1IPERF(1) User Manuals IPERF(1)
2
6 iperf - perform network traffic tests using network sockets. Metrics
7 include throughput and latency.
8
10 iperf -s [options]
11 iperf -c server [options]
13 iperf -u -s [options]
15 iperf -u -c server [options]
17
20 iperf 2 is a testing tool which performs network traffic measurements
21 using network sockets. The performance metrics supported include
22 throughput and latency. Iperf can use both TCP and UDP sockets (or pro‐
23 tocols.) It supports unidirectional, full duplex (same socket) and
24 bidirectional traffic, and supports multiple, simultaneous traffic
25 streams. It supports multicast traffic including source specific multi‐
26 cast (SSM) joins. Its multi-threaded design allows for peak perfor‐
27 mance. Metrics displayed help to characterize host to host network per‐
28 formance. Note: Setting the enhanced (-e) option provides all available
29 metrics.
30 The user must establish both a both a server (to receive traffic) and a
32 client (to generate and send traffic) for a test to occur. The client
33 and server typically are on different hosts or computers but need not
34 be.
35
37 -b, --bandwidth
38 set the target bandwidth and optional standard deviation per
39 <mean>,[<stdev>] (See NOTES for suffixes)
40 -e, --enhanced
42 Display enhanced output in reports otherwise use legacy report
43 (ver 2.0.5) formatting (see NOTES)
44 -f, --format [abkmgBKMG]
46 format to report: adaptive, bits, Bytes, Kbits, Mbits, Gbits,
47 KBytes, MBytes, GBytes (see NOTES for more)
48 -h, --help
50 print a help synopsis
51 --hide-ips
53 obscure ip addresses in output (useful when wanting to publish
54 results and not display the full ip addresses. v4 only)
55 -i, --interval < t | f >
57 sample or display interval reports every t seconds (default) or
58 every frame or burst, i.e. if f is used then the interval will
59 be each frame or burst. The frame interval reporting is experi‐
60 mental. Also suggest a compile with fast-sampling, i.e. ./con‐
61 figure --enable-fastsampling
62 -l, --len n[kmKM]
64 set read/write buffer size (TCP) or length (UDP) to n (TCP de‐
65 fault 128K, UDP default 1470)
66 --l2checks
68 perform layer 2 length checks on received UDP packets (requires
69 systems that support packet sockets, e.g. Linux)
70 -m, --print_mss
72 print TCP maximum segment size
73 --NUM_REPORT_STRUCTS <count>
75 Override the default shared memory size between the traffic
76 thread(s) and reporter thread in order to mitigate mutex lock
77 contentions. The default value of 5000 should be sufficient for
78 1Gb/s networks. Increase this upon seeing the Warning message of
79 reporter thread too slow. If the Warning message isn't seen,
80 then increasing this won't have any significant effect (other
81 than to use some additional memory.)
82 -o, --output filename
84 output the report or error message to this specified file
85 --permit-key [=<value>]
87 Set a key value that must match for the server to accept traffic
88 on a connection. If the option is given without a value on the
89 server a key value will be autogenerated and displayed in its
90 initial settings report. The lifetime of the key is set using
91 --permit-key-timeout and defaults to twenty seconds. The value
92 is required on clients. The value will also be used as part of
93 the transfer id in reports. The option set on the client but not
94 the server will also cause the server to reject the client's
95 traffic. TCP only, no UDP support.
96 -p, --port m[-n]
98 set client or server port(s) to send or listen on per m (default
99 5001) w/optional port range per m-n (e.g. -p 6002-6008) (see
100 NOTES)
101 --sum-dstip
103 sum traffic threads based upon the destination IP address (de‐
104 fault is source ip address)
105 --sum-only
107 set the output to sum reports only. Useful for -P at large val‐
108 ues
109 -t, --time n
111 time in seconds to listen for new traffic connections, receive
112 traffic or send traffic
113 -u, --udp
115 use UDP rather than TCP
116 -w, --window n[kmKM]
118 TCP window size (socket buffer size)
119 -z, --realtime
121 Request real-time scheduler, if supported.
122 -B, --bind host[:port][%dev]
124 bind to host, ip address or multicast address, optional port or
125 device (see NOTES)
126 -C, --compatibility
128 for use with older versions does not sent extra msgs
129 -M, --mss n
131 set TCP maximum segment size using TCP_MAXSEG
132 -N, --nodelay
134 set TCP no delay, disabling Nagle's Algorithm
135 -v, --version
137 print version information and quit
138 -x, --reportexclude [CDMSV]
140 exclude C(connection) D(data) M(multicast) S(settings) V(server)
141 reports
142 -y, --reportstyle C|c
144 if set to C or c report results as CSV (comma separated values)
145 -Z, --tcp-congestion
147 Set the default congestion-control algorithm to be used for new
148 connections. Platforms must support setsockopt's TCP_CONGESTION.
149 (Notes: See sysctl and tcp_allowed_congestion_control for avail‐
150 able options. May require root privileges.)
151
153 -1, --singleclient
154 set the server to process only one client at a time
155 -b, --bandwidth n[kmgKMG]
157 set target read rate to n bits/sec. TCP only for the server.
158 -s, --server
160 run in server mode
161 --histograms[=binwidth[u],bincount,[lowerci],[upperci]]
163 enable latency histograms for udp packets (-u), for tcp writes
164 (with --trip-times), or for either udp or tcp with --isochronous
165 clients. The binning can be modified. Bin widths (default 1 mil‐
166 lisecond, append u for microseconds, m for milliseconds) bin‐
167 count is total bins (default 1000), ci is confidence interval
168 between 0-100% (default lower 5%, upper 95%, 3 stdev 99.7%)
169 --permit-key [=<value>]
171 Set a key value that must match for the server to accept traffic
172 from a client (also set with --permit-key.) The server will au‐
173 togenerate a globally unique key when the option is given with‐
174 out a value. This value will be displayed in the server's ini‐
175 tial settings report. The lifetime of the key is set using
176 --permit-key-timeout and defaults to twenty seconds. TCP only,
177 no UDP support.
178 --permit-key-timeout <value>
180 Set the lifetime of the permit key in seconds. Defaults to 20
181 seconds if not set. A value of zero will disable the timer.
182 --tap-dev <dev>
184 Set the receive interface to the TAP device as specified.
185 --tcp-rx-window-clamp n[kmKM]
187 Set the socket option of TCP_WINDOW_CLAMP, units is bytes.
188 -t, --time n
190 time in seconds to listen for new traffic connections and/or re‐
191 ceive traffic (defaults to infinite)
192 --tos-override n
194 set the socket's IP_TOS (byte) field for reverse or full duplex
195 traffic. Supported in versions 2.1.5 or greater. Previous ver‐
196 sions won't set IP_TOS on reverse traffic.
197 -B, --bind ip | ip%device
199 bind src ip addr and optional src device for receiving
200 -D, --daemon
202 run the server as a daemon. On Windows this will run the speci‐
203 fied command-line under the IPerfService, installing the service
204 if necessary. Note the service is not configured to auto-start
205 or restart - if you need a self-starting service you will need
206 to create an init script or use Windows "sc" commands.
207 -H, --ssm-host host
209 Set the source host (ip addr) per SSM multicast, i.e. the S of
210 the S,G
211 -R, --remove
213 remove the IPerfService (Windows only).
214 -U, --single_udp
216 run in single threaded UDP mode
217 -V, --ipv6_domain
219 Enable IPv6 reception by setting the domain and socket to
220 AF_INET6 (Can receive on both IPv4 and IPv6)
221
223 -b, --bandwidth n[kmgKMG][,n[kmgKMG]] | n[kmgKMG]pps
224 set target bandwidth to n bits/sec (default 1 Mbit/sec) or n
225 packets per sec. This may be used with TCP or UDP. Optionally,
226 for variable loads, use format of mean,standard deviation
227 --bounceback
229 run a tcp bounceback test (set size with -l or --len, defaults
230 to 100 bytes)
231 --bounceback-congest
233 request a concurrent full-duplex TCP stream
234 --bounceback-hold n
236 request the server to insert a delay of n milliseconds between
237 its read and write (default is no delay)
238 --bounceback-period n
240 request the client schedule a send every n milliseconds (w/o op‐
241 tion bouncebacks are immediately scheduled. With command option
242 and no optional value given, the interval will be one second.)
243 --bounceback-no-quickack
245 request the server not set the TCP_QUICKACK socket option (dis‐
246 abling TCP ACK delays) during a bounceback test (see NOTES)
247 --burst-period n
249 Set the burst period in seconds. Defaults to one second. (Note:
250 assumed use case is low duty cycle traffic bursts)
251 --burst-size n
253 Set the burst size in bytes. Defaults to 1M if no value is
254 given. -c, --client host | host%device run in client mode, con‐
255 necting to host where the optional %dev will SO_BINDTODEVICE
256 that output interface (requires root and see NOTES)
257 --connect-only[=n]
260 only perform a TCP connect (or 3WHS) without any data transfer,
261 useful to measure TCP connect() times. Optional value of n is
262 the total number of connects to do (zero is run forever.) Note
263 that -i will rate limit the connects where -P will create bursts
264 and -t will end the client and hence end its connect attempts.
265 --connect-retries[= n]
267 number of times to retry a TCP connect at the application level.
268 See operating system information on the details of TCP connect
269 related settings.
270 -d, --dualtest
272 Do a bidirectional test simultaneous test using two unidirec‐
273 tional sockets
274 --fq-rate n[kmgKMG]
276 Set a rate to be used with fair-queueing based socket-level pac‐
277 ing, in bytes or bits per second. Only available on platforms
278 supporting the SO_MAX_PACING_RATE socket option. (Note: Here the
279 suffixes indicate bytes/sec or bits/sec per use of uppercase or
280 lowercase, respectively)
281 --full-duplex
283 run a full duplex test, i.e. traffic in both transmit and re‐
284 ceive directions using the same socket
285 --histograms[=binwidth[u],bincount,[lowerci],[upperci]]
287 enable select()/write() histograms with --tcp-write-prefetch.
288 The binning can be modified. Bin widths (default 100 microsec‐
289 onds, append u for microseconds, m for milliseconds) bincount is
290 total bins (default 10000), ci is confidence interval between
291 0-100% (default lower 5%, upper 95%, 3 stdev 99.7%)
292 --incr-dstip
294 increment the destination ip address when using the parallel
295 (-P) or port range option
296 --incr-dstport
298 increment the destination port when using the parallel (-P) or
299 port range option
300 --incr-srcip
302 increment the source ip address when using the parallel (-P) or
303 port range option
304 --incr-srcport
306 increment the source ip address when using the parallel (-P) or
307 port range option, requires -B to set the src port
308 --ipg n
310 set the inter-packet gap to n (units of seconds) for packets or
311 within a frame/burst when --isochronous is set
312 --isochronous[=fps:mean,stdev]
314 send isochronous traffic with frequency frames per second and
315 load defined by mean and standard deviation using a log normal
316 distribution, defaults to 60:20m,0. (Note: Here the suffixes in‐
317 dicate bytes/sec or bits/sec per use of uppercase or lowercase,
318 respectively. Also the p suffix is supported to set the burst
319 size in packets, e.g. isochronous=2:25p will send two 25 packet
320 bursts every second, or one 25 packet burst every 0.5 seconds.)
321 --local-only[=1|0]
323 Set 1 to limit traffic to the local network only (through the
324 use of SO_DONTROUTE) set to zero otherwise with optional over‐
325 ride of compile time default (see configure --default-localonly)
326 --near-congestion[=n]
328 Enable TCP write rate limiting per the sampled RTT. The delay is
329 applied after the -l number of bytes have completed. The op‐
330 tional value is the multiplier to the RTT and defines the time
331 delay. This value defaults to 0.5 if it is not set. Values less
332 than 1 are supported but the value cannot be negative. This is
333 an experimental feature. It is not likely stable on live net‐
334 works. Suggested use is over controlled test networks.
335 --no-connect-sync
337 By default, parallel traffic threads (per -P greater than 1)
338 will synchronize after their TCP connects and prior to each
339 sending traffic, i.e. all the threads first complete (or error)
340 the TCP 3WHS before any traffic thread will start sending. This
341 option disables that synchronization such that each traffic
342 thread will start sending immediately after completing its suc‐
343 cessful connect.
344 --no-udp-fin
346 Don't perform the UDP final server to client exchange which
347 means there won't be a final server report displayed on the
348 client. All packets per the test will be from the client to the
349 server and no packets should be sent in the other direction.
350 It's highly suggested that -t be set on the server if this op‐
351 tion is being used. This is because there will be only one
352 trigger ending packet sent from client to server and if it's
353 lost then the server will continue to run. (Requires ver 2.0.14
354 or better)
355 -n, --num n[kmKM]
357 number of bytes to transmit (instead of -t)
358 --permit-key [=<value>]
360 Set a key value that must match the server's value (also set
361 with --permit-key) in order for the server to accept traffic
362 from the client. TCP only, no UDP support.
363 -r, --tradeoff
365 Do a bidirectional test individually - client-to-server, fol‐
366 lowed by a reversed test, server-to-client
367 --tcp-quickack
369 Set TCP_QUICKACK on the socket
370 --tcp-write-prefetch n[kmKM]
372 Set TCP_NOTSENT_LOWAT on the socket and use event based writes
373 per select() on the socket.
374 --tcp-write-times
376 Measure the socket write times
377 -t, --time n|0
379 time in seconds to transmit traffic, use zero for infinite (de‐
380 fault is 10 secs)
381 --trip-times
383 enable the measurement of end to end write to read latencies
384 (client and server clocks must be synchronized)
385 --txdelay-time
387 time in seconds to hold back or delay after the TCP connect and
388 prior to the socket writes. For UDP it's the delay between the
389 traffic thread starting and the first write.
390 --txstart-time n.n
392 set the txstart-time to n.n using unix or epoch time format
393 (supports microsecond resolution, e.g 1536014418.123456) An ex‐
394 ample to delay one second using command substitution is iperf -c
395 192.168.1.10 --txstart-time $(expr $(date +%s) + 1).$(date +%N)
396 -B, --bind ip | ip:port | ipv6 -V | [ipv6]:port -V
398 bind src ip addr and optional port as the source of traffic (see
399 NOTES)
400 -F, --fileinput name
402 input the data to be transmitted from a file
403 -I, --stdin
405 input the data to be transmitted from stdin
406 -L, --listenport n
408 port to receive bidirectional tests back on
409 -P, --parallel n
411 number of parallel client threads to run
412 -R, --reverse
414 reverse the traffic flow (useful for testing through firewalls,
415 see NOTES)
416 -S, --tos n
418 set the socket's IP_TOS (byte) field. Versions 2.1.5 or greater
419 will reflect this tos setting back with --reverse or --full-du‐
420 plex option. (Previous versions won't set tos on the reverse
421 traffic.) Note: use server side --tos-override to override.
422 -T, --ttl n
424 time-to-live, for multicast (default 1) -V, --ipv6_domain Set
425 the domain to IPv6 (send packets over IPv6)
426 -X, --peerdetect
428 run peer version detection prior to traffic.
429 -Z, --linux-congestion algo
431 set TCP congestion control algorithm (Linux only)
432
434 TCP tests (client)
435 iperf -c <host> -e -i 1
437 ------------------------------------------------------------
438 Client connecting to <host>, TCP port 5001 with pid 5149
439 Write buffer size: 128 KByte
440 TCP window size: 340 KByte (default)
441 ------------------------------------------------------------
442 [ 3] local 45.56.85.133 port 49960 connected with 45.33.58.123 port
443 5001 (ct=3.23 ms)
444 [ ID] Interval Transfer Bandwidth Write/Err Rtry
445 Cwnd/RTT NetPwr
446 [ 3] 0.00-1.00 sec 126 MBytes 1.05 Gbits/sec 1006/0 0
447 56K/626 us 210636.47
448 [ 3] 1.00-2.00 sec 138 MBytes 1.15 Gbits/sec 1100/0 299
449 483K/3884 us 37121.32
450 [ 3] 2.00-3.00 sec 137 MBytes 1.15 Gbits/sec 1093/0 24
451 657K/5087 us 28162.31
452 [ 3] 3.00-4.00 sec 126 MBytes 1.06 Gbits/sec 1010/0 284
453 294K/2528 us 52366.58
454 [ 3] 4.00-5.00 sec 117 MBytes 980 Mbits/sec 935/0 373
455 487K/2025 us 60519.66
456 [ 3] 5.00-6.00 sec 144 MBytes 1.20 Gbits/sec 1149/0 2
457 644K/3570 us 42185.36
458 [ 3] 6.00-7.00 sec 126 MBytes 1.06 Gbits/sec 1011/0 112
459 582K/5281 us 25092.56
460 [ 3] 7.00-8.00 sec 110 MBytes 922 Mbits/sec 879/0 56
461 279K/1957 us 58871.89
462 [ 3] 8.00-9.00 sec 127 MBytes 1.06 Gbits/sec 1014/0 46
463 483K/3372 us 39414.89
464 [ 3] 9.00-10.00 sec 132 MBytes 1.11 Gbits/sec 1054/0 0
465 654K/3380 us 40872.75
466 [ 3] 0.00-10.00 sec 1.25 GBytes 1.07 Gbits/sec 10251/0 1196
467 -1K/3170 us 42382.03
468 where (per -e,)
471 ct= TCP connect time (or three way handshake time 3WHS)
472 Write/Err Total number of successful socket writes. Total number
473 of non-fatal socket write errors
474 Rtry Total number of TCP retries
475 Cwnd/RTT (*nix only) TCP congestion window and round trip time
476 (sampled where NA indicates no value)
477 NetPwr (*nix only) Network power defined as (throughput / RTT)
478 TCP tests (server)
481 iperf -s -e -i 1 -l 8K
483 ------------------------------------------------------------
484 Server listening on TCP port 5001 with pid 13430
485 Read buffer size: 8.00 KByte
486 TCP window size: 85.3 KByte (default)
487 ------------------------------------------------------------
488 [ 4] local 45.33.58.123 port 5001 connected with 45.56.85.133 port
489 49960
490 [ ID] Interval Transfer Bandwidth Reads
491 Dist(bin=1.0K)
492 [ 4] 0.00-1.00 sec 124 MBytes 1.04 Gbits/sec 22249
493 798:2637:2061:767:2165:1563:589:11669
494 [ 4] 1.00-2.00 sec 136 MBytes 1.14 Gbits/sec 24780
495 946:3227:2227:790:2427:1888:641:12634
496 [ 4] 2.00-3.00 sec 137 MBytes 1.15 Gbits/sec 24484
497 1047:2686:2218:810:2195:1819:728:12981
498 [ 4] 3.00-4.00 sec 126 MBytes 1.06 Gbits/sec 20812
499 863:1353:1546:614:1712:1298:547:12879
500 [ 4] 4.00-5.00 sec 117 MBytes 984 Mbits/sec 20266
501 769:1886:1828:589:1866:1350:476:11502
502 [ 4] 5.00-6.00 sec 143 MBytes 1.20 Gbits/sec 24603
503 1066:1925:2139:822:2237:1827:744:13843
504 [ 4] 6.00-7.00 sec 126 MBytes 1.06 Gbits/sec 22635
505 834:2464:2249:724:2269:1646:608:11841
506 [ 4] 7.00-8.00 sec 110 MBytes 921 Mbits/sec 21107
507 842:2437:2747:592:2871:1903:496:9219
508 [ 4] 8.00-9.00 sec 126 MBytes 1.06 Gbits/sec 22804
509 1038:1784:2639:656:2738:1927:573:11449
510 [ 4] 9.00-10.00 sec 133 MBytes 1.11 Gbits/sec 23091
511 1088:1654:2105:710:2333:1928:723:12550
512 [ 4] 0.00-10.02 sec 1.25 GBytes 1.07 Gbits/sec 227306
513 9316:22088:21792:7096:22893:17193:6138:120790
514 where (per -e,)
516 Reads Total number of socket reads
517 Dist(bin=size) Eight bin histogram of the socket reads returned
518 byte count. Bin width is set per size. Bins are separated by a
519 colon. In the example, the bins are 0-1K, 1K-2K, .., 7K-8K.
520 TCP tests (server with --trip-times on client) iperf -s -i 1 -w 4M
523 ------------------------------------------------------------
524 Server listening on TCP port 5001
525 TCP window size: 8.00 MByte (WARNING: requested 4.00 MByte)
526 ------------------------------------------------------------
527 [ 4] local 192.168.1.4%eth0 port 5001 connected with 192.168.1.7 port
528 44798 (trip-times) (MSS=1448) (peer 2.0.14-alpha)
529 [ ID] Interval Transfer Bandwidth Burst Latency
530 avg/min/max/stdev (cnt/size) inP NetPwr Reads=Dist
531 [ 4] 0.00-1.00 sec 19.0 MBytes 159 Mbits/sec
532 52.314/10.238/117.155/19.779 ms (151/131717) 1.05 MByte 380.19
533 781=306:253:129:48:18:15:8:4
534 [ 4] 1.00-2.00 sec 20.0 MBytes 168 Mbits/sec
535 53.863/21.264/79.252/12.277 ms (160/131080) 1.08 MByte 389.38
536 771=294:236:126:60:18:24:10:3
537 [ 4] 2.00-3.00 sec 18.2 MBytes 153 Mbits/sec
538 58.718/22.000/137.944/20.397 ms (146/130964) 1.06 MByte 325.64
539 732=299:231:98:52:18:19:10:5
540 [ 4] 3.00-4.00 sec 19.7 MBytes 165 Mbits/sec 50.448/
541 8.921/82.728/14.627 ms (158/130588) 997 KByte 409.00
542 780=300:255:121:58:15:18:7:6
543 [ 4] 4.00-5.00 sec 18.8 MBytes 158 Mbits/sec
544 53.826/11.169/115.316/15.541 ms (150/131420) 1.02 MByte 366.24
545 761=302:226:134:52:22:17:7:1
546 [ 4] 5.00-6.00 sec 19.5 MBytes 164 Mbits/sec
547 50.943/11.922/76.134/14.053 ms (156/131276) 1.03 MByte 402.00
548 759=273:246:149:45:16:18:4:8
549 [ 4] 6.00-7.00 sec 18.5 MBytes 155 Mbits/sec
550 57.643/10.039/127.850/18.950 ms (148/130926) 1.05 MByte 336.16
551 710=262:228:133:37:16:20:8:6
552 [ 4] 7.00-8.00 sec 19.6 MBytes 165 Mbits/sec
553 52.498/12.900/77.045/12.979 ms (157/131003) 1.00 MByte 391.78
554 742=288:200:135:68:16:23:4:8
555 [ 4] 8.00-9.00 sec 18.0 MBytes 151 Mbits/sec 58.370/
556 8.026/150.243/21.445 ms (144/131255) 1.06 MByte 323.81
557 716=268:241:108:51:20:17:8:3
558 [ 4] 9.00-10.00 sec 18.4 MBytes 154 Mbits/sec
559 56.112/12.419/79.790/13.668 ms (147/131194) 1.05 MByte 343.70
560 822=330:303:120:26:16:14:9:4
561 [ 4] 10.00-10.06 sec 1.03 MBytes 146 Mbits/sec
562 69.880/45.175/78.754/10.823 ms (9/119632) 1.74 MByte 260.40
563 62=26:30:5:1:0:0:0:0
564 [ 4] 0.00-10.06 sec 191 MBytes 159 Mbits/sec 54.183/
565 8.026/150.243/16.781 ms (1526/131072) 1.03 MByte 366.98
566 7636=2948:2449:1258:498:175:185:75:48
567 where (per -e,)
569 Burst Latency One way TCP write() to read() latency in mean/min‐
570 imum/maximum/standard deviation format (Note: requires the
571 client's and server's system clocks to be synchronized to a com‐
572 mon reference, e.g. using precision time protocol PTP. A GPS
573 disciplined OCXO is a recommended reference.)
574 cnt Number of completed bursts received and used for the burst
575 latency calculations
576 size Average burst size in bytes (computed average and estimate
577 only)
578 inP inP, short for in progress, is the average number of bytes
579 in progress or in flight. This is taken from the application
580 level write to read perspective. Note this is a mean value. The
581 parenthesis value is the standard deviation from the mean. (Re‐
582 quires --trip-times on client. See Little's law in NOTES.)
583 NetPwr Network power defined as (throughput / one way latency)
584 TCP tests (with one way delay sync check -X and --trip-times on the
587 client)
588 iperf -c 192.168.1.4 -X -e --trip-times -i 1 -t 2
590 ------------------------------------------------------------
591 Client connecting to 192.168.1.4, TCP port 5001 with pid 16762 (1
592 flows)
593 Write buffer size: 131072 Byte
594 TCP window size: 85.0 KByte (default)
595 ------------------------------------------------------------
596 [ 1] Clock sync check (ms): RTT/Half=(3.361/1.680) OWD-
597 send/ack/asym=(2.246/1.115/1.131)
598 [ 1] local 192.168.1.1%ap0 port 47466 connected with 192.168.1.4 port
599 5001 (MSS=1448) (trip-times) (sock=3) (peer 2.1.4-master)
600 [ ID] Interval Transfer Bandwidth Write/Err Rtry
601 Cwnd/RTT NetPwr
602 [ 1] 0.00-1.00 sec 9.50 MBytes 79.7 Mbits/sec 77/0 0
603 2309K/113914 us 87
604 [ 1] 1.00-2.00 sec 7.12 MBytes 59.8 Mbits/sec 57/0 0
605 2492K/126113 us 59
606 [ 1] 2.00-2.42 sec 128 KBytes 2.47 Mbits/sec 2/0 0
607 2492K/126113 us 2
608 [ 1] 0.00-2.42 sec 16.8 MBytes 58.0 Mbits/sec 136/0 0
609 2492K/126113 us 57
610 UDP tests (client)
613 iperf -c <host> -e -i 1 -u -b 10m
615 ------------------------------------------------------------
616 Client connecting to <host>, UDP port 5001 with pid 5169
617 Sending 1470 byte datagrams, IPG target: 1176.00 us (kalman adjust)
618 UDP buffer size: 208 KByte (default)
619 ------------------------------------------------------------
620 [ 3] local 45.56.85.133 port 32943 connected with 45.33.58.123 port
621 5001
622 [ ID] Interval Transfer Bandwidth Write/Err PPS
623 [ 3] 0.00-1.00 sec 1.19 MBytes 10.0 Mbits/sec 852/0 851 pps
624 [ 3] 1.00-2.00 sec 1.19 MBytes 10.0 Mbits/sec 850/0 850 pps
625 [ 3] 2.00-3.00 sec 1.19 MBytes 10.0 Mbits/sec 850/0 850 pps
626 [ 3] 3.00-4.00 sec 1.19 MBytes 10.0 Mbits/sec 851/0 850 pps
627 [ 3] 4.00-5.00 sec 1.19 MBytes 10.0 Mbits/sec 850/0 850 pps
628 [ 3] 5.00-6.00 sec 1.19 MBytes 10.0 Mbits/sec 850/0 850 pps
629 [ 3] 6.00-7.00 sec 1.19 MBytes 10.0 Mbits/sec 851/0 850 pps
630 [ 3] 7.00-8.00 sec 1.19 MBytes 10.0 Mbits/sec 850/0 850 pps
631 [ 3] 8.00-9.00 sec 1.19 MBytes 10.0 Mbits/sec 851/0 850 pps
632 [ 3] 0.00-10.00 sec 11.9 MBytes 10.0 Mbits/sec 8504/0 850 pps
633 [ 3] Sent 8504 datagrams
634 [ 3] Server Report:
635 [ 3] 0.00-10.00 sec 11.9 MBytes 10.0 Mbits/sec 0.047 ms 0/ 8504
636 (0%) 0.537/ 0.392/23.657/ 0.497 ms 850 pps 2329.37
637 where (per -e,)
639 Write/Err Total number of successful socket writes. Total number
640 of non-fatal socket write errors
641 PPS Transmit packet rate in packets per second
642 UDP tests (server) iperf -s -i 1 -w 4M -u
645 ------------------------------------------------------------
646 Server listening on UDP port 5001
647 Receiving 1470 byte datagrams
648 UDP buffer size: 8.00 MByte (WARNING: requested 4.00 MByte)
649 ------------------------------------------------------------
650 [ 3] local 192.168.1.4 port 5001 connected with 192.168.1.1 port 60027
651 (WARN: winsize=8.00 MByte req=4.00 MByte) (trip-times) (0.0) (peer
652 2.0.14-alpha)
653 [ ID] Interval Transfer Bandwidth Jitter Lost/Total
654 Latency avg/min/max/stdev PPS inP NetPwr
655 [ 3] 0.00-1.00 sec 44.5 MBytes 373 Mbits/sec 0.071 ms 52198/83938
656 (62%) 75.185/ 2.367/85.189/14.430 ms 31854 pps 3.64 MByte 620.58
657 [ 3] 1.00-2.00 sec 44.8 MBytes 376 Mbits/sec 0.015 ms
658 59549/143701 (41%) 79.609/75.603/85.757/ 1.454 ms 31954 pps 3.56 MByte
659 590.04
660 [ 3] 2.00-3.00 sec 44.5 MBytes 373 Mbits/sec 0.017 ms
661 59494/202975 (29%) 80.006/75.951/88.198/ 1.638 ms 31733 pps 3.56 MByte
662 583.07
663 [ 3] 3.00-4.00 sec 44.5 MBytes 373 Mbits/sec 0.019 ms
664 59586/262562 (23%) 79.939/75.667/83.857/ 1.145 ms 31767 pps 3.56 MByte
665 583.57
666 [ 3] 4.00-5.00 sec 44.5 MBytes 373 Mbits/sec 0.081 ms
667 59612/322196 (19%) 79.882/75.400/86.618/ 1.666 ms 31755 pps 3.55 MByte
668 584.40
669 [ 3] 5.00-6.00 sec 44.7 MBytes 375 Mbits/sec 0.064 ms
670 59571/381918 (16%) 79.767/75.571/85.339/ 1.556 ms 31879 pps 3.56 MByte
671 588.02
672 [ 3] 6.00-7.00 sec 44.6 MBytes 374 Mbits/sec 0.041 ms
673 58990/440820 (13%) 79.722/75.662/85.938/ 1.087 ms 31820 pps 3.58 MByte
674 586.73
675 [ 3] 7.00-8.00 sec 44.7 MBytes 375 Mbits/sec 0.027 ms
676 59679/500548 (12%) 79.745/75.704/84.731/ 1.094 ms 31869 pps 3.55 MByte
677 587.46
678 [ 3] 8.00-9.00 sec 44.3 MBytes 371 Mbits/sec 0.078 ms
679 59230/559499 (11%) 80.346/75.514/94.293/ 2.858 ms 31590 pps 3.58 MByte
680 577.97
681 [ 3] 9.00-10.00 sec 44.4 MBytes 373 Mbits/sec 0.073 ms
682 58782/618394 (9.5%) 79.125/75.511/93.638/ 1.643 ms 31702 pps 3.55 MByte
683 588.99
684 [ 3] 10.00-10.08 sec 3.53 MBytes 367 Mbits/sec 0.129 ms
685 6026/595236 (1%) 94.967/80.709/99.685/ 3.560 ms 31107 pps 3.58 MByte
686 483.12
687 [ 3] 0.00-10.08 sec 449 MBytes 374 Mbits/sec 0.129 ms
688 592717/913046 (65%) 79.453/ 2.367/99.685/ 5.200 ms 31776 pps (null)
689 587.91
690 where (per -e,)
693 Latency End to end latency in mean/minimum/maximum/standard de‐
694 viation format (Note: requires the client's and server's system
695 clocks to be synchronized to a common reference, e.g. using pre‐
696 cision time protocol PTP. A GPS disciplined OCXO is a recom‐
697 mended reference.)
698 PPS Received packet rate in packets per second
699 inP inP, short for in progress, is the average number of bytes
700 in progress or in flight. This is taken from an application
701 write to read perspective. (Requires --trip-times on client. See
702 Little's law in NOTES.)
703 NetPwr Network power defined as (throughput / latency)
704 Isochronous UDP tests (client)
707 iperf -c 192.168.100.33 -u -e -i 1 --isochronous=60:100m,10m --realtime
709 ------------------------------------------------------------
710 Client connecting to 192.168.100.33, UDP port 5001 with pid 14971
711 UDP isochronous: 60 frames/sec mean= 100 Mbit/s, stddev=10.0 Mbit/s,
712 Period/IPG=16.67/0.005 ms
713 UDP buffer size: 208 KByte (default)
714 ------------------------------------------------------------
715 [ 3] local 192.168.100.76 port 42928 connected with 192.168.100.33
716 port 5001
717 [ ID] Interval Transfer Bandwidth Write/Err PPS
718 frames:tx/missed/slips
719 [ 3] 0.00-1.00 sec 12.0 MBytes 101 Mbits/sec 8615/0 8493 pps
720 62/0/0
721 [ 3] 1.00-2.00 sec 12.0 MBytes 100 Mbits/sec 8556/0 8557 pps
722 60/0/0
723 [ 3] 2.00-3.00 sec 12.0 MBytes 101 Mbits/sec 8586/0 8586 pps
724 60/0/0
725 [ 3] 3.00-4.00 sec 12.1 MBytes 102 Mbits/sec 8687/0 8687 pps
726 60/0/0
727 [ 3] 4.00-5.00 sec 11.8 MBytes 99.2 Mbits/sec 8468/0 8468 pps
728 60/0/0
729 [ 3] 5.00-6.00 sec 11.9 MBytes 99.8 Mbits/sec 8519/0 8520 pps
730 60/0/0
731 [ 3] 6.00-7.00 sec 12.1 MBytes 102 Mbits/sec 8694/0 8694 pps
732 60/0/0
733 [ 3] 7.00-8.00 sec 12.1 MBytes 102 Mbits/sec 8692/0 8692 pps
734 60/0/0
735 [ 3] 8.00-9.00 sec 11.9 MBytes 100 Mbits/sec 8537/0 8537 pps
736 60/0/0
737 [ 3] 9.00-10.00 sec 11.8 MBytes 99.0 Mbits/sec 8450/0 8450 pps
738 60/0/0
739 [ 3] 0.00-10.01 sec 120 MBytes 100 Mbits/sec 85867/0 8574 pps
740 602/0/0
741 [ 3] Sent 85867 datagrams
742 [ 3] Server Report:
743 [ 3] 0.00-9.98 sec 120 MBytes 101 Mbits/sec 0.009 ms 196/85867
744 (0.23%) 0.665/ 0.083/ 1.318/ 0.174 ms 8605 pps 18903.85
745 where (per -e,)
747 frames:tx/missed/slips Total number of isochronous frames or
748 bursts. Total number of frame ids not sent. Total number of
749 frame slips
750 Isochronous UDP tests (server)
753 iperf -s -e -u --udp-histogram=100u,2000 --realtime
755 ------------------------------------------------------------
756 Server listening on UDP port 5001 with pid 5175
757 Receiving 1470 byte datagrams
758 UDP buffer size: 208 KByte (default)
759 ------------------------------------------------------------
760 [ 3] local 192.168.100.33 port 5001 connected with 192.168.100.76 port
761 42928 isoch (peer 2.0.13-alpha)
762 [ ID] Interval Transfer Bandwidth Jitter Lost/Total
763 Latency avg/min/max/stdev PPS NetPwr Frames/Lost
764 [ 3] 0.00-9.98 sec 120 MBytes 101 Mbits/sec 0.010 ms 196/85867
765 (0.23%) 0.665/ 0.083/ 1.318/ 0.284 ms 8585 pps 18903.85 601/1
766 [ 3] 0.00-9.98 sec T8(f)-PDF:
767 bin(w=100us):cnt(85671)=1:2,2:844,3:10034,4:8493,5:8967,6:8733,7:8823,8:9023,9:8901,10:8816,11:7730,12:4563,13:741,14:1
768 (5.00/95.00%=3/12,Outliers=0,obl/obu=0/0)
769 [ 3] 0.00-9.98 sec F8(f)-PDF:
770 bin(w=100us):cnt(598)=15:2,16:1,17:27,18:68,19:125,20:136,21:103,22:83,23:22,24:23,25:5,26:3
771 (5.00/95.00%=17/24,Outliers=0,obl/obu=0/0)
772 where, Frames/lost Total number of frames (or bursts) received. Total
775 number of bursts lost or error-ed
776 T8-PDF(f) Latency histogram for packets
777 F8-PDF(f) Latency histogram for frames
778
782 Note: The environment variable option settings haven't been maintained
783 well. See the source code if these are of interest.
784
786 Numeric options: Some numeric options support format characters per
787 '<value>c' (e.g. 10M) where the c format characters are k,m,g,K,M,G.
788 Lowercase format characters are 10^3 based and uppercase are 2^n based,
789 e.g. 1k = 1000, 1K = 1024, 1m = 1,000,000 and 1M = 1,048,576
790 Rate limiting: The -b option supports read and write rate limiting at
792 the application level. The -b option on the client also supports vari‐
793 able offered loads through the <mean>,<standard deviation> format, e.g.
794 -b 100m,10m. The distribution used is log normal. Similar for the
795 isochronous option. The -b on the server rate limits the reads. Socket
796 based pacing is also supported using the --fq-rate long option. This
797 will work with the --reverse and --full-duplex options as well.
798 Synchronized clocks: The --trip-times option indicates that the
800 client's and server's clocks are synchronized to a common reference.
801 Network Time Protocol (NTP) or Precision Time Protocol (PTP) are com‐
802 monly used for this. The reference clock(s) error and the synchroniza‐
803 tion protocols will affect the accuracy of any end to end latency mea‐
804 surements.
805 Histograms and non-parametric statisitics: The --histograms option pro‐
807 vides the raw data where nothing is averaged. This is useful for non-
808 parametric distribtions, e.g. latency. The standard output does use the
809 central limit theorem to produce average, mininimum, maximum and varia‐
810 tion. This loses information when the underlining distribution is not
811 gaussian. Histograms are supported so this information is made avail‐
812 able.
813 Binding is done at the logical level of port and ip address (or layer
815 3) using the -B option and a colon as the separator between port and
816 the ip addr. Binding at the device (or layer 2) level requires the per‐
817 cent (%) as the delimeter (for both the client and the server.) An ex‐
818 ample for src port and ip address is -B 192.168.1.1:6001. To bind the
819 src port only and let the operating system choose the source ip address
820 use 0.0.0.0, e.g. -B 0.0.0.0:6001. On the client, the -B option af‐
821 fects the bind(2) system call, and will set the source ip address and
822 the source port, e.g. iperf -c <host> -B 192.168.100.2:6002. This con‐
823 trols the packet's source values but not routing. These can be confus‐
824 ing in that a route or device lookup may not be that of the device with
825 the configured source IP. So, for example, if the IP address of eth0
826 is used for -B and the routing table for the destination IP address re‐
827 solves the output interface to be eth1, then the host will send the
828 packet out device eth1 while using the source IP address of eth0 in the
829 packet. To affect the physical output interface (e.g. dual homed sys‐
830 tems) either use -c <host>%<dev> (requires root) which bypasses this
831 host route table lookup, or configure policy routing per each -B source
832 address and set the output interface appropriately in the policy
833 routes. On the server or receive, only packets destined to -B IP ad‐
834 dress will be received. It's also useful for multicast. For example,
835 iperf -s -B 224.0.0.1%eth0 will only accept ip multicast packets with
836 dest ip 224.0.0.1 that are received on the eth0 interface, while iperf
837 -s -B 224.0.0.1 will receive those packets on any interface, Finally,
838 the device specifier is required for v6 link-local, e.g. -c
839 [v6addr]%<dev> -V, to select the output interface.
840 Reverse, full-duplex, dualtest (-d) and tradeoff (-r): The --reverse
842 (-R) and --full-duplex options can be confusing when compared to the
843 older options of --dualtest (-d) and --tradeoff (-r). The newer options
844 of --reverse and --full-duplex only open one socket and read and write
845 to the same socket descriptor, i.e. use the socket in full duplex mode.
846 The older -d and -r open second sockets in the opposite direction and
847 do not use a socket in full duplex mode. Note that full duplex applies
848 to the socket and not to the network devices and that full duplex sock‐
849 ets are supported by the operating systems regardless if an underlying
850 network supports full duplex transmission and reception. It's sug‐
851 gested to use --reverse if you want to test through a NAT firewall (or
852 -R on non-windows systems). This applies role reversal of the test af‐
853 ter opening the full duplex socket. (Note: Firewall piercing may be
854 required to use -d and -r if a NAT gateway is in the path.)
855 Also, the --reverse -b <rate> setting behaves differently for TCP and
857 UDP. For TCP it will rate limit the read side, i.e. the iperf client
858 (role reversed to act as a server) reading from the full duplex socket.
859 This will in turn flow control the reverse traffic per standard TCP
860 congestion control. The --reverse -b <rate> will be applied on transmit
861 (i.e. the server role reversed to act as a client) for UDP since there
862 is no flow control with UDP. There is no option to directly rate limit
863 the writes with TCP testing when using --reverse.
864 Bounceback The bounceback test allows one to measure network respon‐
866 siveness (which, in this test, is an inverse of latency.) Latency is
867 merely delay in units of time. Latency metrics require one to know the
868 delay of what's being measured. For bounceback it's a client write to
869 a server read followed by a server write and then the client read. The
870 original write is bounce backed. Iperf 2 sets up the socket with
871 TCP_NODELAY and possibly TCP_QUICKACK (unless disabled). The client
872 sends a small write (which defaults to 100 bytes unless -l is set) and
873 issues a read waiting for the "bounceback" from the server. The server
874 waits for a read and then optionally delays before sending the payload
875 back. This repeats until the traffic ends.
876 The TCP_QUICKACK socket option will be enabled during bounceback tests
878 when the bounceback-hold is set to a non-zero value. The socket option
879 is applied after every read() on the server and before the hold delay
880 call. It's also applied on the client. Use --bounceback-no-quickack to
881 disable TCP ack delays per the socket.
882 TCP Connect times: The TCP connect time (or three way handshake) can be
884 seen on the iperf client when the -e (--enhanced) option is set. Look
885 for the ct=<value> in the connected message, e.g.in '[ 3] local
886 192.168.1.4 port 48736 connected with 192.168.1.1 port 5001 (ct=1.84
887 ms)' shows the 3WHS took 1.84 milliseconds.
888 Port-range Port ranges are supported using the hyphen notation, e.g.
890 6001-6009. This will cause multiple threads, one per port, on either
891 the listener/server or the client. The user needs to take care that the
892 ports in the port range are available and not already in use per the
893 operating system. The -P is supported on the client and will apply to
894 each destination port within the port range. Finally, this can be used
895 for a workaround for Windows UDP and -P > 1 as Windows doesn't dispatch
896 UDP per a server's connect and the quintuple.
897 Packet per second (pps) calculation The packets per second calculation
899 is done as a derivative, i.e. number of packets divided by time. The
900 time is taken from the previous last packet to the current last packet.
901 It is not the sample interval time. The last packet can land at differ‐
902 ent times within an interval. This means that pps does not have to
903 match rx bytes divided by the sample interval. Also, with --trip-times
904 set, the packet time on receive is set by the sender's write time so
905 pps indicates the end to end pps with --trip-times. The RX pps calcula‐
906 tion is receive side only when -e is set and --trip-times is not set.
907 Little's Law in queuing theory is a theorem that determines the average
909 number of items (L) in a stationary queuing system based on the average
910 waiting time (W) of an item within a system and the average number of
911 items arriving at the system per unit of time (lambda). Mathematically,
912 it's L = lambda * W. As used here, the units are bytes. The arrival
913 rate is taken from the writes.
914 Network power: The network power (NetPwr) metric is experimental. It's
916 a convenience function defined as throughput/delay. For TCP transmits,
917 the delay is the sampled RTT times. For TCP receives, the delay is the
918 write to read latency. For UDP the delay is the end/end latency.
919 Don't confuse this with the physics definition of power (delta en‐
920 ergy/delta time) but more of a measure of a desirable property divided
921 by an undesirable property. Also note, one must use -i interval with
922 TCP to get this as that's what sets the RTT sampling rate. The metric
923 is scaled to assist with human readability.
924 Multicast: Iperf 2 supports multicast with a couple of caveats. First,
926 multicast streams cannot take advantage of the -P option. The server
927 will serialize multicast streams. Also, it's highly encouraged to use a
928 -t on a server that will be used for multicast clients. That is because
929 the single end of traffic packet sent from client to server may get
930 lost and there are no redundant end of traffic packets. Setting -t on
931 the server will kill the server thread in the event this packet is in‐
932 deed lost.
933 TCP_QUICACK: The TCP_QUICKACK socket applied after every read() on the
935 server
936 Fast Sampling: Use ./configure --enable-fastsampling and then compile
938 from source to enable four digit (e.g. 1.0000) precision in reports'
939 timestamps. Useful for sub-millisecond sampling.
940
942 Use ./configure --enable-thread-debug and then compile from source to
943 enable both asserts and advanced debugging of the tool itself.
944
946 See https://sourceforge.net/p/iperf2/tickets/
947
949 Iperf2, based from iperf (originally written by Mark Gates and Alex
950 Warshavsky), has a goal of maintenance with some feature enhancement.
951 Other contributions from Ajay Tirumala, Jim Ferguson, Jon Dugan <jdugan
952 at x1024 dot net>, Feng Qin, Kevin Gibbs, John Estabrook <jestabro at
953 ncsa.uiuc.edu>, Andrew Gallatin <gallatin at gmail.com>, Stephen Hem‐
954 minger <shemminger at linux-foundation.org>, Tim Auckland <tim.auckland
955 at gmail.com>, Robert J. McMahon <rjmcmahon at rjmcmahon.com>
956
958 accept(2),bind(2),close(2),connect(2),fcntl(2),getpeername(2),getsock‐
959 name(2),getsockopt(2),listen(2),read(2),recv(2),select(2),send(2),set‐
960 sockopt(2),shutdown(2),write(2),ip(7),socket(7),tcp(7),udp(7)
961 Source code at http://sourceforge.net/projects/iperf2/
963 "Unix Network Programming, Volume 1: The Sockets Networking API (3rd
965 Edition) 3rd Edition" by W. Richard Stevens (Author), Bill Fenner (Au‐
966 thor), Andrew M. Rudoff (Author)
967NLANR/DAST January 2022 IPERF(1)