Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 26 Oct 2000 15:29:17 +1100 (EST)
From:      Stanley Hopcroft <Stanley.Hopcroft@IPAustralia.Gov.AU>
To:        FreeBSD-ISP@FreeBSD.ORG
Cc:        "Damien.Lederer" <Damien.Lederer@IPAustralia.Gov.AU>, Carl Makin <Carl.Makin@IPAustralia.Gov.AU>
Subject:   Off n(TOP)ic question: ntop-1.1 chews heaps of CPU on 4.x-R
Message-ID:  <Pine.BSF.4.21.0010261511490.21749-100000@stan.aipo.gov.au>

next in thread | raw e-mail | index | archive | help
Dear Ladies and Gentlemen,

I am writing to say that the marvellous ntop-1.1 packet accounting tool

(all the good RMON things like

- top talkers
- top protocols
- traffic matrix in software with a Web or curses interface)

uses a lot of CPU.

In fact it seems to use *all* of the CPU it can get.

This is no big deal but my Solaris admin colleague pointed out that the
same application on a Sparc E4 uses barely any CPU even when exposed to
500 LDAP requests/second.

Here is ntop running on FreeBSD 4.1-RELEASE (P5/166/80 MB RAM) that is
only seeing *boradcast* and data from or to itself. It is a box whose
CPU idle time is usually > 80%.

last pid: 98015;  load averages:  1.27,  1.31,  0.93
up 19+03:59:44  15:18:35
42 processes:  2 running, 40 sleeping
CPU states: 50.6% user,  0.0% nice, 49.0% system,  0.4% interrupt,
0.0% idle
Mem: 39M Active, 44M Inact, 23M Wired, 6712K Cache, 22M Buf, 11M Free
Swap: 256M Total, 256M Free

  PID USERNAME PRI NICE  SIZE    RES STATE    TIME   WCPU    CPU
COMMAND
96811 root      56   0 10928K  9784K RUN      6:08 95.36% 95.36% ntop
98015 anwsmh    29   0  1872K  1088K RUN      0:00  4.13%  1.22% top
  351 netsaint  10   0  1388K   960K nanslp 443:07  0.00%  0.00%
netsaint
   84 root       2 -12  1240K   692K select   3:38  0.00%  0.00% ntpd
32380 root       2   0  2852K  2308K select   1:49  0.00%  0.00% httpd
   82 bind       2   0  3136K  2568K select   1:37  0.00%  0.00% named
  126 root       2   0  1400K   892K select   1:12  0.00%  0.00%
sendmail


and on 4.1.1-RELEASE (PIII/733/128MB RAM) that is getting all the
traffic for a small segment (<= 6 servers).

last pid: 17389;  load averages:  1.03,  1.08,  1.05
up 7+02:28:32  15:19:20
35 processes:  2 running, 33 sleeping
CPU states: 42.4% user,  0.0% nice, 57.6% system,  0.0% interrupt,
0.0% idle
Mem: 91M Active, 7720K Inact, 20M Wired, 5324K Cache, 22M Buf, 492K
Free
Swap: 256M Total, 15M Used, 241M Free, 5% Inuse

  PID USERNAME PRI NICE  SIZE    RES STATE    TIME   WCPU    CPU
COMMAND
17303 root      63   0 19344K 12760K RUN     57:29 91.94% 91.94% ntop


I am not sure this a problem because ntop is such a superb and useful
application, but it seems strange. Both installations are standard out
of the box; in particular both run the bpf and libpcap shipped with
that version of FreeBSD.

One of the reasons for asking about this is that "TCPIP Illustrated"
claims that the Sun NIT interface is very poor compared to the BSD
bpf. Could it be they have radically improved DLPI ?


Thank you.

Yours sincerely,


S Hopcroft

Network Specialist
IP Australia

+61 2 6283 3189
+61 2 6281 1353 FAX



To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-isp" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.BSF.4.21.0010261511490.21749-100000>