ip cache ager is the process which ages the cache. That is to
say, every once in a bit, a process starts up that goes through
1/Nth of the cache, and checks to see if it has not been used, and
prunes it from the tree.
Hence the spikiness; it's particularly noticeable to interactive
work as ip cache ager is (rightfully so) given high priority.
This is normal on a busy router with lots of destinations (esp. at
borders et. al.)
To solve this, you can consider tweaking the parameters on cache
ager (do more less often, fewer spikes, but more significant) or
you can move to a non-cached based routing, like CEF/FIB. The
former is dangerous, the latter significant.
-a
(who is writing this after a 8 hour evil airline experience and
may not be entirely cogent)
Thus spake Alex Bligh (amb@gxn.net)
on or about Sun, Feb 15, 1998 at 02:25:45AM +0000:
>
> I have a busy-ish router or three, one of which seems to be mostly
> busy with a process called "IP Cache Ager" (sometimes peaking at 70%
> of CPU). I'm running NetFlow on it, & just for comparison (with about 60
> or 70 BGP peers and 30-40Mb/s of data running through it):
>
> mae-east1#show proc cpu
> CPU utilization for five seconds: 70%/11%; one minute: 24%; five minutes: 23%
> PID Runtime(ms) Invoked uSecs 5Sec 1Min 5Min TTY Process
> 32 92979364 414543 224301 55.16% 7.13% 6.75% 0 IP Cache Ager
> 26 25452492 24575081 1035 2.20% 2.33% 2.30% 0 IP Input
> 52 14551796 12204107 1192 0.31% 0.38% 0.31% 0 BGP Router
> 54 12516004 192218 65114 0.00% 0.98% 0.82% 0 BGP Scanner
>
> 7507 running rsp-pv-mz.111-12.CA1.bin
>
> It's still a reasonably happy bunny, but I worry about putting too
> much more data through it if these figures are accurate.
>
> Any ideas?
>
> --
> Alex Bligh
> GX Networks (formerly Xara Networks)
>
This archive was generated by hypermail 2b29 : Sun Aug 04 2002 - 04:13:15 EDT