Being back to work I had the opportunity to check how aila does behave on real production systems yesterday morning.
And to do so I had to make a few adjustments that took us to tag v0.3.0_Rev163:
- When aila is executed the following options are now defaults:
- '-0, --zero'
- '-pc, --progress-counter'
- '-dc, --dump-cache'
- As such the related command line options names have changed:
- '-0, --zero' is now '-n0, --no-zero'
- '-pc, --progress-counter' is now '-npc, --no-progress'
- '-dc, --dump-cache' is now '-ndc, --no-dump-cache'
- Corrected a problem for very long log entries (over 1024 bytes) by setting the line read buffer to 2048 and updating the cs_uri_query handling section.
- Change the cache dump file name from cache_dump to cache_dump.csv.
- Overall performance was greatly improved by refactoring the process generating the line summary data (to execute the searches over a single line parse instead of doing function calls to strstr) and by updating the process handling the conversion from http to asci characters (removing the strlen function call out of for loop construct to use a variable instead).
- Added "Resource Guid" entry as a type of Guid, because some guid searches return a resource guid which cannot be mapped directly to a computer resource or a software update or any other type of resource.
- Changed the super cache configuration to use adaptive update intervals, so the first 50 updates are run every 500 cache call, the next 50 run every 2,000 cache call, the next 100 are run every 5,000 cache call and every 10,000 cache call after that until the end of the file.
- Fully converged the Win32 and Linux build by removing the platform specific calls to convert inet address to int32 and back. So now all features available on Linux builds are also available on Windows builds.
- Corrected a defect in the string cache that prevented the super cache to be used with the guid instance. This helps speed up cache lookup process by checking a 50 items array (the super cache) where the top hitters are stored. Currently ~10% of cache hits are out of the super caches which greatly reduce the cache traversals.
Builds are uploaded in the build various directories on Google code: