apache-flink - Apache Flink Stateful Functions 状态缩放

Question

我们能够随心所欲地扩展无状态函数。但是，如果我们在扩展无状态函数时不扩展它们，那么有状态函数可能会成为瓶颈。由于数据在节点之间的分布，缩放状态似乎很棘手。如果有状态函数成为瓶颈，我们也可以扩展它们吗？

Answer 1

Stateful functions are scaled exactly in the same way as stateless.

In StateFun, the remote functions are in fact stateless processes that receive the state just prior to the invocation for a given key, and after a successful invocation the changes to the state are communicated back to the Flink cluster. Therefore scaling stateless or stateful functions is really the same.

To learn more I'd recommend watching a keynote introducing StateFun 2.0 or visiting the distributed architecture page