上QQ阅读APP看书,第一时间看更新
Understanding Kappa architecture
Kappa architecture is simpler than the Lambda pattern as it comprises of the speed and serving layers only. All the computations occur as stream processing and there are no batch recomputations done on the full dataset. Recomputations are only done to support changes and new requirements.
Typically, the incoming real-time data stream is processed in memory and is persisted in a database or HDFS, to support queries, as illustrated in the following figure:
Kappa architecture can be realized using a queueing solution such as Apache Kafka or Kinesis. If the data retention times are bound to several days to weeks then Kafka could also be used to retain the data for the limited period of time.