Towards optimising the data flow in distributed applications

Abstract

Networked applications continuously move towards service- based and modular solutions. At the same time, web tech- nologies, proven to be modular and distributed, are applied to these application areas. However, web technologies have to be adapted to the new characteristics of the involved sys- tems - no explicit client and server roles, use of heteroge- neous devices, or high frequency and low latency data com- munication. To this end, we present an approach for describ- ing distributed applications in terms of graphs of communi- cating nodes. In particular, we develop a formal model for capturing the communication between nodes, by including dynamic and static data producing devices, data consuming client applications, as well as devices that can serve as data produces and consumers at the same time. In our model, we characterise nodes by their frequencies of data exchange. We complement our model with a decision algorithm for deter- mining the pull/push communication direction to optimise the amount of redundantly transferred data (i.e., data that is pushed but cannot be processed or data that is pulled but is not yet updated). The presented work lays the foun- dation for creating distributed applications which can auto- matically optimise data exchange.