1NUTTCP(8) Under Construction NUTTCP(8)
2
3
4
6 nuttcp - network performance measurement tool
7
9 nuttcp -h
10 nuttcp -V
11 nuttcp -t [ -bdDsuv ] [ -cdscp_value ] [ -lbuffer_len ] [ -nnum_bufs ]
12 [ -wwindow_size ] [ -wsserver_window ] [ -wb ]
13 [ -pdata_port ] [ -Pcontrol_port ]
14 [ -Nnum_streams ] [ -Rxmit_rate_limit [m|g] ]
15 [ -Txmit_timeout [m] ] host [ < input ]
16 nuttcp -r [ -bBdsuv ] [ -cdscp_value ] [ -lbuffer_len ] [ -nnum_bufs ]
17 [ -wwindow_size ] [ -wsserver_window ] [ -wb ]
18 [ -pdata_port ] [ -Pcontrol_port ]
19 [ -Nnum_streams ] [ -Rxmit_rate_limit [m|g] ]
20 [ -Txmit_timeout [m] ] [ host ] [ > output ]
21 nuttcp -S [ -Pcontrol_port ]
22 nuttcp -1 [ -Pcontrol_port ]
23
25 nuttcp is a network performance measurement tool intended for use by
26 network and system managers. Its most basic usage is to determine the
27 raw TCP (or UDP) network layer throughput by transferring memory buf‐
28 fers from a source system across an interconnecting network to a desti‐
29 nation system, either transferring data for a specified time interval,
30 or alternatively transferring a specified number of buffers. In addi‐
31 tion to reporting the achieved network throughput in Mbps, nuttcp also
32 provides additional useful information related to the data transfer
33 such as user, system, and wall-clock time, transmitter and receiver CPU
34 utilization, and loss percentage (for UDP transfers).
35
36 nuttcp is based on nttcp, which in turn was an enhancement by someone
37 at Silicon Graphics (SGI) on the original ttcp, which was written by
38 Mike Muuss at BRL sometime before December 1984, to compare the perfor‐
39 mance of TCP stacks by U.C. Berkeley and BBN to help DARPA decide which
40 version to place in the first BSD Unix release. nuttcp has several
41 useful features beyond those of the basic ttcp/nttcp, such as a server
42 mode, rate limiting, multiple parallel streams, and timer based usage.
43 nuttcp is also continuing to evolve to meet new requirements that arise
44 and to add desired new features. nuttcp has been successfully tested
45 and used on a variety of Solaris, SGI, and PPC/X86 Linux systems, and
46 should probably work fine on most flavors of Unix.
47
48 There are two basic modes of operation for nuttcp. The original or
49 classic mode is the transmitter/receiver mode, which is also the way
50 the original ttcp and nttcp worked. In this mode, a receiver is first
51 initiated on the destination host using "nuttcp -r", and then a trans‐
52 mitter must be started on the source host using "nuttcp -t". This mode
53 is somewhat deprecated and is no longer recommended for general use.
54 The preferred and recommended mode of operation for nuttcp is the new
55 client/server mode. With this mode, a server is first started on one
56 system using "nuttcp -S" (or "nuttcp -1"), and then a client may either
57 transmit data to (using "nuttcp -t") or receive data from (using
58 "nuttcp -r") the server system. All the information provided by nuttcp
59 is reported by the client, including the information from the server,
60 thus providing a full snapshot of both the transmitter and receiver
61 ends of the data transfer.
62
63 The server may be started by a normal, non-privileged user by issuing
64 either a "nuttcp -S" or a "nuttcp -1" command. However, the optimal
65 and recommended method of running a server is to run "nuttcp -S" via
66 the inetd/xinetd daemon. This method has several significant advan‐
67 tages, including being more robust, allowing multiple simultaneous con‐
68 nections, and providing for access control over who is allowed to use
69 the nuttcp server via the hosts.allow (and hosts.deny) file. By
70 default, the nuttcp server listens for commands on port 5000, and the
71 actual nuttcp data transfers take place on port 5001.
72
73 The host parameter must be specified for the transmitter, and provides
74 the host name or IP address of the receiver. In classic transmit‐
75 ter/receiver mode, the host parameter may not be specified for the
76 receiver. In client/server mode, when the client is the receiver, the
77 host parameter specifies the host name or IP address of the transmitter
78 (server).
79
80 Normally, a nuttcp data transfer is memory-to-memory. However, by
81 using the "-s" option, it is possible to also perform memory-to-disk,
82 disk-to-memory, and disk-to-disk data transfers. Using the "-s" option
83 with the transmitter will cause nuttcp to read its data from the stan‐
84 dard input instead of using a prefabricated memory buffer, while using
85 the "-s" option on the receiver causes nuttcp to write its data to
86 standard output. All these types of data transfers are possible with
87 the classic transmitter/receiver mode. For security reasons, the "-s"
88 option is disabled on the server, so it is not possible to access the
89 disk on the server side of a data transfer.
90
91 The allowed options to nuttcp are:
92
94 -h Print out a usage statement. Running nuttcp with no arguments
95 will also produce a usage statement.
96
97 -V Prints the nuttcp version number. The nuttcp version is also
98 printed as part of the normal nuttcp output when the "-v" ver‐
99 bose output is used (but not when using the default brief out‐
100 put). In client/server mode, the version number of both the
101 client and server is identified.
102
103 -t Indicates that this nuttcp is the transmitter. In client/server
104 mode, this means the server specified by the host parameter is
105 the receiver.
106
107 -r Indicates that this nuttcp is the receiver. In client/server
108 mode, this means the server specified by the host parameter is
109 the transmitter.
110
111 -S Indicates that this nuttcp is the server. The only option that
112 may be specified to the server is the "-P" option, which allows
113 one to change the control port used by the server, but only when
114 the server is started by a normal, non-privileged user. When
115 the server is initiated by inetd/xinetd, the server control port
116 should be specified in the services file.
117
118 -1 Basically the same as the "-S" option, but indicates a one-shot
119 server, i.e. the server exits after the first data transfer ini‐
120 tiated by a client. The "-1" option should only be used when
121 the server is started by a normal, non-privileged user. This
122 option will probably rarely need to be used, but can be useful
123 for a quick test and eliminates the possibilty of leaving a non-
124 access controlled nuttcp server running on the system (which can
125 happen when using the "-S" option and forgetting to kill the
126 nuttcp server after finishing a series of tests).
127
128 -b Produce brief one-line output, which includes the amount of data
129 transferred in MB (1024**2 bytes), the time interval in seconds,
130 the TCP (or UDP) network throughput in Mbps (millions of bits
131 per second), the transmitter and/or receiver CPU utilization,
132 and for UDP data transfers also outputs the loss percentage. In
133 client/server mode, most of the printed statistics are from the
134 viewpoint of the receiver. This is the default output format.
135
136 -B This option is only valid for the receiver, and forces the
137 receiver to read a full buffer (as specified by the "-l" buffer
138 length option) from the network. It is mainly intended to be
139 used with the "-s" option to only output full buffers to stan‐
140 dard output (e.g. for use with tar). It is also implicitly set
141 whenever the number of streams as specified by the "-N" option
142 is greater than 1. This option is not passed to the server.
143
144 -d For TCP data transfers, sets the SO_DEBUG option on the data
145 socket. This option is not passed to the server. It is a
146 rarely used option which may possibly be removed or renamed in a
147 future version of nuttcp.
148
149 -D This option is only valid for the transmitter, and only for TCP
150 data transfers, in which case it sets the TCP_NODELAY option on
151 the data socket, which turns off the Nagle algorithm causing
152 data packets to be sent as soon as possible without introducing
153 any unnecessary delays. This option is not passed to the
154 server. It is a rarely used option which may possibly be
155 removed or renamed in a future version of nuttcp.
156
157 -s Setting the "-s" option causes nuttcp to either read its data
158 from standard input rather than using prefabricated memory buf‐
159 fers (for "nuttcp -t"), or to write its data out to standard
160 output (for "nuttcp -r"). The "-s" option is disabled for secu‐
161 rity reasons on the server.
162
163 -u Use UDP for the data transfer instead of the default of TCP.
164
165 -v Verbose output that provides some additional information related
166 to the data transfer. In client/server mode, the server is
167 always verbose (implicit "-v" option), but the client controls
168 the extent and type of output via the "-v" and "-b" options.
169
170 -cdscp_value
171 Sets the socket option to support COS. Either takes a dscp
172 value or with the t|T modifier it takes the full TOS field.
173
174 -lbuffer_len
175 Length of the network write/read buffer in bytes for the trans‐
176 mitter/receiver. It defaults to 64 KB (65536) for TCP data
177 transfers and to 8 KB (8192) for UDP. For client/server mode,
178 it sets both the client and server buffer lengths.
179
180 -nnum_bufs
181 Specifies the number of source buffers written to the network
182 (default is unlimited), and is ignored by the receiver. For
183 client/server mode, if the client issues a "nuttcp -r" command
184 making it the receiver, this parameter is passed to the server
185 since the server is the transmitter in this case. This parame‐
186 ter is also ignored if the "-s" parameter is specified to the
187 transmitter.
188
189 -wwindow_size
190 Indicates the window size in KB of the transmitter (for "nuttcp
191 -t") or receiver (for "nuttcp -r"). Actually, to be technically
192 correct, it sets the sender or receiver TCP socket buffer size,
193 which then effectively sets the window size. For client/server
194 mode, both the transmitter and receiver window sizes are set.
195 The default window size is architecture and system dependent.
196 Note recent Linux systems, out of a misguided desire to be help‐
197 ful, double whatever window size is actually specified by the
198 user (this is not a bug with nuttcp but a bug/feature of the
199 Linux kernel). Also, with these Linux systems, the actual win‐
200 dow size that's used on the intervening network, as observed
201 with tcpdump, is greater than the requested window size, but
202 less than the doubled value set by Linux.
203
204 -wsserver_window
205 For client/server mode, the "-ws" option provides a mechanism
206 for setting a different window size on the server than the
207 client window size as specified with the "-w" option.
208
209 -wb Normally, to conserve memory, the transmitter only sets the TCP
210 send socket buffer size and the receiver only sets the TCP
211 receive socket buffer size. However, if the "-wb" option is
212 used, the transmitter will also set the TCP receive socket buf‐
213 fer size and the receiver will also set the TCP send socket buf‐
214 fer size. Under normal circumstances, this should never be nec‐
215 essary. This option was implemented because certain early
216 braindead Solaris 2.8 systems would not properly set the TCP
217 window size unless both the TCP send and receive socket buffer
218 sizes were set (later Solaris 2.8 systems have corrected this
219 deficiency). Thus the 'b' in this option can stand either for
220 "braindead" or "both".
221
222 -pdata_port
223 Port number used for the data connection, which defaults to port
224 5001. If multiple streams are specified with the "-N" option,
225 the "-p" option specifies the starting port number for the data
226 connection. For example, if four streams are specified using
227 the default data connection port number, nuttcp will use ports
228 5001, 5002, 5003, and 5004 for the four TCP (or UDP) connections
229 used to perform the data transfer.
230
231 -Pcontrol_port
232 For client/server mode, specifies the port number used for the
233 control connection between the client and server, and defaults
234 to port 5000. On the server side, the "-P" option should only
235 be used when the server is started manually by the user. If the
236 server is started by inetd/xinetd (the preferred method), the
237 control connection must be specified by adding a nuttcp entry to
238 the services file.
239
240 -Nnum_streams
241 Species the number of parallel TCP (or UDP) data streams to be
242 used for the data transfer, with the default being a single data
243 stream. The maximum number of parallel data streams that can be
244 used is 128. If the number of streams is greater than one, the
245 "-B" option is implicitly set. The current implementation does
246 not fork off separate processes for each data stream, so speci‐
247 fying multiple streams on an SMP machine will not take advantage
248 of its multiple processors. Of course it is always possible to
249 run multiple nuttcp commands in parallel on an SMP system to
250 determine if there is any advantage to running on multiple pro‐
251 cessors. This is especially simple to do when running in
252 client/server mode when the server is started from the
253 inetd/xinetd daemon. When running multiple nuttcp commands in
254 parallel, the "-T" transmitter timeout option may be used to
255 insure that all the nuttcp commands finish at approximately the
256 same time.
257
258 -Rxmit_rate_limit[m|g]
259 The transmitter rate limit throttles the speed at which the
260 transmitter sends data to the network, and by default is in
261 Kbps, although the 'm' or 'g' suffix may be used to specify Mbps
262 or Gbps. This option may be used with either TCP or UDP data
263 streams. Because of the way this option is currently imple‐
264 mented, it will consume all the available CPU on the transmitter
265 side of the connection so the "%TX" stats are not meaningful
266 when using the rate limit option. By default the rate limit is
267 applied to the average rate of the data transfer in real time,
268 and not in CPU time, so if nuttcp is switched out of the proces‐
269 sor for any reason, when it is switched back in, it is possible
270 that the instantaneous rate may momentarily exceed the specified
271 value. There is an 'i' qualifier to the rate limit option
272 (specified as "-Ri") that will restrict the instantaneous rate
273 at any given point in time to the specified value, although in
274 this case the final rate reported by nuttcp may be less than the
275 specified value since nuttcp won't attempt to catch up if other
276 processes gain control of the CPU. The default is no rate
277 limit. Note another way to throttle the throughput of TCP data
278 streams is to reduce the window size.
279
280 -Txmit_time_limit[m]
281 Limits the amount of time that the transmitter will send data to
282 the specified number of seconds, or number of minutes if the 'm'
283 suffix is used. Normally a data transfer will either specify a
284 fixed amount of data to send using the "-n" option, or a fixed
285 period of time to send using the "-T" option. However, if both
286 the "-n" and "-T" options are used, the data transfer will be
287 stopped by whichever option takes affect first. The default is
288 a 10 second time limit for the data transfer.
289
291 Under Construction
292
293 For now, consult the README file for basic usage guidelines.
294
296 Under Construction
297
298 For now, see the examples.txt file for some examples of using nuttcp.
299
301 ping(8), traceroute(8), tracepath(8), pathchar(8), netstat(1),
302 mtrace(8)
303
305 Developed by Bill Fink based on nttcp which in turn was an enhancement
306 of the original ttcp developed by Mike Muuss at BRL. IPv6 capability
307 and some other fixes and enhancements contributed by Rob Scott. Many
308 useful suggestions and testing performed by Phil Dykstra and others.
309
310 The current version is available via anonymous ftp from:
311
312 ftp://ftp.lcp.nrl.navy.mil/pub/nuttcp/
313
314 The authors can be reached at nuttcp@lcp.nrl.navy.mil.
315
317 Please send bug reports to nuttcp-bugs@lcp.nrl.navy.mil.
318
319
320
321nuttcp-5.3.1 6 June 2006 NUTTCP(8)