Convenience functions for microstate accounting

This module implements some convenience functions for analyzing microstate accounting data. For details about how to use the basic api and what the different states represent see erlang:statistics(microstate_accounting).

Basic Scenario

1> msacc:start(1000).
2> msacc:print().
Average thread real-time    : 1000513 us
Accumulated system run-time :    2213 us
Average scheduler run-time  :    1076 us

        Thread      aux check_io emulator       gc    other     port    sleep

Stats per thread:
     async( 0)    0.00%    0.00%    0.00%    0.00%    0.00%    0.00%  100.00%
     async( 1)    0.00%    0.00%    0.00%    0.00%    0.00%    0.00%  100.00%
       aux( 1)    0.00%    0.00%    0.00%    0.00%    0.00%    0.00%   99.99%
 scheduler( 1)    0.00%    0.03%    0.13%    0.00%    0.01%    0.00%   99.82%
 scheduler( 2)    0.00%    0.00%    0.00%    0.00%    0.03%    0.00%   99.97%

Stats per type:
         async    0.00%    0.00%    0.00%    0.00%    0.00%    0.00%  100.00%
           aux    0.00%    0.00%    0.00%    0.00%    0.00%    0.00%   99.99%
     scheduler    0.00%    0.02%    0.06%    0.00%    0.02%    0.00%   99.89%

This first command enables microstate accounting for 1000 milliseconds. See start/0, stop/0, reset/0 and start/1 for more details. The second command prints the statistics gathered during that time. First three general statistics are printed.

Average real-time
The average time spent collecting data in the threads. This should be close to the time which data was collected.
System run-time
The total run-time of all threads in the system. This is what you get if you call msacc:stats(total_runtime,Stats).
Average scheduler run-time
The average run-time for the schedulers. This is the average amount of time the schedulers did not sleep.

Then one column per state is printed with a the percentage of time this thread spent in the state out of it's own real-time. After the thread specific time, the accumulated time for each type of thread is printed in a similar format.

Since we have the average real-time and the percentage spent in each state we can easily calculate the time spent in each state by multiplying Average thread real-time with Thread state %, i.e. to get the time Scheduler 1 spent in the emulator state we do 1000513us * 0.13% = 1300us.





A map containing the different microstate accounting states and the number of microseconds spent in it.



A map containing information about a specific thread. The percentages in the map can be either run-time or real-time depending on if runtime or realtime was requested from stats/2. system is the percentage of total system time for this specific thread.


A map containing the different microstate accounting states. Each value in the map contains another map with the percentage of time that this thread has spent in the specific state. Both the percentage of system time and the time for that specific thread is part of the map.




The different states that a thread can be in. See erlang:statistics(microstate_accounting) for details.


The different options that can be given to print/2.



This function checks whether microstate accounting is available or not.


Start microstate accounting. Returns whether it was previously enabled or disabled.


Resets all counters and then starts microstate accounting for the given milliseconds.


Stop microstate accounting. Returns whether is was previously enabled or disabled.


Reset microstate accounting counters. Returns whether is was enabled or disabled.


Prints the current microstate accounting to standard out. Same as msacc:print(msacc:stats(),#{}).


Print the given microstate statistics values to stdout. Same as msacc:print(DataOrStats,#{}).


Print the given microstate statistics values to standard out. With many states this can be quite verbose. See the top of this reference manual for a brief description of what the fields mean.

It is possible to print more specific types of statistics by first manipulating the DataOrStats using stats/2. For instance if you want to print the percentage of run-time for each thread you can do:


If you want to only print run-time per thread type you can do:



Print percentage of time spent in each state out of system time as well as thread time. Default: false.


Print the given microstate statistics values to the given file or device. The other arguments behave the same way as for print/2.


Returns a runtime system independent version of the microstate statistics data presented by erlang:statistics(microstate_accounting). All counters have been normalized to be in microsecond resolution.


Returns the system time for the given microstate statistics values. System time is the accumulated time of all threads.

Returns all time recorded for all threads.
Returns all time spent doing work for all threads, i.e. all time not spent in the sleep state.

Returns fractions of real-time or run-time spent in the various threads from the given microstate statistics values.

Returns a list of microstate statistics values where the values for all threads of the same type has been merged.


Dumps the current microstate statistics counters to a file that can be parsed with file:consult/1.


Read a file dump produced by to_file(Filename).