stats-rrdb is a high-performance server for collecting, aggregating and storing various statistics. It's a combination of the famous statsd and graphite without the UI from the later.

stats-rrdb provides simple out-of-the box configuration; query language (TCP connections); optimized update language (UDP connections), support for current and historical data; and, of course, high-performance of C++ code.

stats-rrdb is focused on data and data only. It does not provide any graphical interface (though there are examples of how to build one quickly). However, it has a lot of advanced DB features to make it run fast and efficient:

• TCP and UDP servers for queries returning data and fast updates
• Data blocks cache: only frequently accessed data blocks are loaded in memory
• Open file handles cache: set open files limit high in the OS and save time on re-opening files all the time
• Journal file: never lose your data
• Easy integration with your favorite charting package (Google Charts example)

Well, fair question.

First, I really dislike the architecture for statsd and graphite for two main reasons:

• The data storage is split between two systems (statsd and graphite's whisper)
• The graphite engine combines data storage and presentation layers in one package
• The graphite architecture is overly complicated for the task

Second, I don't want yet another "framework" in the production environment. I like Python as much as the next guy, but if you run XXX (insert your favorite non-Python here), then why do you need to install and maintain Python environment? And even if you install Python, then don't forget about packages compatibilities (while installing graphite, I was presented with a choice of either keeping yum or installing graphite... I decided to go with yum).

Last but not the least, it is fun to write server side software :)

For installation instructions see Install.

• 1.x.0 - query language joins to query multiple metrics in the same time