iftop - display bandwidth usage on an interface by host
| [ -nNpblBP
] [-i interface
[-f filter code
] [-F net
] [ -G
listens to network traffic on a named interface
, or on the
first interface it can find which looks like an external interface if none is
specified, and displays a table of current bandwidth usage by pairs of hosts.
must be run with sufficient permissions to monitor all network
traffic on the interface
; see pcap(3)
for more information, but
on most systems this means that it must be run as root.
By default, iftop
will look up the hostnames associated with addresses it
finds in packets. This can cause substantial traffic of itself, and may result
in a confusing display. You may wish to suppress display of DNS traffic by
using filter code such as not port domain
, or switch it off entirely,
by using the -n
option or by pressing r
when the program is
By default, iftop
counts all IP packets that pass through the filter, and
the direction of the packet is determined according to the direction the
packet is moving across the interface. Using the -F
option it is
possible to get iftop
to show packets entering and leaving a given
network. For example, iftop -F 10.0.0.0/255.0.0.0
will analyse packets
flowing in and out of the 10.* network.
Some other filter ideas:
- not ether host ff:ff:ff:ff:ff:ff
- Ignore ethernet broadcast packets.
- port http and not host webcache.example.com
- Count web traffic only, unless it is being directed through a local web
- How much bandwidth are users wasting trying to figure out why the network
- Print a summary of usage.
- Don't do hostname lookups.
- Do not resolve port number to service names
- Run in promiscuous mode, so that traffic which does not pass directly
through the specified interface is also counted.
- Turn on port display.
- Display and count datagrams addressed to or from link-local IPv6
addresses. The default is not to display that address category.
- Don't display bar graphs of traffic.
- -m limit
- Set the upper limit for the bandwidth scale. Specified as a number with a
'K', 'M' or 'G' suffix.
- Display bandwidth rates in bytes/sec rather than bits/sec.
- -i interface
- Listen to packets on interface.
- -f filter code
- Use filter code to select the packets to count. Only IP packets are
ever counted, so the specified code is evaluated as (filter
code ) and ip.
- -F net/mask
- Specifies an IPv4 network for traffic analysis. If specified, iftop will
only include packets flowing in to or out of the given network, and packet
direction is determined relative to the network boundary, rather than to
the interface. You may specify mask as a dotted quad, such as
/255.255.255.0, or as a single number specifying the number of bits set in
the netmask, such as /24.
- -G net6/mask6
- Specifies an IPv6 network for traffic analysis. The value of mask6
can be given as a prefix length or as a numerical address string for more
- -c config file
- Specifies an alternate config file. If not specified, iftop will use
~/.iftoprc if it exists. See below for a description of config
- -t text output mode
- Use text interface without ncurses and print the output to STDOUT.
When running, iftop
uses the whole screen to display network usage. At
the top of the display is a logarithmic scale for the bar graph which gives a
visual indication of traffic.
The main part of the display lists, for each pair of hosts, the rate at which
data has been sent and received over the preceding 2, 10 and 40 second
intervals. The direction of data flow is indicated by arrows, <= and =>.
foo.example.com => bar.example.com 1Kb 500b 100b
<= 2Mb 2Mb 2Mb
shows, on the first line, traffic from foo.example.com
; in the preceding 2 seconds, this averaged 1Kbit/s,
around half that amount over the preceding 10s, and a fifth of that over the
whole of the last 40s. During each of those intervals, the data sent in the
other direction was about 2Mbit/s. On the actual display, part of each line is
inverted to give a visual indication of the 10s average of traffic. You might
expect to see something like this where host foo
is making repeated
HTTP requests to bar
, which is sending data back which saturates a
By default, the pairs of hosts responsible for the most traffic (10 second
average) are displayed at the top of the list.
At the bottom of the display, various totals are shown, including peak traffic
over the last 40s, total traffic transferred (after filtering), and total
transfer rates averaged over 2s, 10s and 40s.
SOURCE / DEST AGGREGATION¶
By pressing s
is running, all traffic for
each source or destination will be aggregated together. This is most useful
is run in promiscuous mode, or is run on a gateway machine.
toggle the display of source and destination ports
will toggle port display on/off.
cycles through the four line display modes; the default 2-line display,
with sent and received traffic on separate lines, and 3 1-line displays, with
sent, received, or total traffic shown.
By default, the display is ordered according to the 10s average (2nd column). By
it is possible to sort by the 1st, 2nd
or 3rd column. By pressing <
the display will be
sorted by source or destination hostname respectively.
allows you to enter a POSIX extended regular expression that will be
used to filter hostnames shown in the display. This is a good way to quickly
limit what is shown on the display. Note that this happens at a much later
stage than filter code, and does not affect what is actually captured. Display
filters DO NOT affect the totals at the bottom of the screen.
PAUSE DISPLAY / FREEZE ORDER¶
will pause the current display.
will freeze the current screen order. This has the side effect that
traffic between hosts not shown on the screen at the time will not be shown at
all, although it will be included in the totals at the bottom of the screen.
will scroll the display of hosts. This feature is most
useful when the display order is frozen (see above).
allows you to edit the filter code whilst iftop running. This can lead
to some unexpected behaviour.
iftop can read its configuration from a config file. If the -c
not specified, iftop will attempt to read its configuration from
, if it exists. Any command line options specified will
override settings in the config file.
The config file consists of one configuration directive per line. Each directive
is a name value pair, for example:
sets the network interface. The following config directives are supported:
- interface: if
- Sets the network interface to if.
- dns-resolution: (yes|no)
- Controls reverse lookup of IP addresses.
- port-resolution: (yes|no)
- Controls conversion of port numbers to service names.
- filter-code: bpf
- Sets the filter code to bpf.
- show-bars: (yes|no)
- Controls display of bar graphs.
- promiscuous: (yes|no)
- Puts the interface into promiscuous mode.
- port-display: (off|source-only|destination-only|on)
- Controls display of port numbers.
- link-local: (yes|no)
- Determines displaying of link-local IPv6 addresses.
- hide-source: (yes|no)
- Hides source host names.
- hide-destination: (yes|no)
- Hides destination host names.
- use-bytes: (yes|no)
- Use bytes for bandwidth display, rather than bits.
- sort: (2s|10s|40s|source|destination)
- Sets which column is used to sort the display.
- Controls the appearance of each item in the display.
- show-totals: (yes|no)
- Shows cumulative total for each item.
- log-scale: (yes|no)
- Use a logarithmic scale for bar graphs.
- max-bandwidth: bw
- Fixes the maximum for the bar graph scale to bw, e.g.
"10M". Note that the value has to always be in bits, regardless
if the option to display in bytes has been chosen.
- net-filter: net/mask
- Defines an IP network boundary for determining packet direction.
- net-filter6: net6/mask6
- Defines an IPv6 network boundary for determining packet direction.
- screen-filter: regexp
- Sets a regular expression to filter screen output.
QUIRKS (aka they're features, not bugs)¶
There are some circumstances in which iftop may not do what you expect. In most
cases what it is doing is logical, and we believe it is correct behaviour,
although I'm happy to hear reasoned arguments for alternative behaviour.
Totals don't add up
There are several reasons why the totals may not appear to add up. The most
obvious is having a screen filter in effect, or screen ordering frozen. In
this case some captured information is not being shown to you, but is included
in the totals.
A more subtle explanation comes about when running in promiscuous mode without
specifying a -F
option. In this case there is no easy way to assign the
direction of traffic between two third parties. For the purposes of the main
display this is done in an arbitrary fashion (by ordering of IP addresses),
but for the sake of totals all traffic between other hosts is accounted as
incoming, because that's what it is from the point of view of your interface.
option allows you to specify an arbitrary network boundary, and
to show traffic flowing across it.
Peak totals don't add up
Again, this is a feature. The peak sent and peak received didn't necessarily
happen at the same time. The peak total is the maximum of sent plus received
in each captured time division.
Changing the filter code doesn't seem to work
Give it time. Changing the filter code affects what is captured from the time
that you entered it, but most of what is on the display is based on some
fraction of the last 40s window of capturing. After changing the filter there
may be entries on the display that are disallowed by the current filter for up
to 40s. DISPLAY FILTERING has immediate effect and does not affect what is
- Configuration file for iftop.
Paul Warren <email@example.com>
$Id: iftop.8,v 1.31 2014/01/05 17:22:39 pdw Exp $
This program 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 of the License, or (at your option) any later
This program 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
this program; if not, write to the Free Software Foundation, Inc., 51 Franklin
Street, Fifth Floor, Boston, MA 02110-1301 USA.