1netperf(1) General Commands Manual netperf(1)
2
6 netperf - a network performance benchmark
7
10 netperf [global options] -- [test specific options]
11
14 Netperf is a benchmark that can be used to measure various aspects of
15 networking performance. Currently, its focus is on bulk data transfer
16 and request/response performance using either TCP or UDP, and the
17 Berkeley Sockets interface. In addition, tests for DLPI, and Unix Do‐
18 main Sockets, tests for IPv6 may be conditionally compiled-in.
19 GLOBAL OPTIONS
22 -4 Use AF_INET (aka IPv4) addressing for the control and possibly
23 data connections.
24 -6 Use AF_INET6 (aka IPv6) addressing for the control and possibly
26 data connections.
27 -a sizespec
29 Alter the send and receive buffer alignments on the local sys‐
30 tem. This defaults to 8 bytes.
31 -A sizespec
33 As -a, but for the remote system.
34 -B brandstr
36 Add brandstr to the output of a test with banners disabled.
37 -c [rate]
39 Request CPU utilization and service demand calculations for the
40 local system. If the optional rate parameter is specified, net‐
41 perf will use that instead of calculating the rate itself.
42 -C [rate]
44 As -c, but for the remote system.
45 -d Increase the quantity of debugging output displayed during a
47 test (possibly at the expense of performance).
48 -D [secs,units] (*)
50 Display interim results at least every secs seconds uning units
51 as the initial guess for units per second. This is only avail‐
52 able when netperf has been configured with --enable-demo.
53 -f GMKgmk
55 Change the units of measure for *_STREAM tests. Capital letters
56 are powers of two, lowercase are powers of ten.
57 -F fill_file
59 Pre-fill the send buffers with data from the named file. This is
60 intended to provide a means for avoiding buffers that are filled
61 with data which is trivially easy to compress. A good choice for
62 a file that should be present on any system is this manpage -
63 netperf.man. Other files may be provided as part of the distri‐
64 bution.
65 -h Display a usage string, and exit.
67 -H name|ip,family (*)
69 Set the hostname (or IP address) and address family to use to
70 establish the control connection to the remote system. Passing a
71 single name with no comma will only set remote_host and will
72 leave selection of address family for the control connection to
73 the stack or by a -4 -r -6 command line option.
74 -i max,min
76 Set the maximum and minimum number of iterations when trying to
77 reach certain confidence levels.
78 -j Instruct netperf to calculate additional statistics on timing
80 when running an omni test. Display of said statistics will de‐
81 pend on the presence of the corresponding output selectors in
82 the output selection. These are MIN_LATENCY, MAX_LATENCY,
83 P50_LATENCY, P90_LATENCY, P99_LATENCY, MEAN_LATENCY and STD‐
84 DEV_LATENCY.
85 -I lvl,[,intvl]
87 Specify the confidence level (either 95 or 99 - 99 is the de‐
88 fault) and the width of the confidence interval as a percentage
89 (default 10)
90 -l testlen
92 Specify the length of the test (default 10 seconds). A negative
93 value sets the number of request/response transactions, or the
94 number of bytes for a stream test.
95 -L name|ip,fam (*)
97 Set the local name|IP and/or address family for the socket used
98 for the control connection to the remote netserver.
99 -n numcpus
101 Specify the number of CPU's in the system on those systems for
102 which netperf has no way to find the number of CPU's programati‐
103 cally.
104 -N This option will tell netperf to not establish a control connec‐
106 tion to a remote netserver. Instead it will try to establish a
107 data connection directly, using only the information supplied by
108 the command line parameters and/or internal defaults. Unless
109 other ports are provided by the command line, by default the
110 data connection will be to the "discard" port for a "STREAM" or
111 "SENDFILE" test, the "echo" port for an "RR" test or the "char‐
112 gen" port for a "MAERTS" test.
113 -o sizespec
115 Set an offset from the alignment specified with -a.
116 -O sizespec
118 As -o, but for the remote system.
119 -p portnum,locport (*)
121 Direct the control connection to a netserver listening on the
122 specified port, rather than using a "netperf" entry in /etc/ser‐
123 vices or the internal default (port 12865). If ",locport" is
124 specified the control connection will be established from that
125 local port number. Specifying a single port number with no
126 comma will specify only the remote netserver port number and
127 will leave local port number selection to the stack.
128 -P 0|1 [22mShow (1) or suppress (0) the test banner.
130 -S This option will cause an attempt to set SO_KEEPALIVE on the
132 ends of the data connection for tests using BSD Sockets. It
133 will be made on the netperf side of classic tests, and both net‐
134 perf and netserver side of an omni or migrated test.
135 -s seconds
137 This will cause netperf to sleep "seconds" seconds before trans‐
138 ferring data over the data connection.
139 -t testname
141 Specify the test to perform. Valid testnames include, but are
142 not limited to, nor always compiled-in:
143 TCP_STREAM
144 TCP_SENDFILE
145 TCP_MAERTS
146 TCP_RR
147 TCP_CRR
148 UDP_STREAM
149 UDP_RR
150 DLCO_STREAM
151 DLCO_RR
152 DLCL_STREAM
153 DLCL_RR
154 STREAM_STREAM
155 STREAM_RR
156 DG_STREAM
157 DG_RR
158 SCTP_STREAM
159 SCTP_STREAM_MANY
160 SCTP_RR
161 SCTP_RR_MANY
162 LOC_CPU
163 REM_CPU
164 -T lcpu,remcpu
166 Request that netperf be bound to CPU lcpu and/or netserver be
167 bound to CPU rcpu.
168 -v verbosity
170 Set the verbosity level for the test (only with -P).
171 -V Display the netperf version and exit.
173 TEST SPECIFIC OPTIONS
176 -h Display a usage string based on the test name set with -t, and
177 exit.
178 Please consult the netperf manual Care and Feeding of Netperf
180 2.5.X (doc/netperf.[pdf|html|txt]) for more information. Or you
181 can join and send email to netperf-talk@netperf.org.
182
185 For those options taking two parms, at least one must be specified;
186 specifying one value without a comma will set both parms to that value,
187 specifying a value with a leading comma will set just the second parm,
188 a value with a trailing comma will set just the first. To set each parm
189 to unique values, specify both and separate them with a comma.
190 * For these options taking two parms, specifying one value with no
192 comma will only set the first parm and will leave the second at the de‐
193 fault value. To set the second value it must be preceded with a comma
194 or be a comma-separated pair. This is to retain previous netperf behav‐
195 iour.
196
200 There are bound to be bugs. If you think you have found a bug, please
201 mention it in netperf-talk@netperf.org. List membership is required to
202 send email to the list. See http://www.netperf.org/cgi-bin/mail‐
203 man/listinfo/netperf-talk . If all else fails send email to netperf-
204 feedback@netperf.org.
205
208 netserver(1)
209 Care and Feeding of Netperf 2.5.X
210 http://www.netperf.org/
211
214 HP Information Networks Division - Networking Performance Team.
215 Rick Jones <rick.jones2@hp.com>
216 Karen Choy HP IND
217 Dave Shield <daves@csc.liv.ac.uk> (man pages)
218 Others too numerous to mention here - see the AUTHORS file
219 netperf(1)