On Thu, Apr 11, 2002 at 05:10:51PM -0700, Brian Pane wrote: > Adding the mutex check will fix one of the three problems in worker that > I know of. The other two--the large-grained overhead of forking a new > process to add concurrency, and the mutex contention on the queue--will > require more radical solutions, so they're better explored in a new MPM.
AIUI, these two problems aren't going to directly impact the user in such a drastic way as the accept() when no threads are available. I'm not terribly sure how forking a new process can be worked around - unless we use a spmt model (my favorite, but the group doesn't like it because it isn't "robust"). Perhaps we could do something where a process can dynamically add threads as needed? (worker has a fixed number of threads per process, right?) Mutex contention is a fact of life. I'd be curious to see how we can reduce it, but I don't think we'll be able to remove it entirely. Correct first, fast second. =) -- justin
