Revision Processing in a Stream Processing Engine: A High-Level Design

Data stream processing systems have become ubiquitous in academic [1, 2, 5, 6] and commercial [11] sectors, with application areas that include financial services, network traffic analysis, battlefield monitoring and traffic control [3]. The append-only model of streams implies that input data is immutable and therefore always correct. But in practice, streaming data sources often contend with noise (e.g., embedded sensors) or data entry errors (e.g., financial data feeds) resulting in erroneous inputs and therefore, erroneous query results. Many data stream sources (e.g., commercial ticker feeds) issue "revision tuples" (revisions) that amend previously issued tuples (e.g. erroneous share prices). Ideally, any stream processing engine should process revision inputs by generating revision outputs that correct previous query results. We know of no stream processing system that presently has this capability.