A distributed and efficient datastore.
RubenSystems released RSGraphite as the new in house datastore. It allows for simple and complex queries with no indexes, reducing the amount of data in RAM and on disks. It is also highly distributed, as it can scale to hundreds of thousands of clusters with ease.
Using an alternate structure to a traditional relational database, RSGraphite is similar to a graph database, using nodes (we call them atoms) and links (electrons). However, it does not have the drawbacks of a graph database such as poor ordering performance. It can also get/store/retrieve trillions of links with ease.
RSGraphite can run matrix similarity searching which comes as a Python extension. This can search for word embeddings, CNN descriptors or even just basic matrixes.
RSGraphite is implemented in C++ and has bindings in Python and Node. Designed for computers with low RAM and disk storage and poor CPU and GPU performance, it can efficiently run on pretty much any machine. We are in the process of developing a custom C++ web server, as it currently runs with the Python Flask library.