Kairion helps online shops getting additional revenues from display advertising, product placement and couponing . In order to do that, data generated with every page impression from one of the connected shops have to be exchanged with the shop system. The data provided by Kairion are read out from a huge data volume in a complex procedure. They must be available in the shop systems within a few milliseconds.
Basing on the programming language Erlang, we developed an application that divides a complex data search and data selection in several parallel, coordinated tasks. The application is seamlessly integrated with a document-orientated database, CouchDB. To guarantee a downstream processing of the data, Couch DB cooperates with another document-orientated database: MongoDB. This application can interact asynchronously as well as synchronously with the shop systems. The chosen combination of “state of the art”-technologies and the conversion in an agile process (SCRUM) have lead to response times which are considerably beneath the ambitious peak values.