Felipe Sateler wirtes at http://bugs.debian.org/751279:
Arbtt log files can become relatively large, and arbtt can take some time to parse it. I have a file containing a bit over 3 years of samples, and it takes a while to process:
% time arbtt-stats -i --logfile capture.log.all <snip> arbtt-stats -i --logfile capture.log.all 20.55s user 0.06s system 100% cpu 20.610 total
Using the right invocations of arbtt-dump, arbtt-import and grep, it is possible to split the log into multiple files, but it is cumbersome.
I'm not sure what exactly is the best way to improve this.
The simplest solution would be to provide a rotate command that (for example) would break the log by year and store each year in a different file, leaving the current year as the default capture file.
A (maybe) better solution would be to convert the logfile into a logdir, where logs are stored by year-month, and then add a flag to arbtt-stats to specify start date, so that older logs are not parsed. This would of course mean more work.