ABSTRACT
This paper introduces Hierarchical Stream Functions (HSF), a novel programming model for nested stream aggregation, enabling multi-granularity stateful processing. HSF allows developers to specify aggregation hierarchies, reducing state storage requirements by up to 60% and maintaining sub-100ms latency for deep hierarchies.
PAPER · PDF
Loading PDF…
Key findings
HSF reduces state storage requirements by up to 60% compared to independent window computations.
HSF maintains sub-100ms latency even for deep hierarchies of 5+ levels.
Formalizes hierarchical stream aggregation semantics and presents efficient incremental evaluation algorithms.
Limitations & open questions
Challenges include watermark propagation across levels, handling out-of-order events, and dynamic hierarchy restructuring at runtime.