Our goal is to create the world's fastest extendable, non-transactional time series database for big data (you know, for kids)!
Log file indexing is our initial focus. For example append only ASCII files produced by libraries like Log4J, or containing FIX messages or JSON objects.
Occursions was built by a small team sick of creating hacks to remotely copy and/or grep through tons of large log files. We use it to index around a terabyte of new log data per day. You can use it too. Who doesn't have `just too many' log files?
Occursions asynchronously tails log files and indexes the individual lines in each log file as each line is written to disk so you don't even have to wait for a second after an event happens to search for it.
Occursions uses custom disk backed data structures to create and search its indexes so it is very efficient at using CPU, memory and disk.
You can extend Occursions with shared libraries to support your own file formats, even binary file formats!
- The B*Tree implementation used by Occursions is interesting because it is based on the property that trees that are created in sorted order do not need rebalancing