Hi aitozi, That's unfortunately true, there is currently ongoing effort to rework shared buffer and nfa to improve it. This improvement is scheduled for flink 1.6. You can track it with FLINK-9418 (https://issues.apache.org/jira/browse/FLINK-9418).
Best, Dawid On 24/05/18 08:53, aitozi wrote: > Hi, guys > > I am using cep librayies these days, and i find that every time we receive > an element from upstream, we will query the rocksdb for the nfa state, if it > is changed still need to update the rocksdb, which makes the performance of > the cep is poor? > > cc @Kl0 > > > > -- > Sent from: > http://apache-flink-user-mailing-list-archive.2336050.n4.nabble.com/