Streaming ETL in Polystore Era

In today’s digital environment, businesses have to access, store and analyze in a real time fashion vast amounts of data issued from streaming graph-structure data sources. To meet these requirements, companies owning the data warehouse (\(\mathcal {DW}\)) technology have to combine hardware and software solutions to reduce the time latency between a \(\mathcal {DW}\) and its data sources. The explosion of advanced hardware deployment platforms such as polystore represents an opportunity as pointed in recent studies. But, deploying a graph-structure \(\mathcal {DW}\) over a polystore is not a simple task, since it requires two important phases which are data partitioning and allocation. We claim that these phases have to be connected to the ETL (Extract, Transform, Load) phase, especially its loading process. This connection questions the initial schedule of ETL and deployment processes. In this paper, we present a new approach that connects ETL and deployment processes and challenges their traditional scheduling to meet real time analysis requirements.