BDB: The Conclusion
We gave up on using Berkeley DB for our metric store. There were just too few advantages over a higher-level solution like MySQL. BDB doesn’t have:
- A notion of schema or data types
- A query language
- Slick integration into Ruby
In the end it would have taken a LOT of manpower to wrap and tune BDB to work more efficiently than MySQL and we would have had a very complex proprietary system to maintain. MySQL simply provides a more standard system which more people understand.
It’s definitely a winner for embedded solutions with simple schema requirements but the implementation complexity grows as O(n^2) with the size of the schema in my estimation.