Group-aware Stream Filtering: Towards Collaborative Data Reduction in Stream Processing Systems

In this dissertation, we (the author and her research collaborators) consider a distributed system that disseminates high-volume event streams to many simultaneous monitoring applications over a low-bandwidth network. For bandwidth efficiency, we propose a ``group-aware stream filtering'' approach, used together with multicasting, that exploits two overlooked, yet important, properties of monitoring applications: 1) many of them can tolerate some degree of ``slack'' in their data quality requirements, and 2) there may exist multiple subsets of the source data satisfying the quality needs of an application. We can thus choose the ``best alternative'' subset for each application to maximize the data overlap within the group to best benefit from multicasting. Here we provide a general framework for the group-aware stream filtering problem, which we prove is NP-hard. We introduce a suite of heuristics-based algorithms that ensure data quality (specifically, granularity and timeliness) while preserving bandwidth. Our evaluation shows that group-aware stream filtering is effective in trading CPU time for bandwidth savings, compared with self-interested filtering.