> > > Unfortunately Benno, your (and my) intuition is wrong, as shown by the > > > testing that Abramo and Steve did about 3 weeks ago. Mine too! I'm still not quite believing it. It is almost too good to be true. > > Interesting ... did they produce some simulations / stats / graphs / testing > > tools which are publicy available ? > > (if someone has links to mailarchives / sites handy please post them) > > There are some graphs at http://plugin.org.uk/spmp/ > > > I'm still a bit sceptical about the out-of-process and 3msec latencies, > > but if it turns out to be true, fine then. > > I think we were all supprised, but it did work. I have a request. Could someone please write up a little (tiny) document that can be posted with the graphs, explaining things better? Basically, I want a simple, clear and direct explanation of why it is that the out-of-process model works to produce such good results, as compared with the in-process model. The closest I can come to understanding this was that I remember a comment from someone that because the process switch doesn't happen really often, the extra time it takes does not have such a huge effect on performance as one might conclude, if comparing the process switch time to the much smaller overhead of a function call. Also I would like to see each graph explained a little. (What are they graphs of, exactly?) And overall, what is the X axis? I think it would be worth it to add some documentation to that web page (and make sure to keep it in the archives) because I expect this issue to come up again and again in the future. The results are so counterintuitive to many people, that it will help keep the matter in the clear with everyone. Again, I totally, absolutely agree with the multiple process model on the basis of its architectural strengths, it's just the performance issue that I'm confused about. - Jay Ts
