Machine learning systems and methods for data placement in distributed storage

A distributed system comprising:

• a plurality of databases each storing data records, wherein each database is physically located in a corresponding geographic location;
• a plurality of mapper servers;
• and
• at least one cartography server,
• wherein the data store, mapper servers and cartography server are communicably connected via a distributed data network,
• wherein the cartography server is configured to generate a machine learning model by:
• determining a set of training feature vectors, each training feature vector being derived from prior metadata values a prior request associated with the data records;
• training the machine learning model using the set of training feature vectors, each training feature vector labelled with a corresponding target primary storage database;
• and
• transmitting copies of the machine learning model to the plurality of mapper servers,
• wherein each one of the plurality of mapper servers is configured to determine a primary storage database for a data record, from the plurality of databases, by:
• receiving a copy of the machine learning model from the cartography server;
• receiving a further request associated with one of the data records, the further request including a plurality of metadata values;
• processing the metadata values to determine a prediction feature vector corresponding to the further request;
• and
• executing the machine learning model using the prediction feature vector as input to identify one database of the plurality of data stores as the primary storage database for the one of the data records associated with the location request.