AuDy: Automatic Dynamic Least-Weight Balancing for Stream Workloads Scalability

Previews research works focus in the separation of data processing issues, heavy database queries and stream data processing over high-data-rate (CEP). In many contexts high rate data streams and database query processing work together, for information correlation, data merging, aggregation with big quantities of previous data. In our work we mention these as Stream-DB workloads. A solution for any database engine is to parallelize the load though many machines or cores. However nodes can still overload. We propose an automated method for scalability and balancing of Stream-DB workloads, called AuDy. This mechanism integrates overload (re)scheduling, automated elasticity, shedding policies, admission control and alerts when more resources are required, to provide total balanced operations avoiding overload problems.