TTCP(1)

NAME

ttcp - test TCP/UDP connection throughput between two hosts

On this page

SYNOPSIS
DESCRIPTION
OPTIONS
EXAMPLES
SEE ALSO
AUTHOR
COPYRIGHT
BUG REPORTS

SYNOPSIS

ttcp -t [options] _host_ < _in_
ttcp -r [options] > _out_

ttcp is a simple network benchmark that measures the throughput of a TCP (or, with -u, UDP) connection between two hosts. It is run in two cooperating instances: a receiver started first with -t omitted (-r), and a transmitter started with -t that names the receiving host.

In transmit mode (-t) ttcp connects to host on the chosen port and either copies data read from standard input to the network, or, in sink/source mode (-s), sends a generated pattern of -n buffers. In receive mode (-r) ttcp listens on the port, accepts a connection, and either copies the received data to standard output or discards it (-s). When the transfer completes, ttcp prints timing and throughput statistics (bytes processed, CPU seconds, and KB/sec and Kbits/sec rates) to standard error.

Both sides must agree on the port (-p), the protocol (TCP or -u UDP), and the buffer length (-l). The default port is 2000 and the default buffer length is 1024 bytes.

OPTIONS

-t: Transmit mode. Requires a host argument naming the receiver.
-r: Receive mode (the default). Listens for an incoming connection.
-l length: Length in bytes of the buffers written to or read from the network (default 1024).
-n count: Number of buffers to send in source mode (-t -s only, default 1024).
-s: Sink/source mode. When transmitting, source a generated pattern instead of reading standard input; when receiving, sink (discard) all data instead of writing standard output.
-p port: TCP/UDP port number to send to or listen on, 0-65535 (default 2000).
-u: Use UDP instead of TCP.
-B: Receive only full blocks of the size given by -l, accumulating partial network reads (useful when feeding fixed-block consumers such as tar).
-d: Enable the SO_DEBUG socket option on the connection, if supported by the platform.

EXAMPLES

Example 1. Measure raw TCP throughput between two hosts

# on the receiving host (e.g. server.example.com):
ttcp -r -s -l65536

# on the sending host:
ttcp -t -s -l65536 -n4096 server.example.com

The receiver discards everything it gets while the transmitter sources 4096 buffers of 65536 bytes each. Both report the achieved throughput on standard error, giving a quick measure of the network path’s bandwidth without needing a real data set.

Example 2. Transfer an actual file (and verify the pipe)

# on the receiving host:
ttcp -r -p3000 > received.dat

# on the sending host:
ttcp -t -p3000 server.example.com < model.g

Here ttcp is used as a plain copy: a BRL-CAD .g geometry database is piped from the transmitter’s standard input to the receiver’s standard output over port 3000, reproducing the file on the far end.