iperflexer 0.1.4

Glob

Although I said above that you pipe input to the parseiperf command you can also use a file-glob to pass in file-names to it. It descends into all sub-directories to look for matches so you need to make sure the glob is restrictive enough to not match unexpected files and you also need to surround it with quotation marks (so that the shell’s interface doesn’t expand the glob before passing it to the command). The output for all the files will be concatenated and output to standard output.

parseiperf -g "*.iperf"

Although my example uses a glob you could just pass in a file name.

Units

The --units tells the iperflexer what units the bandwidths should use. One of the reasons for this code was to allow the iperf raw-data to be output in bits per second and then converted later to the more human-readable Megabits or Gigabits per second. The names for the units is meant to match the iperf headers (so to get megabits-per-second use Mbits).

Saves

As noted above, the default behavior is to send output to standard output, concatenating the values for multiple data-files, but if you passed in a file-name-glob you can use the --save option and it will save the output for each input file to a separate file. The output-file-names will be the input-file-names with _parsed.csv appended to them.

VooDoo

The --voodoo flag will tell iperflexer to re-add the individual threads (otherwise it uses the summed-values that iperf provides). In retrospect this should have been the default, but I didn’t have a complete enough understanding of how iperf’s reporting worked when I created the command-line interface. If you are doing a time-series analysis, the --voodoo interval sums will catch reporting intervals that are missing their summation lines and so should be considered better for that case. The intervals aren’t quiet right for re-calculating the total values though, so you should use the --lastlinebandwidth and --lastlinetransfers if you are trying to find the values for the entire session.

Debugging

The --pudb and --pdb options will start the chosen debugger before running the command.

Tee

The --tee option tells the command to send the un-parsed iperf lines to standard error. This way if you use it in the originally intended way (piping to another command) you can still see what happened or you can pipe it to another command (including another instance of the parseiperf command).

Threads

By default, it is assumed that only one thread is being used. If more than one thread was used the number of threads has to be specified by passing in the --threads option. I might change it to make it infer them but I originally chose to do it this way because it seemed safest in the event of incomplete data output.

Maximum

Sometimes when the bandwidth gets really small iperf seems to react by generating an unbelievably large number. When the iperflexer sees a bandwidth that’s larger than what’s set as the --maximum (default is \(10^6\)) it will report the value as 0.

Last Line Bandwidth

If set, this will grab the last sum-line whose interval is bigger than the interval given (which is assumed to be the bandwidth-rate that iperf calculated for the session). This will generally be different from the value you would get if you were to re-add the interval counts, especially if any of the threads doesn’t report back in time for the interval (it looks like iperf drops the value for the interval’s reporting but adds it in once the thread does report back). Since it’s usually the last line you could probably get it with some clever tail and sed combining or some other command-line combination, but this will also convert the units for you and allow you to process a set of files. Suppose you wanted to create a column of the bandwidth summaries for files that start with ‘client’ and end with ‘iperf’ and you want to store the outcome in ‘client_summaries.csv’. You could do it with:

parseiperf -g "client*iperf" -l > client_summaries.csv

Last Line Transfer

As you might suspect, this tries to grab the last reported data-transfer (the transfer for the entire interval). It is currently mutually-exclusive to the last line bandwidth. This should be changed eventually. An example to gather the server-side information (assuming the names are meaningful):

parseiperf -g "server*iperf" -l > server_transfer_summaries.csv

Interval

The expected reporting interval (e.g. every second is -i 1). If the interval found on a line exceeds the expected interval, then it is assumed to be a total summation line and not used by the interval values, but is used by the SumParser.last_line_bandwidth and last_line_transfer.

Tolerance

The amount that the calculated reporting interval is for a line can differ from the expected interval. This was created to prevent float rounding problems, but probably doesn’t need to be set.