perfSONAR User Q&A and Other Discussion

[Aaron Brown <>]

Hey All,

Does this work better for folks?

Cheers,
Aaron

[aaron@desk146
 bwctl]$ ./bwctl -h
bwctl: 
usage: bwctl [arguments]

Connection Arguments
-4|--ipv4                        Use IPv4 only
-6|--ipv6                        Use IPv6 only
-B|--local_address <address>     Use this as a local address for control 
connection and tests
-c|--receiver <address>          The host that will act as the receiving side 
for a test
-E|--no_endpoint                 Allow tests to occur when the receiver isn't 
running bwctl (Default: False)
-o|--flip                        Have the receiver connect to the sender 
(Default: False)
-s|--sender <address>            The host that will act as the sending side 
for a test

Scheduling Arguments
-a|--allow_ntp_unsync <seconds>  Allow unsynchronized clock - claim good 
within offset
-I|--test_interval <seconds>     Time between repeated bwctl tests
-L|--latest_time <seconds>       Latest time into an interval to allow a test 
to run
-n|--num_tests <num>             Number of tests to perform (Default: 1)
-R|--randomize <percent>         Randomize the start time within this 
percentage of the test's interval (Default: 10%)
--schedule <schedule>            Specify the specific times when a test 
should be run (e.g. --schedule 11:00,13:00,15:00)
--streaming                      Request the next test as soon as the current 
test finishes

Test Arguments
-b|--bandwidth <bandwidth>       Bandwidth to use for tests (bits/sec KM) 
(Default: 1Mb for UDP tests, unlimited for TCP tests)
-D|--dscp <dscp>                 RFC 2474-style DSCP value for TOS byte
-i|--report_interval <seconds>   Tool reporting interval
-l|--buffer_length <bytes>       Length of read/write buffers
-O|--omit <seconds>              Omit time (currently only for iperf3)
-P|--parallel <num>              Number of concurrent connections
-S|--tos <tos>                   Type-Of-Service for outgoing packets
-T|--tool <tool>                 The tool to use for the test
                                   Available Tools:
                                      iperf
                                      iperf3
                                      nuttcp
-t|--test_duration <seconds>     Duration for test (Default: 10)
-u|--udp                         Perform a UDP test
-w|--window <bytes>              TCP window size (Default: system default)
-W|--dynamic_window <bytes>      Dynamic TCP window fallback size (Default: 
system default)
--tester_port <port>             For an endpoint-less test, use this port as 
the server port (Default: tool specific)

Output Arguments
-d|--output_dir <directory>      Directory to save session files to (only if 
-p)
-e|--facility <facility>         Syslog facility to log to
-f|--units <unit>                Type of measurement units to return 
(Default: tool specific)
-p|--print                       Print results filenames to stdout (Default: 
False)
-q|--quiet                       Silent mode (Default: False)
-r|--syslog_to_stderr            Send syslog to stderr (Default: False)
-v|--verbose                     Display verbose output
-x|--both                        Output both sender and receiver results
-y|--format <format>             Output format to use (Default: tool specific)
--parsable                       Set the output format to the machine 
parsable version for the select tool, if available

Misc Arguments
-h|--help                        Display the help message
-V|--version                     Show version number

On Apr 16, 2014, at 7:38 PM, Eli Dart 
<>
 wrote:

> Hi all,
> 
> My opinion only: if groupings make sense, then groupings are fine - they 
> should just be explicit, and options listed alphabetically within the 
> group.  Iperf's man page almost does this - there are sections for options, 
> and the options are alphabetical within the sections (mostly).  I find that 
> to be much more readable than something like the rsync man page which is 
> just a big jumble.
> 
> Again - my opinion only.
> 
> Eli
> 
> 
> 
> On Wed, Apr 16, 2014 at 1:29 PM, Aaron Brown 
> <>
>  wrote:
> Hey Alex,
> 
> They’re roughly grouped by functionality, though I can re-arrange them. In 
> general, I find alphabetical order less useful when I’m reading help 
> options, because I’m usually reading them from a “how do I do X?” 
> perspective as opposed to “what does -x mean?”.
> 
> Cheers,
> Aaron
> 
> On Apr 16, 2014, at 4:23 PM, Alex Moura 
> <>
>  wrote:
> 
>> Hi Aaron,
>> 
>> Just asking (really!): is there any reason for the help options not listed 
>> in aphabetical order?
>> 
>> bwctl: 
>> usage: bwctl [arguments]
>> -c|--receiver <address>          The host that will act as the receiving 
>> side for a test
>> -s|--sender <address>            The host that will act as the sending 
>> side for a test
>> -4|--ipv4                        Use IPv4 only
>> -6|--ipv6                        Use IPv6 only
>> -B|--local_address <address>     Use this as a local address for control 
>> connection and tests
>> -n|--num_tests <num>             Number of tests to perform (default: 1)
>> -I|--test_interval <seconds>     Time between repeated bwctl tests
>> --streaming                      Request the next test as soon as the 
>> current test finishes
>> -L|--latest_time <seconds>       Latest time into an interval to allow a 
>> test to run
>> -R|--randomize <percent>         Randomize the start time within this 
>> percentage of the test's interval (Default: 10%)
>> -T|--tool <tool>                 The tool to use for the test
>>                                    Available Tools:
>>                                       iperf
>>                                       iperf3
>>                                       nuttcp
>> -o|--flip                        Have the receiver connect to the sender 
>> (default: False)
>> -a|--allow_ntp_unsync <seconds>  Allow unsynchronized clock - claim good 
>> within offset
>> -f|--units <unit>                Type of measurement units to return 
>> (Default: tool specific)
>> -x|--both                        Output both sender and receiver results
>> -y|--format <format>             Output format to use (Default: tool 
>> specific)
>> -b|--bandwidth <bandwidth>       Bandwidth to use for tests (bits/sec KM) 
>> (Default: 1Mb for UDP tests, unlimited for TCP tests)
>> -i|--report_interval <seconds>   Tool reporting interval
>> -l|--buffer_length <bytes>       Length of read/write buffers
>> -O|--omit <seconds>              Omit time (currently only for iperf3)
>> -P|--parallel <num>              Number of concurrent connections
>> -D|--dscp <dscp>                 RFC 2474-style DSCP value for TOS byte
>> -S|--tos <tos>                   Type-Of-Service for outgoing packets
>> -t|--test_duration <seconds>     Duration for test (Default: 10)
>> -u|--udp                         Perform a UDP test
>> -w|--window <bytes>              TCP window size (Default: system default)
>> -W|--dynamic_window <bytes>      Dynamic TCP window fallback size 
>> (Default: system default)
>> -v|--verbose                     Display verbose output
>> -p|--print                       Print results filenames to stdout 
>> (Default: False)
>> -d|--output_dir <directory>      Directory to save session files to (only 
>> if -p)
>> -e|--facility <facility>         Syslog facility to log to
>> -q|--quiet                       Silent mode (Default: False)
>> -r|--syslog_to_stderr            Send syslog to stderr (Default: False)
>> -V|--version                     Show version number
>> -h|--help                        Display the help message
>> 
>> Thanks,
>> Alex
>> 
>> 
>> On Fri, Mar 28, 2014 at 10:28 AM, Aaron Brown 
>> <>
>>  wrote:
>> Hey Alan,
>> 
>> There’s a poorly documented (read: not documented) way of doing that for 
>> traceroute (doesn’t work for tracepath, but will when I do another 
>> release), if you specify “-y a”, it’ll skip the hostname checks. 
>> Presumably, I chose ‘a’ for ‘addresses’ or something.
>> 
>> Cheers,
>> Aaron
>> 
>> On Mar 27, 2014, at 1:54 PM, Alan Whinery 
>> <>
>>  wrote:
>> 
>> > One nit, after spending some time messing about with bwctl 1.5.1 --
>> >
>> > The existence of a "-n (number of tests)" arg in bwtraceroute
>> > effectively masks "-n (don't do name resolution)" in tracepath and
>> > traceroute. With many (most?) paths including a hop with some proportion
>> > of hops with no associated reverse DNS record, doing name resolutions
>> > can increase the trace* time-to-complete by several orders of magnitude,
>> > so it's good to be able to send traceroute/tracepath a "-n".  This
>> > creates issues when the extra time stacks up across multiple traces, and
>> > if you really want names at some point you could resolve them from the
>> > collected IP addresses.
>> >
>> > I suggest adding a bwctl switch which tells the tool in question to
>> > not-resolve, not necessarily "-n" itself, like "--no-dns" .
>> >
>> > On 2/18/2014 8:11 AM, Aaron Brown wrote:
>> >> Hi all,
>> >>
>> >> A new version of BWCTL is now available. BWCTL 1.5.1 is a major upgrade 
>> >> over the previously released version.
>> >>
>> >> Notable additions in this release:
>> >> - Support for ping and owping (Run 'bwping -h' to see the available 
>> >> options)
>> >> - Support for traceroute and tracepath (Run 'bwtraceroute -h' to see 
>> >> the available options)
>> >> - Support for iperf3, and improved support for nuttcp
>> >> - Support for working behind a firewall for iperf3, ping, traceroute 
>> >> and tracepath (use the "--flip" flag when the receiver is behind the 
>> >> firewall)
>> >> - Handles NTP unsync a bit better (the daemon start, but will deny 
>> >> tests until NTP is synchronized)
>> >>
>> >> Those who are using the perfSONAR Toolkit, or otherwise have the 
>> >> Internet2 RPM repository enabled, can run 'yum update' to get the 
>> >> latest changes. For those who are installing from source or prefer to 
>> >> install the RPMs manually, the newest version can be downloaded from 
>> >> http://software.internet2.edu/ . After installation, you will need to 
>> >> restart bwctld to use the new features. You can do this by running 
>> >> “/etc/init.d/bwctld restart”.
>> >>
>> >> Please let us know if you have any questions and we look forward to 
>> >> everyone's continued feedback and support!
>> >>
>> >> Thank you,
>> >> The perfSONAR-PS Development Team
>> >
>> 
>> 
> 
> 
> 
> 
> -- 
> Eli Dart, Network Engineer                          NOC: (510) 486-7600
> ESnet Office of the CTO (AS293)                          (800) 333-7638
> Lawrence Berkeley National Laboratory 
> PGP Key fingerprint = C970 F8D3 CFDD 8FFF 5486 343A 2D31 4478 5F82 B2B3