On Tue, 2018-07-17 at 19:39 +0800, King Silver wrote: > @ Oleg： > I think if read or write event concurrently with a timeout > event,there will be an exception in read or write event
But this is bad. We do not want to shut down a channel only because some operation times out. > .handleIOEvent > method in InternalChannel class can > handle the exception appropriate。 > perhaps what i think is not right,please point it out,and i can > try to > improve the code. Before we can make any sensible decision about this feature I would like to know its impact on performance of the I/O reactor code. While working performance optimization of HttpCore 4.3 I put together a *non-scientific* benchmark that nevertheless measures i/o throughput of HttpCore, Jetty and Netty under similar load scenarios that can serve as a base line for comparative analysis of HttpCore performance. https://github.com/ok2c/httpcore-benchmark If you want to make a major contribution to the project please consider forking the benchmark code, upgrading it to the latest releases of all 3 frameworks and adding HttpCore 5 as the forth implementation. Once there are some numbers we can compare with, we can try running the latest HttpCore with your changes and see what kind of impact that makes on the benchmark results. Oleg --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
