The service structure of Apache Spark Streaming can improve. There are a lot of issues with memory management and latency. There is no real-time analytics. We recommend it for the use cases where there is a five-second latency, but not for a millisecond, an IOT-based, or the detection anomaly-based. Flink as a service is much better. Apache Spark Streaming does not have auto-tuning. A customer needs to invest a lot, in terms of management and maintenance.
