 SW<     0  0.0  0.0   0:00 kswapd
    4 root       0   0     0    0     0 SW      0  0.0  0.0   0:00 nfsiod
    5 root       0   0     0    0     0 SW      0  0.0  0.0   0:00 nfsiod
    6 root       0   0     0    0     0 SW      0  0.0  0.0   0:00 nfsiod
    7 root       0   0     0    0     0 SW      0  0.0  0.0   0:00 nfsiod
15942 root       0   0   484  484   380 S       0  0.0  1.5   0:00 mgetty
  275 root       0   0   480  160   160 S       0  0.0  0.5   0:00 login
   21 root       0   0   116   68    52 S       0  0.0  0.2   0:00 kerneld
  299 root       0   0   348  164   164 S       0  0.0  0.5   0:00 bash
  130 root       0   0   244  192   148 S       0  0.0  0.6  30:21 syslogd
  139 root       0   0   336  148   104 S       0  0.0  0.4   0:00 klogd
  150 root       0   0   196  148    96 S       0  0.0  0.4   0:00 crond
  161 bin        0   0   228  160   160 S       0  0.0  0.5   0:00 portmap
  172 root       0   0   140   80    52 S       0  0.0  0.2   0:18 inetd   
   

----------
> From: Paul Braman <braman tsc-corp com>
> To: RedHat Install List <redhat-install-list redhat com>
> Subject: Re: Slow System
> Date: Thursday, April 30, 1998 6:32 AM
> 
> 
> > > Something is fundamentally wrong.
> > > 
> > > If the CPU usage indicator runs full or even more than 30% it is time
to
> > > look at processor usage.
> > > 
> > > Take a look at "top". (e.g. man top)
> > > 
> > > Jim Tarvid
> 
> This is not true.  Something is wrong if you're CPU usage does *not* run 
> near full capacity most of the time.  (Assuming you're actually *doing* 
> something during that time.)
> 
Capacity is something you want in reserve, not use. 

> Linux, like most other Unices, will take full advantage of the system on 
> which it is running.  That means that regardless of what process it 
> happens to be running it wants to get it over with as quick as possible. 

> The only way to do that is to shove as much stuff through the processor 
> as you can.
> 
True as long as the "stuff" is not idle polling loops.

> Now, if your system is *not* running near full capacity there are some 
> things to look at.  First, make sure you're not running out of RAM/swap. 

> You'll most likely be using a lot of RAM, but if you're swapping out 
> processes at the same time you've got a problem.  Second, consider which 
> processes are running.  If the current process is an IO-intensive process

> and you're running slowly then maybe it's your IO subsystem.  Perhaps 
> your disk drive can't shove enough data through to keep the process 
> running efficiently.
> 
True also. If you are running IO intensive tasks a significant portion of
the time, attention to the IO subsystem is in order. I process 1GB data
files on ocassion but I don't do them on my shared server. If I were to do
them even 4 hours a day, I would invest in multiple 10K rpm disk drives and
maybe I could reduce them to an hour a day.

> Paul Braman
> braman tsc-corp com
> 
Jim Tarvid