The performance loss being referred to there is reduced throughput. There's a blog post by Nico Kruber [1] that covers Flink's network stack in considerable detail. The last section on latency vs throughput gives some more insight on this point. In the experiment reported on there, the difference between a buffer timeout of 1 msec and 0 msec resulted in a significant loss of throughput (roughly a factor of 2x), for very little gain in reduced latency.
[1] https://flink.apache.org/2019/06/05/flink-network-stack.html On Wed, Sep 16, 2020 at 4:20 PM Mazen Ezzeddine < mazen.ezzedd...@etu.unice.fr> wrote: > Hi all, > > I have read the below in the documentation : > > "To maximize throughput, set setBufferTimeout(-1) which will remove the > timeout and buffers will only be flushed when they are full. To minimize > latency, set the timeout to a value close to 0 (for example 5 or 10 ms). A > buffer timeout of 0 should be avoided, because it can cause severe > performance degradation." > > > why a 0 BufferTimeout cause severe performance degradation, shouldnt it > provide min latency, what is meant by perf. degradation there. On the > otherhand, can we say that min latency is always > BufferTimeout. > > Best, > > > > -- > Sent from: > http://apache-flink-user-mailing-list-archive.2336050.n4.nabble.com/ >
