Unlike cluster stream processing engines (SPE) like Flink and Spark Streaming; single-node, multi-core stream processing engines can process hundreds of millions of tuples per second with fewer resources and lesser maintenance. However, many developers cannot adopt single-node SPE because they lack fault-tolerance mechanisms or off-load persistence to external distributed messaging systems (e.g. Kafka, Kinesis), thereby obviating the benefits of a single-node deployment. In this context, Theodorakis et al. introduce Scabbard: a single-node fault-tolerant SPE that persists all input data streams to an SSD. Scabbard reduces the required disk I/O bandwidth required for persistence through a persistent operator graph model that enables workload-specific optimizations (i.e. about when to discard persisted data) and through query-specific adaptive compression.