nepim - network pipemeter

• INTRODUCTION
• LICENSE
• PLATFORMS
• USAGE
• EXAMPLE
• HINTS
• DOWNLOAD
• CVS
• SUPPORT
• RELATED

Keywords: network bandwidth throughput measurement estimation testing performance stressing tool

INTRODUCTION

nepim stands for network pipemeter, a tool for measuring available bandwidth between hosts. nepim is also useful to generate network traffic for testing purposes.

nepim operates in client/server mode, is able to handle multiple parallel traffic streams, reports periodic partial statistics along the testing, and supports IPv6.

LICENSE

nepim - network pipemeter
Copyright (C) 2005 Everton da Silva Marques

nepim is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2, or (at your option) any later version.

nepim is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.

You should have received a copy of the GNU General Public License along with nepim; see the file COPYING. If not, write to the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.

PLATFORMS

nepim has been tested under Linux, Solaris and FreeBSD, though it should work under other platforms as well. If you manage to build nepim for different systems, please send the patch.

USAGE

Starting the server:

server$ nepim

Starting the client against the server located at 10.10.10.10:

client$ nepim -c 10.10.10.10 -d

Display brief help about command line options:

$ nepim -h

EXAMPLE

Running nepim in server mode at the server host:

server$ nepim
nepim - network pipemeter - version 0.27
server: tcp_read=32768 tcp_write=32768 udp_read=4096 udp_write=4096
3: TCP socket listening on ::,1234
sock.c: nepim_create_socket: bind(4,0.0.0.0,1234): errno=98: Address already in use
server.c spawn_tcp_listener: TCP listener socket failed for 0.0.0.0,1234: -3
4: UDP socket listening on ::,1234
4: pmtud_mode=1 path_mtu=-11 mss=-19 tos=0 ttl=64 mcast_ttl=1 win_recv=109568 win_send=109
_ka=0 nodelay=-10
5: UDP socket listening on 0.0.0.0,1234
5: pmtud_mode=1 path_mtu=-11 mss=-19 tos=0 ttl=64 mcast_ttl=1 win_recv=109568 win_send=109
_ka=0 nodelay=-10
nepim: server ready

Running nepim in client mode at the client host:

client$ nepim -c localhost -d -r 100000
nepim - network pipemeter - version 0.27
client: tcp_read=32768 tcp_write=32768
not a UNIX domain path: localhost: errno=2: No such file or directory
TCP socket solving localhost,1234
TCP socket trying 127.0.0.1,1234
DEBUG FIXME sock.c nepim_connect_client_socket slow synchronous connect
3: TCP socket connected to 127.0.0.1,1234
3: sending: hello server_send=1 bit_rate=100000 pkt_rate=-1 stat_interval=2 test_duration=
_delay=250000 server_ka_send=0 server_ka_req=0 seed=0xd4d2e835 verify_data=0 random_fill=1
te=0x00 pause_duration=0 sock_ka=1 nagle=-1 overhead=0 password=
3: greetings sent to 127.0.0.1,1234
3: pmtud_mode=1 path_mtu=16436 mss=16383 tos=0 ttl=64 mcast_ttl=1 win_recv=87856 win_send=
ck_ka=1 nodelay=0
                  kbps_in   kbps_out    rcv/s    snd/s
  3  prt     8     100.00     100.00     4.00     4.00
  3  prt     6     100.00     100.00     4.00     4.00
  3  prt     4     100.00     100.00     4.00     4.00
  3  prt     2     100.00     100.00     4.00     4.00
  3  avg     0     100.00     100.00     4.00     4.00
3: pmtud_mode=1 path_mtu=16436 mss=16384 tos=0 ttl=64 mcast_ttl=1 win_recv=87856 win_send=
ck_ka=1 nodelay=0
nepim: no event sink registered
nepim: done

HINTS

• nepim is useful to assess the throughput at the transport layer (TCP or UDP) as seen by applications.
• nepim runs single-threaded and should impose very light burden on your CPU. Unless, of course, your testing hosts have relatively high network bandwidth compared to low CPU power.
• The client/server interaction is friendly to clients hosted behind Dynamic NAT or stateful firewalls: to build a connection, the client needs to reach only one transport-layer port on the server; test options are negotiated in the beginning of that single data connection; there is not any control connection.
• Test duration defaults to 10 seconds. The "-a" client option can supply a distinct duration. Releases higher than 0.31 recognize some time suffixes: m=minute, h=hour, d=day. For instance:

  Up to 0.31	Higher than 0.31
  nepim -c 10.0.0.1 -a 120	nepim -c 10.0.0.1 -a 2m
  nepim -c 10.0.0.1 -a 3600	nepim -c 10.0.0.1 -a 1h
  nepim -c 10.0.0.1 -a 43200	nepim -c 10.0.0.1 -a .5d

• One single server can service multiple clients simultaneously.
• As of nepim 0.11, one single client can interact simultaneously with multiple servers. For instance, suppose you want to test, from a single client, two remote servers, one located at 10.0.0.1,2000 and another at 192.168.0.1,3000:

  client$ nepim -c 10.0.0.1,2000 -c 192.168.0.1,3000

• The server listens to both TCP and UDP sockets. The client by default uses TCP sockets. Use the "-u" client option to switch the client operation to UDP.
• By default, only the server sends traffic towards the client. Use the "-s" client option to reverse the behavior, then only the client will send traffic. Use the "-d" client option to make both client and server to send traffic.
• Starting from nepim 0.32, options -r,-R,-e accept multiplicative suffixes and floating point rates. Examples:

  -r .25g = 250,000,000 bps
  -r .1m  =     100,000 pps
  -R .02k =          20 pps
  

• Use the "-r" client option to establish an upper bit rate limit. Without a rate limiting option, nepim sends as fast as possible. See also "-R" below. Please notice the rate is specified in bps (bits per second); for instance, the following example states a rate limit of 100,000 bps (100 Kbps):

  client$ nepim -c 10.0.0.1 -r 100000

• Use the "-R" client option to establish an upper "packet" rate limit (outbound rate limit for transport layer segments). Without a rate limiting option, nepim sends as fast as possible. If both "-r" and "-R" are given, nepim limits the sending rate at the lower of those bounds.
• The options "-r" and "-R" try to send some packets whenever a constant period is reached. The default period can be changed with the "-D" client option, in microseconds. "-D" is meaningful only when specified before "-r" or "-R". Example:

  client$ nepim -c 10.0.0.1 -D 100000 -R 20

• The "-e send-rate" client option tries to generate a constant data rate with segments sent at regular intervals. It should not be combined with "-D", "-r" or "-R", and might be processor-intensive. Example for 10 segments per second:

  client$ nepim -c 10.0.0.1 -e 10

• Use the "-n" client option to run multiple parallel traffic streams.
• Use the "-b" server option to make the server to listen on specific local addresses.
• Use the "-6" option to disable IPv6 support.
• Use the "-4" option to disable IPv4 support.
• Multicast support is special. Use the "-j" server switch to join a multicast group, then specify the "-M" client switch to enable multicast-compatible options. Example:

  server$ nepim -j ff01::1111

  
  

  client$ nepim -M -c ff01::1111 -r 100000

• The "-k" switch provides a simple password mechanism for client authentication. Just use the same password at both sides:

  server$ nepim -k 321

  
  

  client$ nepim -k 321 -c server-hostname

• One can specify the "-F" switch to force sending out UDP packets regardless of a remote server. Example for sending UDP packets towards 1.1.1.1:

  client$ nepim -F -c 1.1.1.1 -r 100000

• The "-U" option makes the server to listen on UNIX-domain stream sockets. See the example below.

  server$ nepim -U /tmp/sock1

  
  

  client$ nepim -c /tmp/sock1

• By default, nepim uses only transport payload data to compute incoming/outgoing rates. Such behavior is reasonable for large packets (as in the standard 1500 bytes MTU for Ethernet interfaces) since the encapsulation headers remain comparatively small. However, as packets become smaller, the effect of encapsulation on transfer rates grows. The "-o" client option can be used to circumvent gross rate computation errors caused by excessive per-packet encapsulation overhead. The example below shows how the "-o" client option should be used to specify a per-packet overhead of 28 bytes, for UDP payloads of 100 bytes.

  server$ nepim -W 100

  
  

  client$ nepim -u -W 100 -d -r 100000 -c 10.10.10.10 -o 28

  Notes:
  1. The "-o" switch remains untested for TCP segments.
  2. It is probably incorrect to specify "-o" with the default large UDP write size, which typically causes local fragmentation. Thus, in order to use the "-o" option, it is advisable to choose a payload size which would surely fit the UDP segment entirely into a single MTU. For instance, with a 1500-byte MTU, 1400 would likely suffice:

     server$ nepim -W 1400

     
     

     client$ nepim -u -W 1400 -d -r 100000 -c 10.10.10.10 -o

DOWNLOAD

Get latest nepim stable release here:

http://savannah.nongnu.org/download/nepim/

CVS

ViewCVS access to CVS tree:

http://cvs.savannah.nongnu.org/viewcvs/nepim/nepim/

Anonymous access to CVS repository:

export CVS_RSH=ssh

cvs -z3 -d:pserver:anonymous@cvs.savannah.nongnu.org:/sources/nepim co nepim

SUPPORT

For suggestions, patches, bug reports, please visit nepim support page at Savannah:

http://savannah.nongnu.org/projects/nepim/

RELATED

Other tools similar to nepim:

• Iperf - The TCP/UDP Bandwidth Measurement Tool
• NetPerf