The data of an online system is distributed over a number of different physical servers and databases on those servers. The online system shards the data to increase system scalability and replicates the shards to increase data reliability. Features such as sharding and replication in turn increase data storage system complexity. Accordingly, in order to provide simplified data storage semantics to applications that use the online system, the online system provides access to data via asynchronous callback functions that are triggered once data has been written and is available for reading by client code. Thus, applications need only specify code to be used in response to data write events, and can thereafter interact with the online system data using a data-driven paradigm.