Lecture
This lecture covers the evolution of stream processing systems from centralized in-memory to distributed dataflow systems like MapReduce, Spark Streaming, and Flink. It explains the concept of state in stream processing, including windows, aggregates, and user-defined variables. The lecture also discusses state management issues such as scalability, persistence, and consistency, and explores different approaches to handling state, including using synopses, user-defined and system-managed state. Examples of state manipulation in stream processing systems are provided, highlighting the importance of state in expressing operators and the trade-offs between system-managed and user-defined state.