Lx documentation has been synchronized with Poly 2.1.0.

General description of lx and a list of command line options are available elsewhere.

Measurement units

Unless stated otherwise, the following units are used:

Measurement	Unit
time	milliseconds
size	bytes
ratio	percents

See specific object descriptions for exceptions.

Histograms

A typical histogram output looks like this:

# bin   min   max  count     %   acc%
    1     0     1 229934  4.10   4.10
    3     2     2 745027 13.28  17.38
    4     3     3 745050 13.28  30.66
    5     4     4 711251 12.68  43.34
    6     5     5 616906 11.00  54.34
···
   23    22    26  56137  1.00  96.14
   28    27    32  50477  0.90  97.04
   34    33    43  56317  1.00  98.05
   45    44    69  54574  0.97  99.02
   71    70  9663  54912  0.98 100.00

The first line is a histogram header prepended by a ``comment'' character (#). Column meanings:

1. bin: histogram bin number
2. min: minimum value in the bin
3. max: maximum value in the bin
4. count: number of values in the bin
5. %: percentage of number of values in the bin to the total number of values in the histogram
6. acc%: accumulated percentage (the sum of percentage fields of the current bin and the bins above)

Note that min and max columns represent bin boundaries rather than actual values seen during the run. In other words, a bin has not necessarily seen a value equal to min or max.

Because of the comment character in the header, the header line will be automatically ignored by graphing tools like gnuplot. This kludge makes it easier to plot histograms without extra pre-processing.

``X versus Y'' Tables

Many interesting stats are depicted in the ``X vs. Y'' tables. The conn_ttl_vs_use table is one of them:

#uses    response_time
#min max count mean     5% 10% 15% 20% 25% ... 95% 100%
  1   1 335719  1465.53 20  33  43  51 ... 5007 10463
  2   2 196298  2939.19 80 103 121 140 ... 7807 16831
...
 32  63 332193 67664.30 39935 44799 48383 51327 54015 ... 100095 153599
 64 127   7223 94856.94 71167 76287 79615 82431 84991 ... 119551 147455

The names and meaning of the first few columns will change from object to object. For example, min and max columns in conn_ttl_vs_use table specify the number of connection ``uses'' (number of transactions per connection). Most of the columns on the right from ``Y'' specs have the same interpretation regardless of the ``X'' type:

1. count: number of values that correspond to ``Y'' specs on the right (e.g., number of connections that carried a given number of xactions)
2. mean: mean value of the corresponding ``X'' objects (e.g., mean connections duration that carried two xactions)
3. n%: n-th percentile of the corresponding measurement of ``X'' (e.g., 50-th percentile of connection duration for connections that carried two xactions)

Why does the output look ugly?

The output format is intended mostly for automated processing and quick checks by humans.