Watermarks: Time and progress in streaming dataflow and beyond
Watermarks are a system for measuring progress and completeness in out-of-order stream processing systems and are used to emit correct results in a timely way. Given the trend toward out-of-order processing in current streaming systems, understanding watermarks is an increasingly important skill. Slava Chernyak explains watermarks and demonstrates how to apply them using real-world cases.
| Talk Title | Watermarks: Time and progress in streaming dataflow and beyond |
| Speakers | Slava Chernyak (Google) |
| Conference | Strata + Hadoop World |
| Conf Tag | Making Data Work |
| Location | London, United Kingdom |
| Date | June 1-3, 2016 |
| URL | Talk Page |
| Slides | Talk Slides |
| Video | |
Moving from batch to streaming involves changing how we think about time. Streaming data is neither bounded nor typically well ordered in time. However, to make streaming systems useful and deliver on the promise of low-latency results, we often want to know when we have all the data relevant to emitting a correct aggregation. Watermarks provide the foundation for making such decisions, enabling streaming systems to emit timely, correct results when processing out-of-order data. Given the trend toward out-of-order processing in existing streaming systems, understanding watermarks is an increasingly important skill when designing pipelines. This methodology was first discussed in the MillWheel paper and further explored in the Dataflow Model paper, but this approach is not limited to Google’s stream processing efforts. Rather, it is a general problem that must be addressed by any system that wishes to provide timely out-of-order distributed stream processing; solutions have since been pursued by others, including Flink and Qubit (which built a watermark tracking system on top of Spark Streaming for their own internal use). Drawing on his experience developing and using watermarks at Google, Slava Chernyak discusses the details of how watermarks are applied, explains what their strengths and limitations are, and explores real-world use cases. Slava also hints at some of the implementation challenges for computing watermarks with low latency in a highly distributed system. This should provide a practical set of tools for understanding watermarks and time in out-of-order stream processing pipelines.
