Object Oriented Databases High ingest rates
High ingest rates: Inserting new records into most data management systems is usually very slow. It’s often multiple orders of magnitude slower than writing the data directly to disk and very difficult to make completely parallel because of central server architecture.
This is a direct result of organizing data into tables. Since data of the same type ingested from different sources must be organized into the same table, there is a natural bottleneck. Although tables can often be physically distributed, any query on a particular table must be able to access all the data in that table or at least an index to all the data in that table.
When indexes are applied to the table the bottleneck becomes even tighter, since indexes are very difficult to maintain with multiple parallel updaters.
Consider an application whose single data type is a message, but must ingest thousands of messages from thousands of sources simultaneously. Forcing all messages into a single “Message” table provides no value. Instead, it makes much more sense to group the messages, both logically and physically, by their source and/or the time each message was received.