[BACK]Return to invoking.rst CVS log [TXT] [DIR] Up to [ELWIX - Embedded LightWeight unIX -] / embedaddon / iperf / docs

Diff for /embedaddon/iperf/docs/invoking.rst between versions 1.1 and 1.1.1.2

version 1.1, 2016/10/18 13:28:18 version 1.1.1.2, 2021/03/17 00:36:45
Line 6  The manual page is the most up-to-date reference to th Line 6  The manual page is the most up-to-date reference to th
   
 For sample command line usage, see:   For sample command line usage, see: 
   
http://fasterdata.es.net/performance-testing/network-troubleshooting-tools/iperf-and-iperf3/https://fasterdata.es.net/performance-testing/network-troubleshooting-tools/iperf/
   
 Using the default options, iperf3 is meant to show typical well  Using the default options, iperf3 is meant to show typical well
 designed application performance.  "Typical well designed application"  designed application performance.  "Typical well designed application"
Line 20  iperf3 Manual Page Line 20  iperf3 Manual Page
 ------------------  ------------------
   
 This section contains a plaintext rendering of the iperf3 manual page.  This section contains a plaintext rendering of the iperf3 manual page.
It is presented here only for convenience; the authoritative iperf3It is presented here only for convenience; the text here might not
 correspond to the current version of iperf3.  The authoritative iperf3
 manual page is included in the source tree and installed along with  manual page is included in the source tree and installed along with
 the executable.  the executable.
   
 ::  ::
   
   IPERF(1)                         User Manuals                         IPERF(1)   IPERF3(1)                        User Manuals                        IPERF3(1)
   
   
   
    NAME     NAME
   iperf3 − perform network throughput tests          iperf3 - perform network throughput tests
   
    SYNOPSIS     SYNOPSIS
   iperf3 ‐s [ options ]          iperf3 -s [ options ]
   iperf3 ‐c server [ options ]          iperf3 -c server [ options ]
   
   
    DESCRIPTION     DESCRIPTION
   iperf3  is  a  tool for performing network throughput measurements.  It          iperf3  is  a  tool for performing network throughput measurements.  It
   can test either TCP or UDP throughput.  To perform an iperf3  test  the          can test TCP, UDP, or SCTP throughput.  To perform an iperf3  test  the
   user must establish both a server and a client.          user must establish both a server and a client.
   
          The  iperf3  executable  contains both client and server functionality.
           An iperf3 server can be started using either of the -s or --server com-
           mand-line parameters, for example:
    
                  iperf3 -s
    
                  iperf3 --server
    
           Note  that  many  iperf3  parameters  have  both  short  (-s)  and long
           (--server) forms.  In this section we will generally use the short form
           of  command-line  flags,  unless only the long form of a flag is avail-
           able.
    
           By default, the iperf3 server listens on TCP port 5201 for  connections
           from  an iperf3 client.  A custom port can be specified by using the -p
           flag, for example:
    
                  iperf3 -s -p 5002
    
           After the server is started, it will listen for connections from iperf3
           clients  (in  other words, the iperf3 program run in client mode).  The
           client mode can be started using the -c command-line option, which also
           requires a host to which iperf3 should connect.  The host can by speci-
           fied by hostname, IPv4 literal, or IPv6 literal:
    
                  iperf3 -c iperf3.example.com
    
                  iperf3 -c 192.0.2.1
    
                  iperf3 -c 2001:db8::1
    
           If the iperf3 server is running on a non-default TCP  port,  that  port
           number needs to be specified on the client as well:
    
                  iperf3 -c iperf3.example.com -p 5002
    
           The initial TCP connection is used to exchange test parameters, control
           the start and end of the test, and to exchange test results.   This  is
           sometimes  referred  to  as  the "control connection".  The actual test
           data is sent over a separate TCP connection, as a separate flow of  UDP
           packets, or as an independent SCTP connection, depending on what proto-
           col was specified by the client.
    
           Normally, the test data is sent from the client to the server, and mea-
           sures  the  upload  speed  of the client.  Measuring the download speed
           from the server can be done by specifying the -R flag  on  the  client.
           This causes data to be sent from the server to the client.
    
                  iperf3 -c iperf3.example.com -p 5202 -R
    
           Results  are displayed on both the client and server.  There will be at
           least one line of output per measurement interval (by  default  a  mea-
           surement  interval lasts for one second, but this can be changed by the
           -i option).  Each line of output includes (at least) the time since the
           start  of the test, amount of data transferred during the interval, and
           the average bitrate over that interval.  Note that the values for  each
           measurement  interval  are taken from the point of view of the endpoint
           process emitting that output (in other words, the output on the  client
           shows the measurement interval data for the client.
    
           At  the  end of the test is a set of statistics that shows (at least as
           much as possible) a summary of the test as seen by both the sender  and
           the  receiver,  with  lines tagged accordingly.  Recall that by default
           the client is the sender and the server is the  receiver,  although  as
           indicated above, use of the -R flag will reverse these roles.
    
           The  client  can be made to retrieve the server-side output for a given
           test by specifying the --get-server-output flag.
    
           Either the client or the server can produce its output in a JSON struc-
           ture,  useful for integration with other programs, by passing it the -J
           flag.  Because the contents of the JSON structure  are  only  competely
           known after the test has finished, no JSON output will be emitted until
           the end of the test.
    
           iperf3 has a (overly) large set of command-line  options  that  can  be
           used  to  set the parameters of a test.  They are given in the "GENERAL
           OPTIONS" section of the manual page below, as  well  as  summarized  in
           iperf3's help output, which can be viewed by running iperf3 with the -h
           flag.
    
    GENERAL OPTIONS     GENERAL OPTIONS
   ‐p, ‐‐port n          -p, --port n
   set server port to listen on/connect to to n (default 5201)                 set server port to listen on/connect to to n (default 5201)
   
   ‐f, ‐‐format          -f, --format
   [kmKM]   format to report: Kbits, Mbits, KBytes, MBytes                 [kmgtKMGT]   format to report: Kbits/Mbits/Gbits/Tbits
   
   ‐i, ‐‐interval n          -i, --interval n
   pause  n  seconds between periodic bandwidth reports; default is                 pause n seconds between periodic throughput reports; default  is
   1, use 0 to disable                 1, use 0 to disable
   
   ‐F, ‐‐file name          -F, --file name
   client‐side: read from  the  file  and  write  to  the  network,                 Use  a  file  as  the  source  (on  the  sender) or sink (on the
   instead of using random data; server‐side: read from the network                 receiver) of data, rather than just generating  random  data  or
   and write to the file, instead of throwing the data away                 throwing  it  away.  This feature is used for finding whether or
                 not the storage subsystem is the bottleneck for file  transfers.
   ‐A, ‐‐affinity n/n,m                 It  does not turn iperf3 into a file transfer tool.  The length,
   Set the CPU affinity, if possible (Linux and FreeBSD only).   On                 attributes, and in some cases contents of the received file  may
   both  the  client  and  server you can set the local affinity by                 not match those of the original file.
   using the n form of this argument (where n is a CPU number).  In   
   addition,  on  the  client  side  you  can override the server’s          -A, --affinity n/n,m
   affinity for just that one test, using the n,m form of argument.                 Set  the  CPU affinity, if possible (Linux, FreeBSD, and Windows
   Note  that when using this feature, a process will only be bound                 only).  On both the client and server  you  can  set  the  local
   to a single CPU (as opposed to a set containing potentialy  mul‐                 affinity  by using the n form of this argument (where n is a CPU
   tiple CPUs).                 number).  In addition, on the client side you can  override  the
                 server's  affinity for just that one test, using the n,m form of
   ‐B, ‐‐bind host                 argument.  Note that when using this  feature,  a  process  will
   bind to a specific interface                 only  be  bound  to a single CPU (as opposed to a set containing
                 potentialy multiple CPUs).
   ‐V, ‐‐verbose   
   give more detailed output          -B, --bind host
                 bind to the specific interface associated with address host.  If
   ‐J, ‐‐json                 the  host  has multiple interfaces, it will use the first inter-
   output in JSON format                 face by default.
   
   ‐‐logfile file          -V, --verbose
   send output to a log file.                 give more detailed output
   
   ‐d, ‐‐debug          -J, --json
   emit  debugging  output.  Primarily (perhaps exclusively) of use                 output in JSON format
   to developers.   
          --logfile file
   ‐v, ‐‐version                 send output to a log file.
   show version information and quit   
          --forceflush
   ‐h, ‐‐help                 force flushing output at every interval.  Used to avoid  buffer-
   show a help synopsis                 ing when sending output to pipe.
   
          -d, --debug
                  emit  debugging  output.  Primarily (perhaps exclusively) of use
                  to developers.
    
           -v, --version
                  show version information and quit
    
           -h, --help
                  show a help synopsis
    
    
    SERVER SPECIFIC OPTIONS     SERVER SPECIFIC OPTIONS
   ‐s, ‐‐server          -s, --server
   run in server mode                 run in server mode
   
   ‐D, ‐‐daemon          -D, --daemon
   run the server in background as a daemon                 run the server in background as a daemon
   
   ‐I, ‐‐pidfile file          -I, --pidfile file
   write a file with the process ID, most useful when running as  a                 write a file with the process ID, most useful when running as  a
   daemon.                 daemon.
   
          -1, --one-off
                  handle one client connection, then exit.
    
           --rsa-private-key-path file
                  path  to  the  RSA  private key (not password-protected) used to
                  decrypt authentication credentials from  the  client  (if  built
                  with OpenSSL support).
    
           --authorized-users-path file
                  path  to the configuration file containing authorized users cre-
                  dentials to run iperf tests (if  built  with  OpenSSL  support).
                  The  file  is  a  comma separated list of usernames and password
                  hashes; more information on the structure of  the  file  can  be
                  found in the EXAMPLES section.
    
    CLIENT SPECIFIC OPTIONS     CLIENT SPECIFIC OPTIONS
   ‐c, ‐‐client host          -c, --client host
   run in client mode, connecting to the specified server                 run  in  client  mode,  connecting  to the specified server.  By
                  default, a test consists of sending data from the client to  the
                  server, unless the -R flag is specified.
    
           --sctp use SCTP rather than TCP (FreeBSD and Linux)
    
           -u, --udp
                  use UDP rather than TCP
    
           --connect-timeout n
                  set  timeout  for establishing the initial control connection to
                  the server, in milliseconds.  The default behavior is the  oper-
                  ating  system's  timeout for TCP connection establishment.  Pro-
                  viding a shorter value may speed up detection of a  down  iperf3
                  server.
    
           -b, --bitrate n[KM]
                  set  target  bitrate  to n bits/sec (default 1 Mbit/sec for UDP,
                  unlimited for TCP/SCTP).  If  there  are  multiple  streams  (-P
                  flag),  the  throughput  limit  is  applied  separately  to each
                  stream.  You can also add a '/' and  a  number  to  the  bitrate
                  specifier.  This is called "burst mode".  It will send the given
                  number of packets without  pausing,  even  if  that  temporarily
                  exceeds  the  specified  throughput  limit.   Setting the target
                  bitrate to 0 will disable bitrate  limits  (particularly  useful
                  for UDP tests).  This throughput limit is implemented internally
                  inside iperf3, and is available on all platforms.  Compare  with
                  the  --fq-rate flag.  This option replaces the --bandwidth flag,
                  which is now deprecated but (at least for now) still accepted.
    
           --pacing-timer n[KMG]
                  set  pacing  timer  interval  in  microseconds   (default   1000
                  microseconds,  or 1 ms).  This controls iperf3's internal pacing
                  timer for the -b/--bitrate  option.   The  timer  fires  at  the
                  interval  set  by  this parameter.  Smaller values of the pacing
                  timer parameter smooth out the traffic emitted  by  iperf3,  but
                  potentially  at  the  cost  of  performance due to more frequent
                  timer processing.
    
           --fq-rate n[KM]
                  Set a rate to be used with fair-queueing based socket-level pac-
                  ing,  in bits per second.  This pacing (if specified) will be in
                  addition to any pacing due to iperf3's internal throughput  pac-
                  ing  (-b/--bitrate flag), and both can be specified for the same
                  test.  Only available on platforms  supporting  the  SO_MAX_PAC-
                  ING_RATE  socket  option (currently only Linux).  The default is
                  no fair-queueing based pacing.
    
           --no-fq-socket-pacing
                  This option is deprecated and will be removed.  It is equivalent
                  to specifying --fq-rate=0.
    
           -t, --time n
                  time in seconds to transmit for (default 10 secs)
    
           -n, --bytes n[KM]
                  number of bytes to transmit (instead of -t)
    
           -k, --blockcount n[KM]
                  number of blocks (packets) to transmit (instead of -t or -n)
    
           -l, --length n[KM]
                  length  of  buffer to read or write.  For TCP tests, the default
                  value is 128KB.  In the case of UDP, iperf3 tries to dynamically
                  determine  a  reasonable  sending size based on the path MTU; if
                  that cannot be determined it uses 1460 bytes as a sending  size.
                  For SCTP tests, the default size is 64KB.
    
           --cport port
                  bind  data  streams  to  a specific client port (for TCP and UDP
                  only, default is to use an ephemeral port)
    
           -P, --parallel n
                  number of parallel client streams to run. Note  that  iperf3  is
                  single  threaded,  so  if you are CPU bound, this will not yield
                  higher throughput.
    
           -R, --reverse
                  reverse the direction of a test, so that the server  sends  data
                  to the client
   
   ‐‐sctp use SCTP rather than TCP (FreeBSD and Linux)          --bidir
                  bidirectional mode, server and client send and receive data.
    
           -w, --window n[KM]
                  window  size  / socket buffer size (this gets sent to the server
                  and used on that side too)
    
           -M, --set-mss n
                  set TCP/SCTP maximum segment size (MTU - 40 bytes)
    
           -N, --no-delay
                  set TCP/SCTP no delay, disabling Nagle's Algorithm
    
           -4, --version4
                  only use IPv4
    
           -6, --version6
                  only use IPv6
    
           -S, --tos n
                  set the IP type of service. The usual prefixes for octal and hex
                  can be used, i.e. 52, 064 and 0x34 all specify the same value.
    
           --dscp dscp
                  set  the  IP  DSCP  bits.   Both numeric and symbolic values are
                  accepted. Numeric values can be specified in decimal, octal  and
                  hex (see --tos above).
    
           -L, --flowlabel n
                  set the IPv6 flow label (currently only supported on Linux)
    
           -X, --xbind name
                  Bind  SCTP  associations  to  a  specific  subset of links using
                  sctp_bindx(3).  The --B flag will be ignored  if  this  flag  is
                  specified.  Normally SCTP will include the protocol addresses of
                  all active links on the local host when setting up  an  associa-
                  tion.  Specifying at least one --X name will disable this behav-
                  iour.  This flag must be specified for each link to be  included
                  in  the association, and is supported for both iperf servers and
                  clients (the latter are supported by passing the first --X argu-
                  ment  to  bind(2)).  Hostnames are accepted as arguments and are
                  resolved using getaddrinfo(3).  If the  --4  or  --6  flags  are
                  specified,  names  which  do not resolve to addresses within the
                  specified protocol family will be ignored.
    
           --nstreams n
                  Set number of SCTP streams.
    
           -Z, --zerocopy
                  Use a "zero copy" method of sending data, such  as  sendfile(2),
                  instead of the usual write(2).
    
           -O, --omit n
                  Omit the first n seconds of the test, to skip past the TCP slow-
                  start period.
    
           -T, --title str
                  Prefix every output line with this string.
    
           --extra-data str
                  Specify an extra data string field to be included in  JSON  out-
                  put.
    
           -C, --congestion algo
                  Set  the  congestion control algorithm (Linux and FreeBSD only).
                  An older --linux-congestion synonym for this  flag  is  accepted
                  but is deprecated.
    
           --get-server-output
                  Get the output from the server.  The output format is determined
                  by the server (in particular, if the server was invoked with the
                  --json  flag,  the  output  will be in JSON format, otherwise it
                  will be in human-readable format).  If the client  is  run  with
                  --json,  the  server output is included in a JSON object; other-
                  wise it is appended at the bottom of the human-readable  output.
    
           --repeating-payload
                  Use  repeating pattern in payload, instead of random bytes.  The
                  same payload is used in iperf2  (ASCII  '0..9'  repeating).   It
                  might  help  to test and reveal problems in networking gear with
                  hardware compression (including some WiFi access points),  where
                  iperf2  and  iperf3  perform  differently, just based on payload
                  entropy.
    
           --username username
                  username to use for authentication to the iperf server (if built
                  with OpenSSL support).  The password will be prompted for inter-
                  actively when the test is run.  Note, the password to use can
                  also be specified via the IPERF3_PASSWORD environment variable.
                  If this variable is present, the password prompt will be
                  skipped.
   
   ‐u, ‐‐udp          --rsa-public-key-path file
   use UDP rather than TCP                 path to the RSA public key used to encrypt  authentication  cre-
                 dentials (if built with OpenSSL support)
   ‐b, ‐‐bandwidth n[KM]   
   set  target bandwidth to n bits/sec (default 1 Mbit/sec for UDP,   
   unlimited for TCP).  If there are multiple  streams  (‐P  flag),   EXAMPLES
   the  bandwidth  limit is applied separately to each stream.  You      Authentication - RSA Keypair
   can also add a ’/’ and a  number  to  the  bandwidth  specifier.          The  authentication  feature  of iperf3 requires an RSA public keypair.
   This  is  called "burst mode".  It will send the given number of          The public key is used to encrypt the authentication  token  containing
   packets without pausing, even if that  temporarily  exceeds  the          the  user  credentials,  while  the  private key is used to decrypt the
   specified bandwidth limit.          authentication token.  An example of a set of  UNIX/Linux  commands  to
          generate correct keypair follows:
   ‐t, ‐‐time n   
   time in seconds to transmit for (default 10 secs)               > openssl genrsa -des3 -out private.pem 2048
               > openssl rsa -in private.pem -outform PEM -pubout -out public.pem
   ‐n, ‐‐bytes n[KM]               > openssl rsa -in private.pem -out private_not_protected.pem -out-
   number of bytes to transmit (instead of ‐t)               form PEM
   
   ‐k, ‐‐blockcount n[KM]          After these commands, the public key will be contained in the file pub-
   number of blocks (packets) to transmit (instead of ‐t or ‐n)          lic.pem and the  private  key  will  be  contained  in  the  file  pri-
          vate_not_protected.pem.
   ‐l, ‐‐length n[KM]   
   length  of  buffer to read or write (default 128 KB for TCP, 8KB      Authentication - Authorized users configuration file
   for UDP)          A  simple plaintext file must be provided to the iperf3 server in order
          to specify the authorized user credentials.  The file is a simple  list
   ‐P, ‐‐parallel n          of  comma-separated  pairs  of  a username and a corresponding password
   number of parallel client streams to run          hash.  The password hash is a SHA256 hash of the string  "{$user}$pass-
          word".   The file can also contain commented lines (starting with the #
   ‐R, ‐‐reverse          character).  An example of commands to generate the password hash on  a
   run in reverse mode (server sends, client receives)          UNIX/Linux system is given below:
   
   ‐w, ‐‐window n[KM]               > S_USER=mario S_PASSWD=rossi
   TCP window size / socket buffer size  (this  gets  sent  to  the               > echo -n "{$S_USER}$S_PASSWD" | sha256sum | awk '{ print $1 }'
   server and used on that side too)   
          An example of a password file (with an entry corresponding to the above
   ‐M, ‐‐set‐mss n          username and password) is given below:
   set TCP maximum segment size (MTU ‐ 40 bytes)               > cat credentials.csv
               # file format: username,sha256
   ‐N, ‐‐no‐delay               mario,bf7a49a846d44b454a5d11e7acfaf13d138bbe0b7483aa3e050879700572709b
   set TCP no delay, disabling Nagle’s Algorithm   
   
   ‐4, ‐‐version4   
   only use IPv4 
 
   ‐6, ‐‐version6 
   only use IPv6 
 
   ‐S, ‐‐tos n 
   set the IP ’type of service’ 
 
   ‐L, ‐‐flowlabel n 
   set the IPv6 flow label (currently only supported on Linux) 
 
   ‐Z, ‐‐zerocopy 
   Use  a  "zero copy" method of sending data, such as sendfile(2), 
   instead of the usual write(2). 
 
   ‐O, ‐‐omit n 
   Omit the first n seconds of the test, to skip past the TCP slow‐ 
   start period. 
 
   ‐T, ‐‐title str 
   Prefix every output line with this string. 
 
   ‐C, ‐‐linux‐congestion algo 
   Set the congestion control algorithm (linux only). 
 
 
    AUTHORS     AUTHORS
   Iperf  was  originally  written by Mark Gates and Alex Warshavsky.  Man          A list of the contributors to iperf3 can be found within the documenta-
   page and maintence by Jon Dugan <jdugan at x1024 dot net>.  Other  con‐          tion located at https://software.es.net/iperf/dev.html#authors.
   tributions  from  Ajay  Tirumala,  Jim Ferguson, Feng Qin, Kevin Gibbs,   
   John Estabrook <jestabro at ncsa.uiuc.edu>, Andrew  Gallatin  <gallatin   
   at gmail.com>, Stephen Hemminger <shemminger at linux‐foundation.org> 
 
 
    SEE ALSO     SEE ALSO
   libiperf(3), https://github.com/esnet/iperf          libiperf(3), https://software.es.net/iperf
   
   
   
   ESnet                            February 2014                        IPERF(1)   ESnet                              June 2018                         IPERF3(1)
   
 The iperf3 manual page will typically be installed in manual  The iperf3 manual page will typically be installed in manual
 section 1.  section 1.
Removed from v.1.1  
changed lines
  Added in v.1.1.1.2

FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>