Skip to content

Latest commit

 

History

History
56 lines (44 loc) · 2.24 KB

README.md

File metadata and controls

56 lines (44 loc) · 2.24 KB
Raw

Performance Logging

This application provides different kinds of statistics about performance in erlang enviroment.

There are diffrerent types of counters:

  • delta -- time samples (deltas) of various operations
  • magnitude
  • size
  • value
  • count -- number of items
  • combo -- size and timing samples (size + delta)
  • combo_count -- number of items and timing samples (count + delta)

Usage example

Consider a situation when you want to get execution time of some function -- let it be json to xml convertation. With plog you could simply write:

Xml = plog:delta(hot_part, json_to_xml, fun () ->
        json_to_xml_converter:convert(Json)
    end).

Now you should turn plog application (e.g. application:start(plog)) and enable category hot_part for activating logging: application:set_env(plog, hot_part, true).

That is all.

Now you could watch the statistics: attach to the node and execute plog:print() or plog:print(delta) to print only delta-statistic. The output would be looks like:

    json_to_xml                                     990epm,20ms@95%,20ms@98%
       ms   eps epm
        1     3 169  [616  878  792  779 |3969814] 300avg
        3     5 308 [1183 1435 1499 1461 |6294807] 300avg
        5     4 220  [792 1063 1157 1142 |4573967] 300avg
       10     4 212  [726 1041 1176 1269 |4850060] 300avg
       20     1  74  [259  357  411  412 |1853697] 300avg
       30     0   6   [21   24   51   35  |214992] 300avg
       50     0   1    [3    1    5    2   |46119] 300avg
       75     0   0    [0    1    2    0   |17593] 300avg
      100     0   0    [0    0    4    0    |8161] 300avg
      200     0   0    [0    0    3    0     |926] 300avg
      300     0   0    [0    0    0    0      |70] 300avg
      500     0   0    [0    0    0    0      |53] 300avg
      750     0   0    [0    0    0    0      |10] 300avg
    total    16 990

It is mean that in the last minute it was about 990 event in total. 169 events per minute took <= 1 ms to execute, 308 events took from 1 to 3 ms to execute and so on. You could also see that 95% (and 98% in particular this case too) of all events took less than 20ms.

Please refer to src/plog.erl to obtain more information.