Linux Performance Monitoring
Sarah07 said:
they told me to fix them up cause thay seems to be slower than what they are specially dial up server.
you might
start with this website
xinetd: controls services which are auto started
the control list is typically located at /etc/xinetd.conf
and depending upon which specific Linux you're running
may have other files in the suibdir /etc/xinitd.d/*
XINETD is the first place to start reducing workload.
you should know WHY each and every service is made active and
WHY you need it. If it's just 'it would be nice ...' make it
disabled! set ONLY the processes you need, disable the rest,
and send KILL -1 $pid-xinet.d-process to reconfigure the
running system w/o a reboot.
now the quesiton becomes, what's chewing up the resources?
System Load: use
uptime
the last three numbers are important. see man uptime(1)
load average should be no more that 3, better still 2 or lower
only two soultions here: add more memory OR reduce process/thread counts
Monitor Paging / Virtual Memory
a combination pipe:
watch -n 1 -d free
actual Pagefile info is seen
swapon -s
Linux Paging: man
vmstat(1) which gives you
DETAILS on swap in/out
swap-out occurs for hi memory usage, dirty pages and
the need to switch to another process. this will later
generate a corresponding swap-in
swap-in can occur for just simple page stealling
(ie, not dirty and memory needed elsewhere)
Processes of course: PS
you find HOW MAN and WHICH ONES are using the cpu but
but here's the winner!
TOP
man top(1) will give you a better snapshot including
uptime
processes
cpu usage
memory usage
swap space usage (need vmstat 1 for swap i/o)
Network monitoring? ouch! there's a ton of stuff
google 'linux monitor network performance'
might like to see also