> So it looks most of the performance loss originated from GLDv3. I guess the > next step is to decide whether try to remove those locks or try the > multiple-lower-streams approach. Note that even we remove all the locks, > there is still some performance degrade (about 1%)
Thanks for the very useful data. My personal opinion is that if we can get to < 1% on the receive-side (along with 0% on the send side), we will not have a customer adoption problem. However, if the GLDv3 code has to become incredibly complicated as a result, then we still may want to pursue the other design. -- meem
