Hi Lukas, On Fri, Nov 28, 2014 at 05:19:31PM +0100, Lukas Tribus wrote: > I hope that one day we can move the SSL handshake to dedicated threads, > completely eliminating the event loop blocking and allowing a single > process to forward all the traffic while some parallel threads do all > the heavy SSL handshake lifting.
I'm delighted to read this because that's *exactly* what I'm having in mind while reworking the I/O layer and the scheduler :-) In 1.6-dev the I/O callbacks are not called directly from the poll() loops anymore but exclusively from the cache. That way we'll be able to assign some affinity masks between certain tasks and the threads they're allowed to run on, and for example we could decide that when an SSL handshake is pending, we exclude certain low-latency threads. Well, for now only the I/O layer part was adapted. There are so many other parts to do :-/ Cheers, Willy
